A Survey of Computer Vision Technologies In Urban and

Controlled-environment Agriculture

JIAYUN LUO, Nanyang Technological University, Singapore

BOYANG LI∗ , Nanyang Technological University, Singapore
CYRIL LEUNG, Nanyang Technological University, Singapore
In the evolution of agriculture to its next stage, Agriculture 5.0, artificial intelligence will play a central role. Controlled-environment
arXiv:2210.11318v1 [cs.CV] 20 Oct 2022

agriculture, or CEA, is a special form of urban and suburban agricultural practice that offers numerous economic, environmental, and
social benefits, including shorter transportation routes to population centers, reduced environmental impact, and increased produc-
tivity. Due to its ability to control environmental factors, CEA couples well with computer vision (CV) in the adoption of real-time
monitoring of the plant conditions and autonomous cultivation and harvesting. The objective of this paper is to familiarize CV re-
searchers with agricultural applications and agricultural practitioners with the solutions offered by CV. We identify five major CV
applications in CEA, analyze their requirements and motivation, and survey the state of the art as reflected in 68 technical papers
using deep learning methods. In addition, we discuss five key subareas of computer vision and how they related to these CEA prob-
lems, as well as nine vision-based CEA datasets. We hope the survey will help researchers quickly gain a bird-eye view of the striving
research area and will spark inspiration for new research and development.

CCS Concepts: • Computing methodologies → Computer vision; Neural networks; • Applied computing → Agriculture; •
General and reference → Surveys and overviews.

Additional Key Words and Phrases: agriculture 5.0, controlled-environment agriculture, multimodality, pest and disease detection,
growth monitoring, flower and fruit detection

ACM Reference Format:

Jiayun Luo, Boyang Li, and Cyril Leung. 2018. A Survey of Computer Vision Technologies In Urban and Controlled-environment
Agriculture. In Woodstock ’18: ACM Symposium on Neural Gaze Detection, June 03–05, 2018, Woodstock, NY . ACM, New York, NY, USA,
31 pages. https://doi.org/10.1145/1122445.1122456

1 INTRODUCTION
Artificial intelligence (AI), especially computer vision (CV), is finding an ever broadening range of applications in
modern agriculture. The next stage of agricultural technological development, Agriculture 5.0 [15, 83, 208, 317], will
constitute AI-driven autonomous decision making as a central component. The term Agriculture 5.0 stems from a
chronology [317] that begins with Agriculture 1.0, which heavily depends on human labor and animal power, and
Agriculture 2.0, enabled by synthetic fertilizers, pesticide, and combustion-powered machinery, and develops to Agri-
culture 3.0 and 4.0, characterized by GPS-enabled precision control, and Internet-of-Thing (IoT) driven data collection
∗ The authors can be reached at the following address: 50 Nanyang Avenue, School of Computer Science and Engineering, Nanyang Technological Uni-
versity, Singapore 639798. Boyang Li is the corresponding author. The research is funded by WeBank-NTU Joint Research Center and China-Singapore
International Joint Research Institute.

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not
made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components
of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or
to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from [email protected].
© 2018 Association for Computing Machinery.
Manuscript submitted to ACM

1
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

[228]. Built upon the rich agricultural data collected, Agriculture 5.0 holds the promise to further increase productivity,
satiate the food demand of a growing global population, and mitigate the negative environmental impact of existing
agricultural practices.
As an integral component of Agriculture 5.0, controlled-environment agriculture (CEA), a farming practice carried
out within urban, indoor, resource-controlled, and sensor-driven factories, is particularly suitable for the application of
AI and CV. This is because CEA provides ample infrastructure support for data collection and autonomous execution
of algorithmic decisions. In terms of productivity, CEA could produce higher yield per unit area of land [8, 9] and boost
the nutritional content of agricultural products [139, 274]. In terms of environmental impact, CEA farms can insulate
environmental influences, relieve the need for fertilizer and pesticides, and efficiently utilize recycled resources like
water, thereby may be much more environmental friendly and self-sustainable than traditional farming.
In the light of current global challenges, such as disruptions to global supply chains and the threat of climate change,
CEA appears especially appealing as a food source for urban population centers. Under pressures of deglobalization
brought by geopolitical tensions [326] and global pandemics [209, 245], CEA provides the possibility to build farms
close to large cities, which shortens the transportation distance and maintains secure food supplies even when long-
distance routes are disrupted. The city-state Singapore, for example, has promised to source 30% of its food domestically
by 2030 [1, 276], which is only possible through suburban farms such as CEAs. Furthermore, CEA, as a form of precision
agriculture, is by itself a viable solution to the reduction the emission of greenhouse gases [9, 33, 220]. CEA can also
shield plants from adverse climate conditions exacerbated by climate change as its environments are fully controlled
[94] and is able to effectively reuse the arable land eroded due to climate change [328].
We argue that AI and CV are critical to the economic viability and long-term sustainability of CEAs as these tech-
nologies could alleviate cost associated with production and improve productivity. Suburban CEAs have high land cost.
An analysis in Victoria, Australia [34] shows that, due to the higher land cost resulted from proximity to cities, with
an estimated 50-fold productivity improvement per land area, it still takes 6 to 7 years for a CEA to reach the break-
even point. Thus, further productivity improvement from AI would act as strong drivers for CEA adoption. Moreover,
vertical or stacked setup of vertical farms impose additional difficulty for farmers to perform daily surveillance and
operations. Automated solutions empowered by computer vision could effectively solve this problem. Finally, AI and
CV technologies have the potential to fully characterize the complex, individually different, time-varying, and dynamic
conditions of living organisms [35], which will enable precise and individualized management and further elevate yield.
Thus, AI and CV technologies appear to be a natural fit to CEAs.
Most of the recent development of AI can be attributed to the newly discovered capability to train deep neural
networks [150] that can (1) automatically learn multi-level representations of input data that are transferable to diverse
downstream tasks [54, 116], (2) easily scale up to match the growing size of data [258], and (3) conveniently utilize
massively parallel hardware architectures like GPUs [96, 294]. As function approximators, deep learning proves to be
surprisingly effective in generalizing to previously unseen data [318]. Deep learning has achieved tremendous success
in computer vision [267], natural language processing [41, 71, 100], multimedia [19, 74], robotics [265], game playing
[247], and many other areas.
The AI revolution in agriculture is already underway. State-of-the-art neural network technologies, such as ResNet
[113] and MobileNet [118] for image recognition, and Faster R-CNN [215], Mask R-CNN [112], and YOLO [211] for
object detection, have been applied to the management of crops [172], livestock [121, 271], and indoor and vertical
farms [216, 321]. AI has been used to provide decision support in a myriad of tasks from DNA analysis [172] and
growth monitoring [216, 321] to disease detection [232] and profit prediction [24].
2
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

As the volume of research in smart agriculture grows rapidly, we hope the current review article can bridge re-
searchers from AI and agriculture and create a mild learning curve when they wish to familiarize themselves in the
other area. We believe computer vision has the closest connections with, and is the most immediately applicable in,
urban agriculture and CEAs. Hence, in this paper, we focus on reviewing deep-learning based computer vision tech-
nologies in urban farming and CEAs. We focus on deep learning because it is the predominant approch in AI and CV
research. The contributions of this paper are two-fold, with the former targeted at mainly AI researchers and the latter
targeted at agriculture researchers:

• We identify five major CV applications in CEA and analyze their requirements and motivation. Further, we
survey the state of the art as reflected in 68 technical papers and 9 vision-based CEA datasets.
• We discuss five key subareas of computer vision and how they related to these CEA problems. In addition, we
identify four potential future directions for research in CV for CEA.

We structure the survey as follows: First, to provide a bird-eye view of CV capabilities available to researchers in
smart agriculture, we summarize several major CV problems and influential technical solutions in §2. Next, we re-
viewed 68 papers with respect to the application of computer vision in CEA system in §3. The discussion is organized
into five subsections including Growth Monitoring, Fruit and Flower Detection, Fruit Counting, Maturity Level Clas-
sification, and Pest and Disease Detection. In the discussion, we focus on fruits and vegetables that are suitable for
CEA, including tomato [10, 13, 106, 316], mango [7], guava [246, 298], strawberry [90, 311], capsicum [152], banana
[5], lettuce [323], cucumber [10, 107, 178], citrus [4] and blueberry [2]. Next, we provide a summary of nine publicly
available datasets of plants and fruits in §4 to facilitate future studies in Controlled-environment agriculture. Finally,
we highlight a few research directions that could generate high-impact research in the near future in §5.

2 COMPUTER VISION CAPABILITIES RELEVANT TO SMART AGRICULTURE

2.1 Image Recognition
The classic problem of image recognition is to classify an image containing a single object to the corresponding object
class. The success of deep convolutional networks in this area dates (at least) back to LeNet [151] of 1998, which rec-
ognizes hand-written digits. The fundamental building block of such networks is the convolution operation. Using the
principles of local connections and weight sharing, convolutional networks benefit from an inductive bias of transla-
tional invariance. That is, a convolutional network applies (approximately) the same operation to all pixel locations of
the image.
The victory of AlexNet [143] in the 2012 ImageNet Large Scale Visual Recognition Challenge [224] is often consid-
ered as a landmark event that introduced deep neural networks into the AI mainstream. Subsequently, many variants
of convolutional networks [129, 148, 248, 262] have been proposed. Due to space limits, here we provide a brief review
of a few influential works, which is by no means exhaustive. ResNet [114] introduces residual connections that allow
the training of networks of more than 100 layers. ResNeXT [301] and MobileNet [119] employ grouped convolution
that reduces interaction between channels and improves the efficiency of the network parameters. ShuffleNet [329] uti-
lizes the shuffling of channels, which complements group convolution. EfficientNet [266] shows simultaneous scaling
of the network width, height, and image resolution is key to efficient use of parameters.
Recently, the transformer model has proven to be a highly competitive architecture for image recognition and other
computer vision tasks [75]. These models cut the input image into a sequence of small image patches and often apply
3
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

strong regularization such as RandAugment [63]. Variants such as CaiT [273], CeiT [315], Swin Transformer [173], and
others [61, 66, 302, 335] achieve outstanding performance on ImageNet.
Despite the maturity of the technology for image classification, the assumption that an image contains only one
object may not be easily satisfied in real-world scenarios. Thus, it is often necessary to adopt a problem formulation
as object detection or semantic / instance segmentation.

2.2 Object Detection

The object detection task is to identify and locate all objects in the image. It can be understood as the task resulted
from relaxing the assumption that the input image contains a single object. This is one natural problem formulation
for real-world images and has seen wide adoption in agricultural applications.
In broad strokes, contemporary object detection methods can be categorized into anchor-box-based and point-based
/ proposal-free approaches. In anchor-box methods [93, 214], the process starts a number of predefined anchor boxes
are periodically tiled to cover the entire input image. For each anchor box, the network makes two types of predictions.
First, it determines if the anchor box contains one of the predefined object classes. Second, if the box contains an object,
the network attempts to move and reshape the box to become closer to the ground-truth location of the object. One-
stage anchor-box detectors [65, 84, 165, 171, 212, 331] make these predictions all at once. In comparison, two-stage
detectors [93, 111, 164, 214], in the first stage discard anchor boxes that do not contain any object and classify the
remaining boxes into finer object categories in the second stage. The location adjustment, known as bounding box
regression, can happen in both stages. It is also possible to employ more than two stages [42]. When the objects have
diverse shapes and scales, these methods must create a large number of proposal boxes and evaluate them all, which
can lead to high computational cost.
While point-based object detectors [76, 138, 149, 272, 336] still need to identify rectangular boxes around the objects,
they make predictions at the level of grid locations on the feature maps. The networks predict if a grid location is a
corner or the center of an object bounding box. After that, the algorithm assembles the corners and centers into bound-
ing boxes. The point-based approaches can reduce the total number of decisions to be made. A careful comparison and
analysis of anchor-box methods and point-based methods can be found in [324].

2.3 Semantic, Instance, and Panoptic Segmentation

Segmentation is a pixel-level classification task, aiming to classify every pixel in the image into a type of object or an
object instance. The variations of the task differ by their definitions of the classes. In semantic segmentation [62, 78,
95, 145, 174], each type of object, such as cat, cow, grass, or sky, is its own class, but different instances of the same
object type (e.g., two cats) share the same class. In instance segmentation [64, 108, 110, 203], different instances of the
same object type become unique classes, so that two cats are no longer the same class. However, object types such as
sky or grass, which are not easily divided into instances, are ignored. In the recently proposed panoptic segmentation
[58, 92, 135, 156, 167, 327], objects are first separated into things and stuff. Things are countable and each instance
of things is its own class, whereas stuff is uncountable, impossible to separate into instances, appearing as texture or
amorphous regions [12], and remains as one class. We note that the distinction between things and stuff is not rigid and
can change depending on the application. For example, grass is typically considered as stuff, but in the leaf instance
segmentation task, each leaf of a plant becomes an instance and is a separate class.
The primary requirement of pixel-level classification is to learn pixel-level representations that consider sufficient
context and within reasonable computational budget. A typical solution is to introduce a series of downsampling
4
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

followed by a series of upsampling operations. since classic works such as the Fully Convolutional Network (FCN)
[174] and U-Net [223], this has been the mainstream strategy for various segmentation strategies.
Due to its use in leaf segmentation, a problem in plant phenotyping, instance segmentation may be the most relevant
segmentation formulation for urban farming. Despite the apparent similarity to semantic segmentation, instance seg-
mentation poses challenges due to the variable number of instance classes and possible permutation of class indices
[68]. This could be handled by combining proposal-based object detection and segmentation [50, 57, 108, 158, 204].
Mask-RCNN [111] exemplifies this approach. Leveraging its object detection capability, the network associates each
object with a bounding box. After that, the network predicts a binary mask for the object within the bounding box.
However, such methods may not perform well when there is substantial occlusion among objects or when objects are
of irregular shapes [68].
Departing from the detect-then-segment paradigm, recurrent methods [213, 222, 229] that outputs one segmentation
mask at one time may be considered as implicitly modeling occlusion. Pixel embedding methods [51, 68, 191, 199, 292,
296, 309] learn vector representations for every pixel and cluster the vectors. These methods are especially suitable for
segmenting plant leaves and we will discuss them in greater detail in Section 3.1. Taking a page from the proposal-free
object detector YOLO [211], SOLO [283] and SOLOv2 [284] divide the image into grids. The grid that the center an
object falls into is responsible for predicting the segmentation mask of the object.

2.4 Uncertainty Quantification

Real-world applications often require qualification of the amount of uncertainty in the predictions made by machine
learning, especially when the predictions carry serious implications. For example, if the system incorrectly determines
that fruits are not mature enough, it may delay harvesting and cause overripe fruits with diminished values. Thus,
users of the ML system are justified to ask how certain we are about the decision. In addition, when facing real-world
input, it is desirable for the network to answer “I don’t know” when facing an input that it does not recognize [161].
Well-calibrated uncertainty measurements may enable such a capability.
However, research shows that deep neural networks exhibit severe vulnerability to overconfidence, or under-estimation
of the uncertainty in its own decisions [99, 181]. That is, the accuracy of the network decision is frequently lower than
the probability that the network assigns to the decision. As a result, proper calibration of the networks should be a
concern for systems built for real-world applications.
Calibration of deep neural networks may be performed post-doc (after training) using temperature scaling and
histogram binning [73, 99, 279]. Also, regularization during training such as label smoothing [263] and mixup [117]
have been shown to improve calibration [189, 201, 270]. Researchers propose new loss functions to replace existing
ones that are susceptible to overconfidence [188, 308]. Moreover, ensemble methods such as Vertical Voting [300],
Batch Ensemble [289], and Multi-input Multi-output [109] can derive uncertainty estimates.

2.5 Interpretability
Modern AI systems are known for its inability to provide faithful and human-understandable explanations for its own
decisions. The unique characteristics of deep learning, such as network over-parameterization, large amount of training
data, and stochastic optimization, while being beneficial to the predictive accuracy (e.g., [23, 157, 251, 256]), all create
obstacles toward understand how and why a neural network reaches its decisions. The lack of human-understandable
explanations leads to difficulties in the verification and trust of network decisions [45, 330].
5
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

We categorize model interpretation techniques into a few major classes, including visualization, feature attribution,
instance attribution, inherently explainable models, and approximation by simple models. Visualization techniques
present holistically what the model has learned from the training data by visualizing the model weights for direct visual
inspection [31, 77, 81, 180, 187, 264]. In comparison, feature attribution and instance attribution are often considered as
local explanations as they aim to explain model predictions on individual samples. Feature attribution methods [18, 47,
49, 185, 207, 233, 243, 250, 261, 305] generate a saliency map of an image or video frame, which highlights the pixels that
contribute the most to its prediction. Instance attribution methods [28, 40, 56, 136, 206, 243, 306] attribute a network
decision to training instances that, through the training process, exert positive or negative influence on the particular
decision. Moreover, inherently explainable models [30, 48, 153, 237, 310] incorporate explainable components into the
network architecture, which reduces the need to apply post-hoc interpretation techniques. In contrast, researchers
also try to post-hoc approximate complex neural networks with simple models such as rule-based models [72, 85, 97,
133, 200, 285] or linear models [14, 88, 89, 140, 218] that are easily understandable.
The most significant benefit of interpretation in the context of CEA lies in its ability to aid with the auditing and
debugging of AI systems and datasets. With feature attribution, users can make sure the system captures the robust
features, or semantically meaningful features, that generalize to real-world data. As in the well-known case of husky
vs. wolf image classification, due to a spurious correlation, the neural network learns to classify all images with white
backgrounds as wolf and those with green backgrounds as husky [184]. Such shortcut learning can be identified by
feature attribution and subsequently corrected. Moreover, instance attribution allows researchers to pinpoint outliers
or incorrectly labeled training data that may lead to misclassification [56].

3 CONTROLLED-ENVIRONMENT AGRICULTURE
Controlled-environment agriculture (CEA) is the farming practice carried out within urban, indoor, resource-controlled
factories, often accompanied by stacked growth levels (i.e., vertical farming), renewable energy and recycling of water
and waste. CEA has recently been adopted in nations around the world [34, 70] such as Singapore [141], North America
[6], Japan [9, 242], and UK [8].
CEA has economic and environmental benefits. Compared to traditional farming, CEA farms produce higher yield
per unit area of land [8, 9]. Controlled environments shield the plants from seasonality and extreme weather, so that
plants can grow all year round given suitable lighting, temperature and irrigation [34]. The growing conditions can
be further optimized to boost growth and nutritional content [139, 274]. Rapid turnover increases farmers’ flexibility
in plant choice to catch the trend of consumption [32]. Moreover, farms investment on pesticides, herbicides, and
transportation can be cut down due to reduced contamination from the outside environment and proximity to urban
consumers.
CEA farms, when designed properly, can become much more environmental friendly and self-sustainable than
traditional farming. With optimized growing condition and limited external interference, the need for fertilizer and
pesticides decreases, so that we can reduce the amount of chemicals go into the environment as well as the resulted
pollution. Furthermore, CEA farms can save water and energy through the use of renewable energy and aggressive
water recycling. For instance, CEA farms from Spread, a Japanese company, recycle 98% of used water and reduce
the energy cost per head of lettuce by 30% with LED lightening [9]. Finally, CEA farm can be situated in urban or
suburban areas, thereby reducing transportation and storage cost. A simulation for different farm designs in Lisbon
shows vertical tomato farms with appropriate designs emit less greenhouse gas than conventional farms, mainly due
to reduced water use and transportation distance [33].
6
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

A significant drawback of CEA, however, lies in its high cost, which may be partially addressed by computer vision
technologies. According to [34], the higher land cost in Victoria, Australia means that the yield of vertical farms has
to be at least 50 times more than traditional farming to break even. Computer vision holds the promise of boosting the
level of automation and increasing yield, thereby making CEA farms economically viable.
CEA can take diverse form factors [32], such as glasshouses with transparent shells and completely enclosed facili-
ties. Depending on the arrangement, vertical farms can be classified into stacked horizontal systems, vertical growth
surfaces, and multi-floor towers. The form factors may pose different requirements for computer vision technologies.
For example, glasshouses with transparent shells utilize natural light to reduce energy consumption but may not pro-
vide sufficient lighting for CV around the clock. In comparison, a completely enclosed facility can have greater control
of lighting conditions. In a stacked horizontal system, the growth surface on the top may be better lit than those on
the bottom. In addition, in order to take pictures of plants on different shelf levels or in vertical growth surfaces, it may
be necessary to have the hardware that moves the camera vertically.
In the following, we investigate the application of autonomous computer vision techniques on Growth Monitoring,
Maturity Level Classification and Pest and Disease Detection to increase production efficiency. In addition to existing
applications, we also include techniques that can be easily applied to vertical farms even though they have not yet
been applied to them.

3.1 Growth Monitoring

Growth monitoring, a critical component of plant phenotyping, aims to understanding the life cycle of plants and
estimating yield [127] by monitoring various growth indicators such as the plant size, number of leaves, leaf sizes, land
area covered by the plant, and so on. Plant growth monitoring also facilitates in quantifying the effects of biological /
environmental factors on growth and thus is crucial for finding the optimal growing condition and developing high-
yield crops [190, 268].
As early as 1903, Wilhelm Pfeffer has recognized the potential of image analysis in monitoring plant growth [202,
255]. Traditional machine vision techniques such as gray-level pixel thresholding [198], Bayesian statistics [39] and
shallow learning techniques [128, 313], have been applied to segment the objects of interest, such as leaves and stems,
from the background to analyze plant growth. Compared to traditional methods, deep-learning techniques provide
automatic representation learning and are less sensitive to image quality variations. For this reason, deep learning
techniques for growth monitoring have recently gained popularity.
Among various growth indicators, leaf size and number of leaves per plant are the most commonly used [103, 127,
147, 231]. Therefore, in the section below, we first discuss leaf instance segmentation, which can support both indicators
at the same time, followed by a discussion of techniques for only leaf counting or for other growth indicators.

3.1.1 Leaf Instance Segmentation. Due to the popularity of the CVPPP dataset [182], the segmentation of leaf instance
has attracted special attention from the computer vision community and warrants its own section. These methods
include recurrent network methods [213, 222] and pixel embedding methods [51, 68, 199, 290, 296]. Parallel proposal
methods are popular for general-purpose segmentation (see Section 2.3), but are ill-suited for leaf segmentation. As
most leaves have irregular shapes, the rectangle proposal boxes used in these methods do not fit the leaves well,
resulting in many poorly positioned boxes. In addition, the density of leaves causes many proposal boxes to overlap
and compounds the fitting problem. As a result, it is difficult to pick out the best proposal box from the large number
of parallel proposals. Therefore, we focus on recurrent network based methods and pixel embedding based methods
7
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

Table 1. Performance of various leaf instance segmentation techniques on the CVPPP A1 test set. Higher SBD and lower |DiC|
indicate better performance. (GT-FG) indicates model making use of ground-truth foregrounds

Category Technique SBD (↑) |DiC| (↓)

End-to-end instance segmentation [213] 84.9 0.8
Sequential
RNN-SIS [230] 74.7 1.1
RIS [222] 66.6 1.1
Semantic Instance Segmentation [68] 84.2 1.0
Object-aware Embedding [51] 83.1 0.73
Pixel Embedding RHN + Cosine Embeddings [199] 84.5 1.5
Crop Leaf and Plant Instance Segmentation [290] 91.1 1.8
W-Net (GT-FG) [296] 91.9 -
SPOCO (GT-FG) [293] 93.2 1.7

in this section. Quality metrics for leaf segmentation include Symmetric Best Dice (SBD) and Absolute Difference in
Count (|DiC|). SBD calculates the average overlap between the predicted mask and the ground truth for all leaves. DiC
calculates the average number of miscalculated leaves over the entire test set.
Recurrent network based methods output masks for single leaves sequentially. Their decisions are usually informed
by the already segmented parts of the image, which are summarized by the recurrent network. [213] applied LSTM
and DeconvNet to segment one leaf at a time. The network first locates a bounding box for the next leaf, and performs
segmentation within that box. After that, leaves segmented in all previous iterations are aggregated by the recurrent
network and passed to the next iteration as contextual information. [222] employs convolution-based LSTMs (ConvL-
STM) with FCN feature maps as input. At each time step, the network outputs a single-leaf mask and a confidence
score. During inference, the segmentation stops when the confidence score drops below 0.5. [230] proposed another
similar method that combines feature maps with different abstraction levels for prediction.
Pixel embedding methods learn vector representations for the pixels so that pixels in irregularly shaped leaves can
become regularly shaped clusters in the representation space. With that, we can directly cluster the pixels. [290] per-
forms simultaneous instance segmentation of leaves and plants. The method predicts the leaf centroids and the offset
of each leaf pixel to its leaf centroid. The pixel locations plus their offset vectors would be close to their leaf centroids,
so that the clustering in the 2D metric space becomes easy. The authors propose an encoder-decoder framework, based
on ERFNet [221], with two decoders. One decoder predicts the centroids of plant and leaves. The other decoder predicts
the offset of each leaf pixels to the leaf centroid. The pixel location plus the offset vector hence should be very close to
the leaf centroid. The dispersion among all pixels of the same leaf can be modeled as a Gaussian distribution, whose
covariance matrix is also predicted by the decoder and whose mean is from the first decoder. The training maximizes
the Gaussian likelihood for all pixels of the same leaf. The same process is applied to pixels of the same plant.
[51, 199, 296] are three similar pixel embedding methods. They encourage pixels from the same leaf to have similar
embeddings and pixels from different neighboring leaves to have different embeddings and perform clustering in
the embedding space. Their networks consists of two modules, the distance regression module and pixel embedding
module. [199, 296] arrange the two modules in sequence, while [51] place them in parallel. The distance regression
module predicts the distance between the pixel and the closest object boundary. The pixel embedding module generates
an embedding vector for each pixel, so that pixels from the same leaves have similar embeddings and pixels from
8
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

different neighboring leaves have different embeddings. During inference, pixels are clustered around leaf centers,
which are identified as local maxima in the distance map from the distance regression module.
Lastly, [68, 293] take a large-margin approach. They ensure that embeddings of pixels from the same leaf are within
a circular margin of the leaf center, and the embedding of leaf centers are far away from each other. This removes the
need to determine the leaf centroids during inference because the embeddings are already well separated. [293] built
upon the method in [68] to perform pixel embedding and clustering of leaves under weak supervision, with annotation
on only a subset of instances in the images. In addition, a differentiable instance-level loss for a single leaf is formed to
overcome the non-differentiability of assigning pixels to instances by comparing a Gaussian shape soft mask with the
corresponding ground truth mask. Finally, consistency regularization, which encourage accordance of two embedding
framework is applied to to improve embedding for unlabeled pixels.
Comparing different approaches, proposal-free pixel embedding techniques seem to the best choice for the leaf
segmentation problem. As can be seen from Table 1, pixel embedding methods obtain both the highest SBD and lowest
|DiC|. One thing to note here, however, is that superior result of W-Net [296] and SPOCO [293] could be attributed to the
inclusion of ground-truth foreground masks during inference. Even though the recurrent approach does not generate
a large number of proposal boxes at once, it still uses rectangular proposals, which means that it still suffers from the
fitting problem to irregular leaf shapes. Moreover, the recurrent methods are usually slower than pixel embeddings,
due to the temporal dependence between the leaves.

3.1.2 Leaf Count and Other Growth Metrics. Leaf counts may be estimated without leaf segmentation. [275] employs
synthetic data in leaf counting task. In this study, a synthetic dataset of Arabidopsis rosette is generated using the
L-system-based plant simulator lpfg [3, 205]. The authors test a CNN, trained with only synthetic data, on real data
from CVPPP and obtain superior result than a model trained with CVPPP data only. In addition, the CNN trained
with the combination of synthetic and real data obtain approximately 27% reduction in the mean absolute count error
compared to CNN using only real data. These result demonstrate the potential of synthetic data in plant phenotyping.
Besides leaf size and leaf count, leaf fresh weight, leaf dry weight, and plant coverage (the area of land covered by
the plant) are also used as metrics of growth. [323] applies a CNN to regress leaf fresh weight, leaf dry weight, and
leaf area of lettuce with RGB images. [217] makes use of Mask R-CNN, a parallel proposal method, for lettuce instance
segmentation. With the segmentation mask and bounding box obtained, plant attributes such as contour, side view
area, height, and width are derived using preset formulas. Growth rate is then estimated from the changes in area of
plant at each time step. Additionally, fresh weight is estimated by a linear model, which model the relationship of plant
measured weight and attributes calculated from the preset formulas. [176] leverages COCO dataset pretrained Mask
R-CNN with ResNet-50 as backbone to segment leaf area of lettuce. Then the daily change of mean leaf area of plant
is used for growth rate calculation.

3.2 Fruit and Flower Detection

Algorithms for fruit and flower detection find the location and spatial distribution of fruits and fruit flowers. This task
supports various downstream applications such as fruit count estimation, size estimation, weight estimation, robotic
pruning, robotic harvesting, and disease detection [27, 91, 177, 307]. In addition, fruit or flower detection may also
facilitate in devising plantation management strategies [91, 106] because fruit or flower statistics such as positions,
facing directions (the directions the flowers face), and spatial scatter can reveal the status of the plant and the suitability
of environmental conditions. For example, the knowledge of flower distribution may allow pruning strategies that focus
9
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

on regions of excessive density and achieve even distribution of fruits, in order to optimize the delivery of nutrient to
the fruits.
Traditional approaches for fruit detection rely on manual feature engineering and feature fusion. As fruits tend to
have unique colors and shapes, one natural thought is to apply thresholding on color [197, 288] and shape information
[170, 196]. Additionally, [46, 162, 186] employ a combination of color, shape, and texture features. However, manual
feature extraction suffers from brittleness when the image distribution changes due to different camera resolutions,
camera angles, and illumination, and may not generalize easily to other species [26].
Deep learning methods for fruit detection include object detection and segmentation. Specifically, [316] applied SSD
for cherry tomato detection. [120] leverages Faster R-CNN to detect tomatoes. Inside the generated bounding boxes,
color thresholding and fuzzy-rule-based morphological processing methods are applied to remove image background
and obtain the contours of individual tomatoes. [226] leverages Faster R-CNN with VGG-16 as the backbone for sweet
pepper detection. RGB and near-infrared (NIR) images are used together for detection. Two fusion approaches, early
and late fusion, are proposed. Early fusion alters the first pretrained layer to allow 4 input channels (RGB and NIR),
whereas late fusion aggregates the two modalities by training independent proposal models for each modality and then
combining the proposed boxes by averaging the predicted class probabilities. [320] trains three multi-task cascaded
convolutional networks (MTCNN) [319] for detecting apples, strawberries and oranges. MTCNN contains a proposal
network, a bounding box refinement network, and an output network in a feature pyramid architecture with gradually
increased input sizes for each network. The model is trained on artificial images, which are generated by combining
randomly cropped negative patches with fruits patches, in addition to real-world images. [311] proposed R-YOLO with
MobileNet-V1 as the backbone to detect ripe strawberries. Different from regular horizontal bounding boxes in object
detection, the model generates rotated bounding boxes by adding a rotation-angle parameter to the anchors.
Delicate fruits, such as strawberries and tomatoes, are particular vulnerable to damage during harvesting. Therefore,
much research has been devoted to segmenting such fruits from backgrounds in order to determine the precise picking
point. Precise fruit masks are expected to enable robotic fruit picking while avoiding damages on the neighboring
fruits. [163] performs semantic segmentation for guava fruits and determines their poses using FCN with RGB-D
images as input. The FCN outputs a binary mask for fruits and another binary mask for branches. With the fruit
binary mask, the authors employ Euclidean clustering [225] to cluster single guava fruit. From the clustering result
and the branch binary mask, fruit centroids and the closest branch are located. Finally, the system predicts the vertical
axis of the fruit as the direction perpendicular to the closest branch to facilitate robotic harvesting. Similarly, [13]
leverages Mask R-CNN with ResNet as backbone for semantic segmentation of tomato. Additionally, the authors filter
the false positive detection due to tomatoes from the non-targeted rows by setting a depth threshold. [90] utilizes
Mask R-CNN with a ResNet101 backbone to perform instance segmentation of ripe strawberries, raw strawberries,
straps and tables. Depth images are aligned with the segmentation mask to project the shape of strawberries into 3D
space to facilitate automatic harvesting. [312] also applies Mask R-CNN with a ResNet101 + FPN backbone to perform
instance segmentation and ripeness classification on strawberries. [122] leverages a similar network to segment tomato
instances. With the segmentation mask, the systems determine the cut points of the fruits.
Besides accuracy, the processing speed of neural networks is also important for deployment on mobile devices or
agricultural robots. As an example, [240] performs network pruning on YOLOv3-tiny and obtains a lightweight mango
detection network. A YOLOv3-tiny pretrained on the COCO dataset has learned to extract fruit-relevant features
because the COCO dataset contains apple and orange images, but it also has learned irrelevant features. The authors
thus use a generalized attribution method [244] to determine the contribution of each layer to fruit features extraction
10
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

and remove convolution kernels responsible for detecting non-fruit classes. They find that the lower level features are
shared across all classes detection and pruning in the higher layers does not harm fruit detection performance. After
pruning, the network achieve significantly lowers float-point operations (FLOPs) at the same level of accuracy.
Object detection is also applied for flower detection. [177] propose a modified YOLOv4-Tiny with cascade fusion
(CFNet) to detect citrus buds, citrus flowers, and gray mold, which is a disease commonly found on citrus plants. The
authors propose a block module with channel shuffle and depth separable convolution for YOLOv4-Tiny. [259] shrinks
the anchor boxes of Faster-RCNN to fit small fruits and applies soft non-maximum suppression to retain boxes that
may contain occluded objects. As flowers usually have similar morphological characteristics, flowers from other non-
targeted species could possibly be used as training data in a transfer learning scenario. In [260], the authors finetune a
DeepLab-ResNet model [52] for fruit flower detection. The model is trained on apple flower dataset but achieves high
F1 scores on pear and peach flower images (0.777 and 0.854 respectively).

3.3 Fruit Counting

Pre-harvest estimation of yields plays an important role in the planning of harvesting resources and marketing strate-
gies [115, 303]. As fruits are usually sold to consumers as a pack of uniformly sized fruits or individual fruits, the fruit
count also provides an effective yield metric [137], besides the distribution of fruit sizes. Traditional yield estimation
is obtained through manual counting of samples from a few randomly selected areas [115]. Nonetheless, when the
production is large-scale, to counteract the effect of plant variability, a large quantity of samples from different areas
of the field is needed for accurate estimation, resulting in high cost. Thus, researchers resort to CV-based counting
methods.
A direct counting method is to regress on the image and output the fruit count. In [210], the authors apply a modified
version of Inception-ResNet for direct tomato counting. The authors train the model on simulated images and test on
real images, which suggest, once again, the viability of using simulated images to circumvent the cost for formulating
large dataset.
Besides direct regression, object detection [137, 286], semantic segmentation [134], and instance segmentation [194]
have also been applied for fruit counting. These methods provide an intermediate level of results from which the
count can be easily gathered. [137] proposes MangoYOLO based on YOLOv2-tiny and YOLOv3 for mango detection
and counting. As fruit are usually small in the images, the authors extract features with higher spatial resolution
than the original YOLOv3 in order to preserve details in the image. [134] performs semantic segmentation for mango
counting using a modification of FCN. The coordinates of blob-like regions in the semantic segmentation mask is used
to generate bounding boxes corresponding to mango fruits in the original images. Finally, [194] applies Mask R-CNN
to for instance segmentation of blueberries. The model classifies the maturity of individual blueberries and count the
number of berries according to the masks.
Occlusion poses a difficult challenge for counting. Due to this issue, automatic count from detection or segmentation
results is almost always lower than the actual number of fruits on the plant. To deal with this problem, [137] calculates
a correction factor which is the ratio of the actual hand harvest count to the fruit count in images, and applies the
correction to all predictions. [137], also uses dual views of mango trees (front and back) to detect mangos that may
be occluded from one angle. Taking this idea one step further [286] use dual-view video to detect and track mangos
when the camera moves. The authors apply MangoYOLO for mango detection in each frame of the video. Individual
mangos are tracked based on distance across frames. When a tracked mango disappears in the subsequent frames,
Kalman Filter is used to predict the position of occluded fruit for maximum of 15 frames to prevent double counting. If
11
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

the fruit disappears for more than 15 frames, it is removed from tracking list and will be counted again if it reappears
later. Even though [286] manages to recognize around 20% more fruits, the detected count is still significantly lower
than the actual number, underscoring the research challenge of exhaustive and accurate counting.

3.4 Maturity Level Classification

Maturity level classification aims to determine the ripeness of fruits or vegetables to aid in proper harvesting and food
quality assurance. Premature harvesting results in plants that are unpalatable or incapable of ripening, while delayed
harvesting can result in overripe plants or food decay [122].
The optimal maturity level differs for different targeted products and destinations. Fruits and vegetables can be
consumed at different growing stages. For example, lettuce can be consumed either as baby lettuce or fully grown
lettuce. The same situation happens with baby corn and normal corn. Products are to be transported to different
destinations, so we must consider the length of transportation and ripening speed when deciding the correct maturity
level at harvest [322].
Manually distinguishing the subtle differences in maturity levels is time-consuming, prone to inconsistency, and
costly. The labor cost of harvesting accounts for a large percentage of operation cost in farms, with 42% of variable
production expenses in U.S. fruit and vegetable farms being spent on labor for harvesting [123]. Automatic maturity
level classification with computer vision, in contrast, can assist automatic harvesting [17, 90, 322] and reduce cost.
Similar to fruit detection, we can apply thresholding methods on color to detect ripeness. For example, [21] trans-
forms image color to HSI and YIQ color spaces before thresholding. [269] applies linear color models. [154] utilizes the
combination of color and texture features. [80, 144, 234, 235, 295] apply shallow learning methods based on a multitude
of features.
More recently, researchers evaluate the performance of deep learning based computer vision methods on maturity
level classification and attain satisfactory results. For example, [321] applied a network with five convolution layer and
a fully connected layer to classify tomato maturity into five levels. However, to further facilitate automatic harvesting,
object detection and instance segmentation are more commonly used for getting the exact shape, location and maturity
level of fruits and position of peduncle for robotic end effectors to cut on.
With object detection, [311] applies the R-YOLO network described in the fruit detection (§3.2) to detect ripe
strawberries. [104] proposes pre-trained Faster R-CNN network, building upon DeepFruits [226], to estimate both the
ripeness and quantity of sweet pepper. Two formulations of the model are tested. One treats ripe/unripe as additional
classes on top of foreground/background, and the other performs foreground/background classification first and then
performs ripeness classification on foreground regions. The second approach generates better ripeness classification
result as the ripe/unripe classes are more balanced when only the foreground regions are considered. Additionally, the
authors design a tracking sub-system which calculates the quantity of fruit available for harvesting. The sub-system
identifies new fruits by measuring the IoU between and comparing the boundary of detected and new fruits.
Using the segmentation methods discussed in §3.2, [13] performs semantic segmentation of and classification of
ripe and raw tomatoes. [90, 312] perform instance segmentation and classification of ripe and raw strawberries. [122]
perform instance segmentation on tomatoes first. Then, by transforming the mask region into HSV color space, the
authors apply a fuzzy system to classify tomatoes into four classes: immature (completely green), breaker (green to
tannish), preharvest (light red), and harvest (fully colored).

12
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

3.5 Pest and Disease Detection

Plants are susceptible to environmental disorders caused by temperature, humidity, nutritional excess/deficiency, light
changes and biotic disorders due to fungi, bacteria, virus or other pests [86, 249]. Infectious diseases or pest pandemic
induce inferior plant quality or plant death, resulting in at least 10% of global food production losses [257].
Although controlled vertical farming restrict the entry of pests and diseases, it cannot eliminated them completely.
Accidental contamination from employees, seeds, irrigation water and nutrient solution, poorly maintained environ-
ment or phytosanitation protocols, unsealed entrance and ventilation systems can all be sources of pests and diseases
[219]. For this reason, pest and disease detection is still worth studying in the context of CEA.
Manual diagnosis of plant is complex due to the large quantity of vertically arranged plants in the field and numerous
possible symptoms of diseases on different species. In addition, plants show different patterns along infection cycles
and their symptoms can vary in different part of the plant [37]. Consequently, autonomous computer vision systems
that recognize diseases according to the species and plant organs are gaining traction. From a technological perspective,
we sort existing techniques into three parts, namely single- and multi-label classificaiton, handling unbalanced class
distributions, as well as label noise and uncertainty estimates.

3.5.1 Single- and Multi-label Classification. Studies applied classification networks for classifying diseases of either
one single species [20, 232, 249, 325] or multiple species [79]. [325] creates a lightweight version of AlexNet to clas-
sify six type of cucumber diseases by replacing the fully connected network with a global pooling layer. [249] lever-
ages CNNs for the classification of leaves into mango leaves, diseased mango leaves and other plants leaves. [20]
utilizes AlexNet and VGG16 to recognize five types of pests and diseases of tomatoes. [232] assigns one disease label to
each image of banana plants and locates diseases using object detection. [79] applies AlexNet, AlexNetOWTBn [142],
GoogLeNet, Overfeat [236], and VGG for classifying 25 different healthy or diseased plants.
There is one possible drawback with single-label datasets. In the real world, one plant or one leaf sometimes carries
multiple diseases or multiple areas with disease symptoms. If a model is trained with multiple types of diseases, it could
detect multiple targeted area or disease classes from the images. However, models trained on single-label datasets will
only recognize one disease. To deal with the possibility of having multiple diseases or multiple areas of diseases on
one plant simultaneously, two types of methods are proposed. [179] first segments out different infection area on cu-
cumber leaves using color thresholding following [178], then applies DCNN on segmented images for classification
of four types of cucumber diseases. Nevertheless, the color thresholding technique may not generalize to different
environment conditions and plant species. Another type of methods leverages object detection or segmentation for
locating and classifying infection areas on plants. [86] compared Faster R-CNN, R-FCN and SSD for detecting nine
classes of diseases and pests that affect tomato plants. Multiple diseases and pests in one plant are detected simultane-
ously. [314] applies improved DeepLab v3+ for segmentation of multiple black rot spots on grape leaves. The efficient
channel attention mechanism [282] is added to the backbone of DeepLab v3+ for capturing local cross-channel inter-
action. Feature pyramid network and Atrous Spatial Pyramid Pooling [53] are utilized for fusing feature maps from
the backbone network at different scale to improve segmentation.

3.5.2 Handling Unbalanced Class Distributions. A common obstacle encountered in disease detection is unbalanced
disease class distributions. There are typically much fewer diseased plants than healthy plants; as diseases have unequal
frequency, it is usually difficult to find images of rare diseases. The data unbalance leads to difficulty for model training.
13
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

Thus, researchers have proposed remediation methods such as weakly supervised learning [38], generative adversarial
network (GAN) [98], and few-shot learning [160, 195].
Specifically, [38] applies multiple instance learning (MIL), a type of weakly supervised learning method, for multi-
class classification of six mite species on citrus. In MIL, learner receives a set of labeled bags, containing multiple image
instances. We know that at least one instance is associated with the class label, but do not know the exact instance.
The MIL algorithm tries to identify the common characteristic shared by images in the positively labeled bags. In
this work, a CNN is first trained with labeled bags. Next, by calculating saliency maps of images in bags, the model
identifies salient patches that have high probability of containing mites. These patches inherit labels from their bags
and are used to refine the CNN trained above.
[98] leverages generative adversarial network (GAN) to generate realistic image patches of tip-burn lettuce and train
model for tip-burn segmentation. For the generation stage, lettuce canopy image patches are inputted into Wasserstein
GANs [22] to generate stressed (tip-burned) patches so that there are equal number of stressed and healthy patches.
Then, in the segmentation stage, the authors generate a binary label map for the images using a classifier and a edge
map. The binary label map labels each mini-patches (super-pixels) as stressed or healthy. The authors then feed the
label map, alongside the original images, as input to U-net for mask segmentation.
In few-shot meta-learning, we are given a meta-train set and a meta-test set, with the two sets containing mutually
exclusive image classes (i.e. classes in training set do not appear in testing set). Meta-train or meta-test set contain a
number of episodes, each of which consists of some training (supporting) images and some test (query) images. The
rationale of meta-learning is to equip the model with the ability to quickly learn to classify the test images from a small
number of training images within each episode. The model acquires this meta-learning capability on the meta-train
set and is evaluated on the meta-test set.
As an example of few-shot meta-learning, [195] performs classification of plant pest and diseases. The model frame-
work consists of an embedding module and a distance module. The embedding module first projects supporting images
in to an embedding space using ResNet-18, then feed embedding vectors into a transformer to incorporate information
of other support samples in the same episode. After that, the distance module calculate the Mahalanobis distance[87]
of the query and support samples to classify the query. Similarly, [160] use a shallow CNN for embedding and the
Euclidean distance for calculating the similarity between the embeddings of the query and support samples.

3.5.3 Label Noise and Uncertainty Estimates. [241] is another example of meta-learning, where it is used to improve
the network’s robustness against label noise. The model consist of two phrases. The first phrase is the conventional
training of a CNN for classification. In the second phrase, the authors generate ten synthetic mini batches of images,
containing real images with the labels taken from similar images. As a result, these mini-batches could contain noisy
labels. After one step update on the synthetic instances, the network is trained to output similar predictions with the
CNN from the first phrase. The result is a model that is not easily affected by noisy training data.
Finally, having a confidence score associated with the model prediction allows farmers to make decision selectively
under different confidence levels and boost the acceptance of deep learning models in agriculture. As an example,
[82] perform classification of tomato diseases and pair the prediction with a confidence score following [67]. The
confidence score, calculated using Bayes’ rule, is defined as the probability of the true class label conditioned on the
class probability predicted by the CNN. In addition, the authors build an ontology of disease classification. For example,
the parent node “stressed plant” has as children “bacteria infection” and “virus infection”, which in turn has “mosaic
virus” as a child. If the confidence score of a specific terminal disease label is below certain threshold, the model switch
14
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

to its more general parent label in the tree for higher confidence. By the axiom of probability, the predicted probability
of the parent label is the summation of all the predicted probability of its direct descendants. For a general discussion
of machine learning techniques that create well-calibrated uncertainty estimates, we refer readers to §2.4.

4 DATASETS
High-quality datasets with human annotations are one of the most important factors in the success of a machine
learning project [183, 192, 291]. In this section, we review datasets for computer vision in CEA or datasets related
to plants suitable for CEA. We exclude datasets for plants as we have not found literature regarding their suitability
in CEA, such as apples, broccoli, and dates. We have manually checked every dataset listed and made sure that they
are available for downloading at the time of writing. By summarizing dataset related to CEA, we aim to facilitate
interested researchers on their future studies. In the mean time, we would like encourage scholars to publish more
datasets dedicated to CEA environments.
As listed in Table 2, we discover nine datasets in CEA, with one for Growth Monitoring, five for Fruit Detection,
and four for Pest and Disease Detection. Each targeted task contain at least one dataset that covers multiple species to
facilitate training of generalizable and transferable models. The largest dataset is CVPPP with 6,287 and 165,120 RGB
images for Arabidopsis and Tobacco respectively, aiming for growth monitoring related tasks. In total, eight datasets
contains real images, with the only exception being the Capsicum Annuum dataset. While real images provide realistic
data, synthetic datasets feature balanced class distribution and accurate labeling. One point noteworthy is that many
real images are also collected under simplified laboratory environments, which may bias the data toward specific
lighting conditions, backgrounds, plant orientation, or camera positions. When applying the models learned on such
data to the real world, practitioners may consider finetuning on more realistic data.

5 FUTURE RESEARCH DIRECTIONS

So far we have discussed the objectives, benefits, and realizations of Growth Monitoring, Fruit and Flower Detection,
Fruit Counting, Maturity Level Classification, and Pest and Disease Detection in CEA precision farming. Based on cur-
rent status of the research area and existing technical capabilities of computer vision, we would like to point out several
areas where computer vision technologies could provide short- to mid-term benefits to urban and suburban CEA. We
identify three such areas, including realistic datasets that are unbalanced and noisy, uncertainty quantification, and
multi-task learning / system integration.

5.1 Handling Realistic Data

The ability to handle realistic data is a critical competence that has not received sufficient research attention (with a
few notable exceptions [38, 98, 160, 195, 241]). Unlike well-curated datasets that have accurate and abundant labels and
relatively balanced label distributions, real-world data exhibit skewed label distribution as well as substantial noise in
the labels. For effective real-world application, it is important that the CV algorithms can maintain good predictive
performance under these conditions. In addition, the algorithmic tolerance of data imperfection can lower annotation
cost and enable wider applications of CV. There has been substantial research on these topics in the computer vision
community, such as long-tail recognition [69, 168, 239, 280, 333, 334], few-shot and zero-shot learning [155, 252–
254, 299], as well as noise-resistant classification [16, 59, 131, 287, 332] and metric learning [126, 166, 277]. We believe
that research on intelligent urban agriculture could benefit from the existing body of literature.
15
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

Table 2. Dataset for CV tasks in CEA

Target Task Dataset Data Description URL

Growth CVPPP dataset [182] 6,287 and 165,120 RGB images of Ara- https://www.plant-phenotyping.org/datasets-downl
Monitoring bidopsis and Tobacco respectively. An-
notations include bounding boxes and
segmentation masks for every plant
and every leaf, and the leaf centers.

DeepFruits[227] RGB images of sweet pepper, rock https://drive.google.com/drive/folders/1CmsZb1cagg

melon, apple, mango, orange and
strawberry annotated with rectangular
bounding boxes. Each fruit has 42-170
images.
Orchard Fruit[26] 1,120, 1,964 and 620 RGB images of ap- http://data.acfr.usyd.edu.au/ag/treecrops/2016-multi
ple, mango and almond, respectively.
Fruit Apples are annotated with bounding
Detection circles; mango and almond are anno-
tated with rectangular bounding boxes
MangoYOLO[137] 1,730 RGB Mango images annotated https://figshare.com/articles/dataset/MangoYOLO_d
with rectangular bounding boxes; pho-
tos are under artificial lighting
Capsicum Annuum 10,500 synthetic RGB images of sweet https://data.4tu.nl/repository/uuid%3a884958f5-b868
dataset[29] pepper with pixel-level segmentation
masks of 8 plant parts, including stem,
node, side shoot, leaf, peduncle, fruit
and flower

Banana pest and dis- 18,000 RGB images of banana plants, https://pestdisplace.org/
eases detection [232] leaves, fruit bunch, cut fruits, pseu-
dostem and corms of 7 classes of ba-
nana diseases and pest
Plant Village [124] 61,486 RGB images of 39 different https://data.mendeley.com/datasets/tywbtsjrjv/1
classes of diseased and healthy plant
leaves
Pest and
Disease Crop Pests Recognition 5,629 RGB images of 10 pest classes, https://bit.ly/2DdUFza
Detection [159] with each class containing over 400 im-
ages
Plant and Pest [160] 6,000 RGB images of 20 different classes https://zenodo.org/record/4529076#.YupE_-xBzlw
of plant leaves and pests from Plant Vil-
lage [124] and Crop Pests Recognition
[159]

16
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

5.2 Quantifying Uncertainty and Interpretability

Real-world applications call for reliable estimation of the quality of automated decision. An incorrect prediction made
by an AI system may have profound implications. For example, if the system incorrectly determines that fruits are not
mature enough, it may delay harvesting and cause overripe fruits with diminished values. However, it is impossible to
eliminate incorrect or uncertain predictions, as they originate from factors difficult to control and precisely measure,
including model assumptions, test data shift, incomplete training data and so on [11, 125]. Thus, we argue that uncer-
tainty quantification is another crucial factor for real-world deployment. Such quantification would allows farmer to
make informed decisions on whether to follow the machine recommendation or not. For the convenience of readers,
we provide a brief review of such deep learning techniques in Section 2.4.
Besides uncertainty quantification, pair the model with explanation on its decisions could boost user confidence
and assist auditing and debugging of the AI system. Specifically, instance attribution methods, as discussed in §2.5,
enable detection of the biased or low quality data points with extreme influence on prediction [56]. For example, if
the model is trained with a image of dry leaves with dust that resemble a certain disease of the plant, in the inference
process, the model might misclassify diseased leaves as normal dry leaves or vice versa and induce plant death or
unnecessary treatments. With instance attribution interpretation, researchers can identify misleading data points and
perform adversarial training to improve model accuracy.

5.3 Multi-task Learning and System Integration

Real-world deployment usually requires the coordination of multiple CV capabilities provided by different networks.
When the system is designed well, these networks could facilitate each other and achieve synergistic effects. For exam-
ple, instance segmentation can be used for fruit and flower localization (Section 3.2), growth monitoring (Section 3.1),
and fruit maturity level detection (Section 3.4). However, academic research tends to study these problems in isolation,
thereby unable to reap benefits of multi-task learning.
Multi-task learning [25, 44, 169] focuses on leveraging mutually beneficial supervisory signals from multiple corre-
lated tasks. Recently, CV researchers have built large-scale networks [55, 60, 102, 130, 132, 175, 281, 337] that perform a
wide range of tasks, which achieve state-of-the-art results on most tasks. This demonstrates the benefits of multi-task
learning and could inspire similar work dedicated to smart farming in CEAs.
Another motivation for considering multi-task learning and system integration is that errors can propagate in a
pipeline architecture. For example, a network could first incorrectly detect a leaf as a fruit and then classify it as
an immature fruit. As a result, simply concatenating multiple techniques will result in inferior overall performance
than what practitioners may expect. Thus, we encourage system designers to consider end-to-end training, or other
innovative techniques [101, 297, 304] for aligning and interfacing different components within a system.

5.4 Effective Use of Multimodality

Fusion of multi-modal data enhances inference ability of models by incorporating complementary view of data [146].
In the context of CEA, thermal or depth images capture the depth or temperature differences between foreground and
background and enable filtering of non-target objects (e.g., fruits or leaves). Abnormal temperature changes during
growth cycle can also indicate disease infection before visual symptoms appear. Moreover, LiDAR and RGB-D systems
allow the generation of high density 3D point clouds of plants and fruits, which facilitate size measurement or cut-point
detection during harvesting.
17
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

Existing works have demonstrated the efficacy of depth images and hyper-spectral images [13, 36, 193, 278]. Hyper-
spectral images as the sole modality in early disease detection. With a GAN-based data augmentation method, [278]
performs early detection of tomato spotted wilt virus before visible symptoms appear using hyper-spectral images
whose wavelengths range from 395 nanometers (nm) to 1005nm. [193] performs early detection of grapevine vein-
clearing virus and shows the discriminative power of infrared imaging.
However, systematic exploration of fusion techniques for multimodal input remains relatively rare in CEA appli-
cations. Many existing approaches adopt pipeline-based multimodal integration techniques that does not exhaust the
potential of deep learning due to the lack of end-to-end training. For example, in [13], the authors set a depth threshold
to filter false positive tomato detections from the background. [36] first performs broccoli segmentation on the RGB
image. Within the segmentation mask, the authors find the mode of the depth value distribution, which is used to
calculate the diameter of the broccoli head. [163] conducts semantic segmentation for guava fruits using RGB images
and reconstructs their 3D positions from the depth input. These methods use the two modalities separately and do not
apply end-to-end training of the pipeline. As an exception, [226] proposes late fusion of RGB and near-infrared images
in sweet pepper detection.
In computer vision research, numerous techniques for fusing and joint utilization of multimodal information have
been proposed over the years, which we believe could contribute to CEA applications. Due to space limits, we list only
a few examples here. [238] proposes two different ways to combine multiple modalities in object detection, Concatena-
tion and Element-wise Cross Product. The former combines feature maps from different modalities along the channel
dimension and let the network discover the best way to combine them from data. The latter technique, Element-wise
Cross Product applies element-wise multiplication to every possible pair of feature maps from the two modalities. [43]
experiments with a variety of fusion techniques for RGB and optical flow and discovers a high-performing late-fusion
strategy in action recognition. In self-supervised learning, [105] identifies similar data points using one modality and
treat them as positive pairs in another modality. This technique provides another paradigm to leverage the comple-
mentary nature of multimodality.

6 CONCLUSIONS
Smart agriculture, and particularly computer vision for controlled-environment agriculture (CV4CEA), are rapidly
emerging as an interdisciplinary area of research that could potentially lead to enormous economic, environmental
and social benefits. In this survey, we first provide brief overviews of existing CV technologies that range from image
recognition to structured understanding such as segmentation, from uncertain quantification to interpretable machine
learning. Next, we systematically review existing applications of CV4CEA, including growth monitoring, fruit and
flower detection, fruit counting, maturity level classification, and pest / disease detection. Finally, we highlight a few
research directions that could generate high-impact research in the near future.
Like any interdisciplinary area, research progress in CV4CEA requires expertise in both computer vision and agri-
culture. However, it could take substantial amount of time for any researcher to acquire in-depth understanding of
both subjects. By reviewing existing applications, available CV technologies, and identifying possible future research
directions, we aim to provide a quick introduction of CV4CEA to researchers with expertise in agriculture or computer
vision alone. It is our hope that the current survey will serve as a bridge between researchers from diverse backgrounds
and contribute to accelerated innovation in the next decade.

18
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

7 ACKNOWLEDGMENTS
The authors gratefully acknowledge the support from the WeBank-NTU Joint Research Center under grant NWJ-2020-
008, and that from China-Singapore International Joint Research Institute under grant 206-A021002.

REFERENCES
[1] [n. d.]. 30 by 30: Strengthening our food security. https://www.ourfoodfuture.gov.sg/30by30. Accessed: 2022-8-15.
[2] [n. d.]. AeroFarms Partners With Hortifrut to Grow Blueberries, Caneberries Via Vertical Farming.
https://thespoon.tech/aerofarms-partners-with-hortifrut-to-grow-blueberries-caneberries-via-vertical-farming/. Accessed: 2022-7-28.
[3] [n. d.]. Algorithmic Botany. http://www.algorithmicbotany.org/virtual_laboratory/. Accessed: 2022-6-20.
[4] [n. d.]. All in(doors) on citrus production. https://www.hortibiz.com/newsitem/news/all-indoors-on-citrus-production/. Accessed: 2022-7-28.
[5] [n. d.]. Greenhouse in Shanghai successfully plants bananas on water. https://www.hortidaily.com/article/9369964/greenhouse-in-shanghai-successfully-plants-bananas-on-water/.
Accessed: 2022-7-28.
[6] [n. d.]. Introducing VertiCrop™. https://verticrop.com/. Accessed: 2022-5-24.
[7] [n. d.]. Mango trees cultivation under greenhouse conditions. https://horti-generation.com/mango-trees-cultivation-under-greenhouse-conditions/.
Accessed: 2022-7-28.
[8] [n. d.]. Saturn Bioponics. http://www.saturnbioponics.com/. Accessed: 2022-05-25.
[9] [n. d.]. Spread-A new way to grow vegetable. https://spread.co.jp/en/environment/. Accessed: 2022-05-24.
[10] [n. d.]. Tomatoes and cucumbers in a vertical farm without daylight. https://www.hortidaily.com/article/9212847/tomatoes-and-cucumbers-in-a-vertical-farm-without-daylight/.
Accessed: 2022-7-28.
[11] Moloud Abdar, Farhad Pourpanah, Sadiq Hussain, Dana Rezazadegan, Li Liu, Mohammad Ghavamzadeh, Paul Fieguth, Xiaochun Cao, Abbas
Khosravi, U Rajendra Acharya, et al. 2021. A review of uncertainty quantification in deep learning: Techniques, applications and challenges.
Information Fusion 76 (2021), 243–297.
[12] Edward H. Adelson. 2001. On seeing stuff: the perception of materials by humans and machines. In Human Vision and Electronic Imag-
ing VI, Bernice E. Rogowitz and Thrasyvoulos N. Pappas (Eds.), Vol. 4299. International Society for Optics and Photonics, SPIE, 1 – 12.
https://doi.org/10.1117/12.429489
[13] Manya Afonso, Hubert Fonteijn, Felipe Schadeck Fiorentin, Dick Lensink, Marcel Mooij, Nanne Faber, Gerrit Polder, and Ron Wehrens. 2020.
Tomato fruit detection and counting in greenhouses using deep learning. Frontiers in plant science 11 (2020), 571299.
[14] Isaac Ahern, Adam Noack, Luis Guzman-Nateras, Dejing Dou, Boyang Li, and Jun Huan. [n. d.]. NormLime: A New Feature Importance Metric
for Explaining Deep Neural Networks. arXiv Preprint 1909.04200 ([n. d.]). https://doi.org/10.48550/ARXIV.1909.04200
[15] Latief Ahmad and Firasath Nabi. 2021. Agriculture 5.0: Artificial Intelligence, IoT and Machine Learning. CRC Press.
[16] Görkem Algan and Ilkay Ulusoy. 2020. Meta soft label generation for noisy labels. In ICPR.
[17] H Altaheri, M Alsulaiman, M Faisal, and G Muhammed. 2019. Date fruit dataset for automated harvesting and visual yield estimation. In Proc.
IEEE DataPort.
[18] Marco Ancona, Enea Ceolini, Cengiz Öztireli, and Markus Gross. 2017. Towards better understanding of gradient-based attribution methods for
deep neural networks. arXiv preprint arXiv:1711.06104 (2017).
[19] Peter Anderson, Xiaodong He, Chris Buehler, Damien Teney, Mark Johnson, Stephen Gould, and Lei Zhang. 2018. Bottom-Up and Top-Down
Attention for Image Captioning and Visual Question Answering. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition
(CVPR).
[20] Krishnaswamy R Aravind, Purushothaman Raja, Rajendran Ashiwin, and Konnaiyar V Mukesh. 2019. Disease classification in Solanum melongena
using deep learning. Spanish Journal of Agricultural Research 17, 3 (2019), e0204–e0204.
[21] Arman Arefi, Asad Modarres Motlagh, Kaveh Mollazade, and Rahman Farrokhi Teimourlou. 2011. Recognition and localization of ripen tomato
based on machine vision. Australian Journal of Crop Science 5, 10 (2011), 1144–1149.
[22] Martin Arjovsky, Soumith Chintala, and Léon Bottou. 2017. Wasserstein generative adversarial networks. In International conference on machine
learning. PMLR, 214–223.
[23] Sanjeev Arora, Nadav Cohen, and Elad Hazan. 2018. On the optimization of deep networks: Implicit acceleration by overparameterization. In
International Conference on Machine Learning. PMLR, 244–253.
[24] S Aruul Mozhi Varman, Arvind Ram Baskaran, S Aravindh, and E Prabhu. 2017. Deep Learning and IoT for Smart Agriculture Using WSN. In 2017
IEEE International Conference on Computational Intelligence and Computing Research (ICCIC). 1–6. https://doi.org/10.1109/ICCIC.2017.8524140
[25] BJ Bakker and TM Heskes. 2003. Task clustering and gating for bayesian multitask learning. (2003).
[26] Suchet Bargoti and James Underwood. 2017. Deep fruit detection in orchards. In 2017 IEEE International Conference on Robotics and Automation
(ICRA). IEEE, 3626–3633.
[27] Suchet Bargoti and James P Underwood. 2017. Image segmentation for fruit detection and yield estimation in apple orchards. Journal of Field
Robotics 34, 6 (2017), 1039–1060.

19
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

[28] Elnaz Barshan, Marc-Etienne Brunet, and Gintare Karolina Dziugaite. 2020. Relatif: Identifying explanatory training samples via relative influence.
In International Conference on Artificial Intelligence and Statistics. PMLR, 1899–1909.
[29] R. Barth, J. IJsselmuiden, J. Hemming, and E.J. Van Henten. 2018. Data synthesis methods for semantic segmentation in agriculture: A Capsicum
annuum dataset. Computers and Electronics in Agriculture 144 (2018), 284–296. https://doi.org/10.1016/j.compag.2017.12.001
[30] Jasmijn Bastings, Wilker Aziz, and Ivan Titov. 2019. Interpretable neural predictions with differentiable binary variables. arXiv preprint
arXiv:1905.08160 (2019).
[31] David Bau, Bolei Zhou, Aditya Khosla, Aude Oliva, and Antonio Torralba. 2017. Network dissection: Quantifying interpretability of deep visual
representations. In Proceedings of the IEEE conference on computer vision and pattern recognition. 6541–6549.
[32] Andrew M Beacham, Laura H Vickers, and James M Monaghan. 2019. Vertical farming: a summary of approaches to growing skywards. The
Journal of Horticultural Science and Biotechnology 94, 3 (2019), 277–283.
[33] Khadija Benis, Christoph Reinhart, and Paulo Ferrão. 2017. Development of a simulation-based decision support workflow for the implementation
of Building-Integrated Agriculture (BIA) in urban contexts. Journal of cleaner production 147 (2017), 589–602.
[34] Kurt Benke and Bruce Tomkins. 2017. Future food-production systems: vertical farming and controlled-environment agriculture. Sustainability:
Science, Practice and Policy 13, 1 (2017), 13–26.
[35] Daniel Berckmans. 2017. General introduction to precision livestock farming. Animal Frontiers 7, 1 (2017), 6–11.
[36] Pieter M. Blok, Eldert J. van Henten, Frits K. van Evert, and Gert Kootstra. 2021. Image-based size estimation of broccoli heads under varying
degrees of occlusion. Biosystems Engineering 208 (2021), 213–233. https://doi.org/10.1016/j.biosystemseng.2021.06.001
[37] CH Bock, PE Parker, AZ Cook, and TR Gottwald. 2008. Visual rating and the use of image analysis for assessing different symptoms of citrus
canker on grapefruit leaves. Plant Disease 92, 4 (2008), 530–541.
[38] Edson Bollis, Helio Pedrini, and Sandra Avila. 2020. Weakly supervised learning guided by activation mapping applied to a novel citrus pest
benchmark. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. 70–71.
[39] Charles A Bouman and Michael Shapiro. 1994. A multiscale random field model for Bayesian image segmentation. IEEE Transactions on image
processing 3, 2 (1994), 162–177.
[40] Jonathan Brophy and Daniel Lowd. 2020. TREX: Tree-Ensemble Representer-Point Explanations. arXiv preprint arXiv:2009.05530 (2020).
[41] Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry,
Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey
Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin Chess, Jack Clark, Christopher Berner,
Sam McCandlish, Alec Radford, Ilya Sutskever, and Dario Amodei. 2020. Language Models are Few-Shot Learners. arXiv 2005.14165 (2020).
[42] Zhaowei Cai and Nuno Vasconcelos. 2018. Cascade R-CNN: Delving Into High Quality Object Detection. In Proceedings of the IEEE Conference on
Computer Vision and Pattern Recognition (CVPR).
[43] Joao Carreira and Andrew Zisserman. 2017. Quo vadis, action recognition? a new model and the kinetics dataset. In proceedings of the IEEE
Conference on Computer Vision and Pattern Recognition. 6299–6308.
[44] Rich Caruana. 1997. Multitask learning. Machine learning 28, 1 (1997), 41–75.
[45] Diogo V Carvalho, Eduardo M Pereira, and Jaime S Cardoso. 2019. Machine learning interpretability: A survey on methods and metrics. Electronics
8, 8 (2019), 832.
[46] Supawadee Chaivivatrakul, Jednipat Moonrinta, and Matthew N Dailey. 2010. Towards Automated Crop Yield Estimation-Detection and 3D
Reconstruction of Pineapples in Video Sequences.. In VISAPP (1). 180–183.
[47] Hila Chefer, Shir Gur, and Lior Wolf. 2021. Transformer interpretability beyond attention visualization. In Proceedings of the IEEE/CVF Conference
on Computer Vision and Pattern Recognition. 782–791.
[48] Howard Chen, Jacqueline He, Karthik Narasimhan, and Danqi Chen. 2022. Can Rationalization Improve Robustness? arXiv preprint
arXiv:2204.11790 (2022).
[49] Hanjie Chen, Guangtao Zheng, and Yangfeng Ji. 2020. Generating hierarchical explanations on text classification via feature interaction detection.
arXiv preprint arXiv:2004.02015 (2020).
[50] Kai Chen, Jiangmiao Pang, Jiaqi Wang, Yu Xiong, Xiaoxiao Li, Shuyang Sun, Wansen Feng, Ziwei Liu, Jianping Shi, Wanli Ouyang, et al. 2019.
Hybrid task cascade for instance segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 4974–4983.
[51] Long Chen, Martin Strauch, and Dorit Merhof. 2019. Instance Segmentation of Biomedical Images with an Object-aware Em-
bedding Learned with Local Constraints. In International Conference on Medical Image Computing and Computer-Assisted Intervention.
https://doi.org/10.48550/ARXIV.2004.09821
[52] Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos, Kevin Murphy, and Alan L Yuille. 2017. Deeplab: Semantic image segmentation with
deep convolutional nets, atrous convolution, and fully connected crfs. IEEE transactions on pattern analysis and machine intelligence 40, 4 (2017),
834–848.
[53] Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam. 2018. Encoder-decoder with atrous separable convolution
for semantic image segmentation. In Proceedings of the European conference on computer vision (ECCV). 801–818.
[54] Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. 2020. A Simple Framework for Contrastive Learning of Visual Repre-
sentations. arXiv 2002.05709 (2020).

20
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

[55] Ting Chen, Saurabh Saxena, Lala Li, David J Fleet, and Geoffrey Hinton. 2021. Pix2seq: A language modeling framework for object detection.
arXiv preprint arXiv:2109.10852 (2021).
[56] Yuanyuan Chen, Boyang Li, Han Yu, Pengcheng Wu, and Chunyan Miao. 2021. Hydra: Hypergradient data relevance analysis for interpreting
deep neural networks. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 35. 7081–7089.
[57] Yi-Ting Chen, Xiaokai Liu, and Ming-Hsuan Yang. 2015. Multi-instance object segmentation with occlusion handling. In 2015 IEEE Conference on
Computer Vision and Pattern Recognition (CVPR). 3470–3478. https://doi.org/10.1109/CVPR.2015.7298969
[58] Bowen Cheng, Maxwell D. Collins, Yukun Zhu, Ting Liu, Thomas S. Huang, Hartwig Adam, and Liang-Chieh Chen. 2020. Panoptic-DeepLab: A
Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition
(CVPR). 12472–12482. https://doi.org/10.1109/CVPR42600.2020.01249
[59] De Cheng, Tongliang Liu, Yixiong Ning, Nannan Wang, Bo Han, Gang Niu, Xinbo Gao, and Masashi Sugiyama. 2022. Instance-Dependent Label-
Noise Learning with Manifold-Regularized Transition Matrix Estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern
Recognition. 16630–16639.
[60] Jaemin Cho, Jie Lei, Hao Tan, and Mohit Bansal. 2021. Unifying vision-and-language tasks via text generation. In International Conference on
Machine Learning. PMLR, 1931–1942.
[61] Xiangxiang Chu, Zhi Tian, Yuqing Wang, Bo Zhang, Haibing Ren, Xiaolin Wei, Huaxia Xia, and Chunhua Shen. 2021. Twins: Revisiting the Design
of Spatial Attention in Vision Transformers. arXiv 2104.13840 (2021).
[62] Dan Ciresan, Alessandro Giusti, Luca Gambardella, and Jürgen Schmidhuber. 2012. Deep Neural Networks Segment Neuronal Membranes in
Electron Microscopy Images. In Advances in Neural Information Processing Systems, F. Pereira, C.J. Burges, L. Bottou, and K.Q. Weinberger (Eds.),
Vol. 25. Curran Associates, Inc. https://proceedings.neurips.cc/paper/2012/file/459a4ddcb586f24efd9395aa7662bc7c-Paper.pdf
[63] Ekin D. Cubuk, Barret Zoph, Jonathon Shlens, and Quoc V. Le. 2019. RandAugment: Practical automated data augmentation with a reduced search
space. arXiv 1909.13719 (2019).
[64] Jifeng Dai, Kaiming He, Yi Li, Shaoqing Ren, and Jian Sun. 2016. Instance-Sensitive Fully Convolutional Networks. In Computer Vision – ECCV
2016, Bastian Leibe, Jiri Matas, Nicu Sebe, and Max Welling (Eds.). Springer International Publishing, Cham, 534–549.
[65] Jifeng Dai, Haozhi Qi, Yuwen Xiong, Yi Li, Guodong Zhang, Han Hu, and Yichen Wei. 2017. Deformable Convolutional Networks. In 2017 IEEE
International Conference on Computer Vision (ICCV). 764–773.
[66] Zihang Dai, Hanxiao Liu, Quoc V Le, and Mingxing Tan. 2021. CoAtNet: Marrying Convolution and Attention for All Data Sizes. arXiv preprint
arXiv:2106.04803 (2021).
[67] Jim Davis, Tong Liang, James Enouen, and Roman Ilin. 2019. Hierarchical semantic labeling with adaptive confidence. In International Symposium
on Visual Computing. Springer, 169–183.
[68] Bert De Brabandere, Davy Neven, and Luc Van Gool. 2017. Semantic Instance Segmentation with a Discriminative Loss Function. CVPR 2017
Workshop on Deep Learning for Robotic Vision. https://arxiv.org/abs/1708.02551
[69] Jiankang Deng, Jia Guo, Niannan Xue, and Stefanos Zafeiriou. 2019. Arcface: Additive angular margin loss for deep face recognition. In CVPR.
4690–4699.
[70] Dickson Despommier. 2010. The vertical farm: feeding the world in the 21st century. Macmillan.
[71] Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language
Understanding. arXiv 1810.04805 (2019).
[72] Amit Dhurandhar, Pin-Yu Chen, Ronny Luss, Chun-Chen Tu, Paishun Ting, Karthikeyan Shanmugam, and Payel Das. 2018. Explanations based
on the missing: Towards contrastive explanations with pertinent negatives. Advances in neural information processing systems 31 (2018).
[73] Zhipeng Ding, Xu Han, Peirong Liu, and Marc Niethammer. 2021. Local Temperature Scaling for Probability Calibration. In Proceedings of the
IEEE/CVF International Conference on Computer Vision (ICCV). 6889–6899.
[74] Pelin Dogan, Boyang Li, Leonid Sigal, and Markus Gross. 2018. A Neural Multi-sequence Alignment TeCHnique (NeuMATCH). In The Conference
on Computer Vision and Pattern Recognition (CVPR).
[75] Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias
Minderer, Georg Heigold, Sylvain Gelly, et al. 2021. An image is worth 16x16 words: Transformers for image recognition at scale. In International
Conference on Learning Representations.
[76] Kaiwen Duan, Song Bai, Lingxi Xie, Honggang Qi, Qingming Huang, and Qi Tian. 2019. CenterNet: Keypoint Triplets for Object Detection. In
Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV).
[77] Dumitru Erhan, Yoshua Bengio, Aaron Courville, and Pascal Vincent. 2009. Visualizing higher-layer features of a deep network. University of
Montreal 1341, 3 (2009), 1.
[78] Clement Farabet, Camille Couprie, Laurent Najman, and Yann LeCun. 2013. Learning Hierarchical Features for Scene Labeling. IEEE Transactions
on Pattern Analysis and Machine Intelligence 35, 8 (2013), 1915–1929. https://doi.org/10.1109/TPAMI.2012.231
[79] Konstantinos P Ferentinos. 2018. Deep learning models for plant disease detection and diagnosis. Computers and electronics in agriculture 145
(2018), 311–318.
[80] Roemi Fernández, Carlota Salinas, Héctor Montes, and Javier Sarria. 2014. Multisensory system for fruit harvesting robots. Experimental testing
in natural scenarios and with different kinds of crops. Sensors 14, 12 (2014), 23885–23904.

21
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

[81] Ruth Fong and Andrea Vedaldi. 2018. Net2vec: Quantifying and explaining how concepts are encoded by filters in deep neural networks. In
Proceedings of the IEEE conference on computer vision and pattern recognition. 8730–8738.
[82] Logan Frank, Christopher Wiegman, Jim Davis, and Scott Shearer. 2021. Confidence-Driven Hierarchical Classification of Cultivated Plant Stresses.
In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 2503–2512.
[83] Evan DG Fraser and Malcolm Campbell. 2019. Agriculture 5.0: reconciling production with planetary health. One Earth 1, 3 (2019), 278–280.
[84] Cheng-Yang Fu, Wei Liu, Ananth Ranga, Ambrish Tyagi, and Alexander C. Berg. 2017. DSSD : Deconvolutional Single Shot Detector. arXiv
Preprint 1701.06659 (2017).
[85] LiMin Fu. 1991. Rule Learning by Searching on Adapted Nets.. In AAAI, Vol. 91. 590–595.
[86] Alvaro Fuentes, Sook Yoon, Sang Cheol Kim, and Dong Sun Park. 2017. A robust deep-learning-based detector for real-time tomato plant diseases
and pests recognition. Sensors 17, 9 (2017), 2022.
[87] Pedro Galeano, Esdras Joseph, and Rosa E Lillo. 2015. The Mahalanobis distance for functional data with applications to classification. Techno-
metrics 57, 2 (2015), 281–291.
[88] Damien Garreau and Dina Mardaoui. 2021. What does LIME really see in images?. In Proceedings of the 38th International Confer-
ence on Machine Learning (Proceedings of Machine Learning Research, Vol. 139), Marina Meila and Tong Zhang (Eds.). PMLR, 3620–3629.
https://proceedings.mlr.press/v139/garreau21a.html
[89] Damien Garreau and Ulrike von Luxburg. 2020. Explaining the Explainer: A First Theoretical Analysis of LIME. In Proceedings of the Twenty Third
International Conference on Artificial Intelligence and Statistics (Proceedings of Machine Learning Research, Vol. 108), Silvia Chiappa and Roberto
Calandra (Eds.). PMLR, 1287–1296. https://proceedings.mlr.press/v108/garreau20a.html
[90] Yuanyue Ge, Ya Xiong, Gabriel Lins Tenorio, and Pål Johan From. 2019. Fruit Localization and Environment Perception for Strawberry Harvesting
Robots. IEEE Access 7 (2019), 147642–147652. https://doi.org/10.1109/ACCESS.2019.2946369
[91] Jordi Gené-Mola, Ricardo Sanz-Cortiella, Joan R Rosell-Polo, Josep-Ramon Morros, Javier Ruiz-Hidalgo, Verónica Vilaplana, and Eduard Gregorio.
2020. Fruit detection and 3D location using instance segmentation neural networks and structure-from-motion photogrammetry. Computers and
Electronics in Agriculture 169 (2020), 105165.
[92] Daan de Geus, Panagiotis Meletis, Chenyang Lu, Xiaoxiao Wen, and Gijs Dubbelman. 2021. Part-aware Panoptic Segmentation. In 2021 IEEE/CVF
Conference on Computer Vision and Pattern Recognition (CVPR). 5481–5490. https://doi.org/10.1109/CVPR46437.2021.00544
[93] Ross Girshick. 2015. Fast R-CNN. In ICCV.
[94] Andrea Gomez-Zavaglia, Juan Carlos Mejuto, and Jesus Simal-Gandara. 2020. Mitigation of emerging implications of climate change on food
production systems. Food Research International 134 (2020), 109256.
[95] Stephen Gould, Richard Fulton, and Daphne Koller. 2009. Decomposing a scene into geometric and semantically consistent regions. In 2009 IEEE
12th International Conference on Computer Vision. 1–8. https://doi.org/10.1109/ICCV.2009.5459211
[96] Priya Goyal, Piotr Dollár, Ross Girshick, Pieter Noordhuis, Lukasz Wesolowski, Aapo Kyrola, Andrew Tulloch, Yangqing Jia, and Kaiming He.
2018. Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour. arXiv 1706.02677 (2018).
[97] Yash Goyal, Ziyan Wu, Jan Ernst, Dhruv Batra, Devi Parikh, and Stefan Lee. 2019. Counterfactual visual explanations. In International Conference
on Machine Learning. PMLR, 2376–2384.
[98] Riccardo Gozzovelli, Benjamin Franchetti, Malik Bekmurat, and Fiora Pirri. 2021. Tip-burn stress detection of lettuce canopy grown in Plant
Factories. In Proceedings of the IEEE/CVF International Conference on Computer Vision. 1259–1268.
[99] Chuan Guo, Geoff Pleiss, Yu Sun, and Kilian Q. Weinberger. 2017. On Calibration of Modern Neural Networks. In Proceedings of the 34th Interna-
tional Conference on Machine Learning - Volume 70 (Sydney, NSW, Australia) (ICML’17). JMLR.org, 1321–1330.
[100] Xu Guo, Boyang Li, Han Yu, and Chunyan Miao. 2021. Latent-Optimized Adversarial Neural Transfer for Sarcasm Detection. In 2021 Annual
Conference of the North American Chapter of the Association for Computational Linguistics (NAACL-HLT 2021).
[101] Xu Guo, Boyang Li, Han Yu, and Chunyan Miao. 2021. Latent-Optimized Adversarial Neural Transfer for Sarcasm Detec-
tion. In 2021 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL-HLT 2021).
http://www.boyangli.org/paper/XuGuo-NAACL-2021.pdf
[102] Tanmay Gupta, Amita Kamath, Aniruddha Kembhavi, and Derek Hoiem. 2022. Towards General Purpose Vision Systems: An End-to-End Task-
Agnostic Vision-Language Architecture. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 16399–16409.
[103] Felix G Gustafson and Elnore Stoldt. 1936. Some relations between leaf area and fruit size in tomatoes. Plant Physiology 11, 2 (1936), 445.
[104] Michael Halstead, Christopher McCool, Simon Denman, Tristan Perez, and Clinton Fookes. 2018. Fruit quantity and ripeness estimation using a
robotic vision system. IEEE robotics and automation LETTERS 3, 4 (2018), 2995–3002.
[105] Tengda Han, Weidi Xie, and Andrew Zisserman. 2020. Self-supervised co-training for video representation learning. Advances in Neural Informa-
tion Processing Systems 33 (2020), 5679–5690.
[106] X Hao, X Guo, J Zheng, L Celeste, S Kholsa, and X Chen. 2015. Response of greenhouse tomato to different vertical spectra of LED lighting
under overhead high pressure sodium and plasma lighting. In International Symposium on New Technologies and Management for Greenhouses-
GreenSys2015 1170. 1003–1110.
[107] Xiuming Hao and Athanasios P Papadopoulos. 1999. Effects of supplemental lighting and cover materials on growth, photosynthesis, biomass
partitioning, early yield and quality of greenhouse cucumber. Scientia Horticulturae 80, 1-2 (1999), 1–18.

22
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

[108] Bharath Hariharan, Pablo Arbeláez, Ross Girshick, and Jitendra Malik. 2014. Simultaneous Detection and Segmentation. In Computer Vision –
ECCV 2014, David Fleet, Tomas Pajdla, Bernt Schiele, and Tinne Tuytelaars (Eds.). Springer International Publishing, Cham, 297–312.
[109] Marton Havasi, Rodolphe Jenatton, Stanislav Fort, Jeremiah Zhe Liu, Jasper Snoek, Balaji Lakshminarayanan, Andrew M. Dai, and Dustin Tran.
2021. Training independent subnetworks for robust prediction. In ICLR.
[110] Zeeshan Hayder, Xuming He, and Mathieu Salzmann. 2016. Boundary-aware Instance Segmentation. arXiv Preprint 1612.03129 (2016).
https://doi.org/10.48550/ARXIV.1612.03129
[111] Kaiming He, Georgia Gkioxari, Piotr Dollár, and Ross Girshick. 2017. Mask R-CNN. In ICCV.
[112] Kaiming He, Georgia Gkioxari, Piotr Dollár, and Ross Girshick. 2018. Mask R-CNN. arXiv 1703.06870 (2018).
[113] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2015. Deep Residual Learning for Image Recognition. arXiv 1512.03385 (2015).
[114] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2015. Deep Residual Learning for Image Recognition. (2015). arXiv:1512.03385 [cs.CV]
[115] Leilei He, Wentai Fang, Guanao Zhao, Zhenchao Wu, Longsheng Fu, Rui Li, Yaqoob Majeed, and Jaspreet Dhupia. 2022. Fruit yield prediction and
estimation in orchards: A state-of-the-art comprehensive review for both direct and indirect methods. Computers and Electronics in Agriculture
195 (2022), 106812.
[116] Katherine L. Hermann and Andrew K. Lampinen. 2020. What shapes feature representations? Exploring datasets, architectures, and training.
arXiv2006.12433 (2020).
[117] Yann N. Dauphin David Lopez-Paz Hongyi Zhang, Moustapha Cisse. 2018. mixup: Beyond Empirical Risk Minimization. In International Conference
on Learning Representations. https://openreview.net/forum?id=r1Ddp1-Rb
[118] Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, and Hartwig Adam. 2017.
MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications. arXiv 1704.04861 (2017).
[119] Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, and Hartwig Adam. 2017.
MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications. arXiv PrePrint 1704.04861 (2017).
[120] Chunhua Hu, Xuan Liu, Zhou Pan, and Pingping Li. 2019. Automatic detection of single ripe tomato on plant combining faster R-CNN and
intuitionistic fuzzy set. IEEE Access 7 (2019), 154683–154696.
[121] Xiaoping Huang, Zelin Hu, Xiaorun Wang, Xuanjiang Yang, Jian Zhang, and Daoling Shi. 2019. An Improved Single Shot Multibox Detector
Method Applied in Body Condition Score for Dairy Cows. Animals 9, 7 (2019). https://doi.org/10.3390/ani9070470
[122] Yo-Ping Huang, Tzu-Hao Wang, and Haobijam Basanta. 2020. Using fuzzy mask R-CNN model to automatically identify tomato ripeness. IEEE
Access 8 (2020), 207672–207682.
[123] Wallace E Huffman. 2012. The status of labor-saving mechanization in US fruit and vegetable harvesting. Choices 27, 316-2016-6262 (2012).
[124] David Hughes, Marcel Salathé, et al. 2015. An open access repository of images on plant health to enable the development of mobile disease
diagnostics. arXiv preprint arXiv:1511.08060 (2015).
[125] Eyke Hüllermeier and Willem Waegeman. 2021. Aleatoric and epistemic uncertainty in machine learning: An introduction to concepts and
methods. Machine Learning 110, 3 (2021), 457–506.
[126] Sarah Ibrahimi, Arnaud Sors, Rafael Sampaio de Rezende, and Stéphane Clinchant. 2022. Learning with label noise for image retrieval by selecting
interactions. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 2181–2190.
[127] Joana Iljazi. 2017. Deep learning for image-based prediction of plant growth in City Farms. (2017).
[128] David Ireri, Eisa Belal, Cedric Okinda, Nelson Makange, and Changying Ji. 2019. A computer vision system for defect discrimination and grading
in tomatoes using machine learning and image processing. Artificial Intelligence in Agriculture 2 (2019), 28–37.
[129] Jörn-Henrik Jacobsen, Arnold Smeulders, and Edouard Oyallon. 2018. i-RevNet: Deep Invertible Networks. In International Conference on Learning
Representations (ICLR).
[130] Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew
Brock, Evan Shelhamer, et al. 2021. Perceiver io: A general architecture for structured inputs & outputs. arXiv preprint arXiv:2107.14795 (2021).
[131] Lu Jiang, Mason Liu Di Huang, and Weilong Yang. 2020. Beyond synthetic noise: Deep learning on controlled noisy labels. In ICML.
[132] Amita Kamath, Christopher Clark, Tanmay Gupta, Eric Kolve, Derek Hoiem, and Aniruddha Kembhavi. 2022. Webly Supervised Concept Expan-
sion for General Purpose Vision Models. arXiv preprint arXiv:2202.02317 (2022).
[133] Kentaro Kanamori, Takuya Takagi, Ken Kobayashi, and Hiroki Arimura. 2020. DACE: Distribution-Aware Counterfactual Explanation by Mixed-
Integer Linear Optimization.. In IJCAI. 2855–2862.
[134] Ramesh Kestur, Avadesh Meduri, and Omkar Narasipura. 2019. MangoNet: A deep semantic segmentation architecture for a method to detect
and count mangoes in an open orchard. Engineering Applications of Artificial Intelligence 77 (2019), 59–69.
[135] Alexander Kirillov, Kaiming He, Ross Girshick, Carsten Rother, and Piotr Dollar. 2019. Panoptic Segmentation. In Proceedings of the IEEE/CVF
Conference on Computer Vision and Pattern Recognition (CVPR).
[136] Pang Wei Koh and Percy Liang. 2017. Understanding black-box predictions via influence functions. In International conference on machine learning.
PMLR, 1885–1894.
[137] Anand Koirala, KB Walsh, Zhenglin Wang, and C McCarthy. 2019. Deep learning for real-time fruit detection and orchard fruit load estimation:
Benchmarking of ‘MangoYOLO’. Precision Agriculture 20, 6 (2019), 1107–1135.
[138] Tao Kong, Fuchun Sun, Huaping Liu, Yuning Jiang, Lei Li, and Jianbo Shi. 2020. FoveaBox: Beyound Anchor-Based Object Detection. IEEE
Transactions on Image Processing 29 (2020), 7389–7398. https://doi.org/10.1109/TIP.2020.3002345
23
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

[139] Dean A Kopsell, Carl E Sams, and Robert C Morrow. 2015. Blue wavelengths from LED lighting increase nutritionally important metabolites in
specialty crops. HortScience 50, 9 (2015), 1285–1288.
[140] Maxim S. Kovalev, Lev V. Utkin, and Ernest M. Kasimov. 2020. SurvLIME: A method for explaining machine learning survival models. Knowledge-
Based Systems 203 (2020), 106164. https://doi.org/10.1016/j.knosys.2020.106164
[141] R Krishnamurthy. 2014. vertical farming: Singapore’s Solution to feed the local urban Population. Permaculture Research Institute (2014).
[142] Alex Krizhevsky. 2014. One weird trick for parallelizing convolutional neural networks. arXiv preprint arXiv:1404.5997 (2014).
[143] Alex Krizhevsky, Ilya Sutskever, and Geoffrey E. Hinton. 2012. ImageNet Classification with Deep Convolutional Neural Networks. In Proceedings
of the 25th International Conference on Neural Information Processing Systems (Lake Tahoe, Nevada) (NIPS’12). Curran Associates Inc., Red Hook,
NY, USA, 1097–1105.
[144] Ferhat Kurtulmus, Won Suk Lee, and Ali Vardar. 2014. Immature peach detection in colour images acquired in natural illumination conditions
using statistical classifiers and neural network. Precision agriculture 15, 1 (2014), 57–79.
[145] L’ubor Ladický, Chris Russell, Pushmeet Kohli, and Philip H.S. Torr. 2009. Associative hierarchical CRFs for object class image segmentation. In
2009 IEEE 12th International Conference on Computer Vision. 739–746. https://doi.org/10.1109/ICCV.2009.5459248
[146] Dana Lahat, Tülay Adali, and Christian Jutten. 2015. Multimodal data fusion: an overview of methods, challenges, and prospects. Proc. IEEE 103,
9 (2015), 1449–1477.
[147] Peter D Lancashire, Hermann Bleiholder, T van den Boom, P Langelüddeke, Reinhold Stauss, Elfriede Weber, and A Witzenberger. 1991. A uniform
decimal code for growth stages of crops and weeds. Annals of applied Biology 119, 3 (1991), 561–601.
[148] Gustav Larsson, Michael Maire, and Gregory Shakhnarovich. 2017. FractalNet: Ultra-Deep Neural Networks without Residuals. In ICLR.
[149] Hei Law and Jia Deng. 2018. CornerNet: Detecting Objects as Paired Keypoints. In Proceedings of the European Conference on Computer Vision
(ECCV).
[150] Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. 2015. Deep learning. Nature 521 (2015), 436–444.
[151] Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner. 1998. Gradient-based learning applied to document recognition. Proc. IEEE 86, 11 (1998), 2278–2324.
https://doi.org/10.1109/5.726791
[152] Joon-Woo Lee, Taewon Moon, and Jung-Eek Son. 2021. Development of Growth Estimation Algorithms for Hydroponic Bell Peppers Using
Recurrent Neural Networks. Horticulturae 7, 9 (2021). https://doi.org/10.3390/horticulturae7090284
[153] Tao Lei, Regina Barzilay, and Tommi Jaakkola. 2016. Rationalizing neural predictions. arXiv preprint arXiv:1606.04155 (2016).
[154] Han Li, Won Suk Lee, and Ku Wang. 2016. Immature green citrus fruit detection and counting based on fast normalized cross correlation (FNCC)
using natural outdoor colour images. Precision Agriculture 17, 6 (2016), 678–697.
[155] Kai Li, Martin Renqiang Min, and Yun Fu. 2019. Rethinking zero-shot learning: A conditional visual classification perspective. In Proceedings of
the IEEE/CVF international conference on computer vision. 3583–3592.
[156] Yanwei Li, Xinze Chen, Zheng Zhu, Lingxi Xie, Guan Huang, Dalong Du, and Xingang Wang. 2019. Attention-Guided Unified
Network for Panoptic Segmentation. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 7019–7028.
https://doi.org/10.1109/CVPR.2019.00719
[157] Yuanzhi Li and Yingyu Liang. 2018. Learning Overparameterized Neural Networks via Stochastic Gradient Descent on Structured Data. In
Advances in Neural Information Processing Systems, S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett (Eds.),
Vol. 31. Curran Associates, Inc. https://proceedings.neurips.cc/paper/2018/file/54fe976ba170c19ebae453679b362263-Paper.pdf
[158] Yi Li, Haozhi Qi, Jifeng Dai, Xiangyang Ji, and Yichen Wei. [n. d.]. Fully Convolutional Instance-Aware Semantic Segmentation. In 2017 IEEE
Conference on Computer Vision and Pattern Recognition (CVPR).
[159] Yanfen Li, Hanxiang Wang, L Minh Dang, Abolghasem Sadeghi-Niaraki, and Hyeonjoon Moon. 2020. Crop pest recognition in natural scenes
using convolutional neural networks. Computers and Electronics in Agriculture 169 (2020), 105174.
[160] Yang Li and Jiachen Yang. 2021. Meta-learning baselines and database for few-shot classification in agriculture. Computers and Electronics in
Agriculture 182 (2021), 106055.
[161] Zhizhong Li and Derek Hoiem. 2020. Improving Confidence Estimates for Unfamiliar Examples. In 2020 IEEE/CVF Conference on Computer Vision
and Pattern Recognition (CVPR). 2683–2692. https://doi.org/10.1109/CVPR42600.2020.00276
[162] Guichao Lin, Yunchao Tang, Xiangjun Zou, Juntao Xiong, and Yamei Fang. 2020. Color-, depth-, and shape-based 3D fruit detection. Precision
Agriculture 21, 1 (2020), 1–17.
[163] Guichao Lin, Yunchao Tang, Xiangjun Zou, Juntao Xiong, and Jinhui Li. 2019. Guava detection and pose estimation using a low-cost RGB-D
sensor in the field. Sensors 19, 2 (2019), 428.
[164] Tsung-Yi Lin, Piotr Dollar, Ross Girshick, Kaiming He, Bharath Hariharan, and Serge Belongie. 2017. Feature Pyramid Networks for Object
Detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR).
[165] Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollar. 2017. Focal Loss for Dense Object Detection. In Proceedings of the IEEE
International Conference on Computer Vision (ICCV).
[166] Chang Liu, Han Yu, Boyang Li, Zhiqi Shen, Zhanning Gao, Peiran Ren, Xuansong Xie, Lizhen Cui, and Chunyan Miao. 2021. Noise-resistant
Deep Metric Learning with Ranking-based Instance Selection. In The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).
http://www.boyangli.org/paper/ChangLiu-CVPR-2021.pdf

24
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

[167] Huanyu Liu, Chao Peng, Changqian Yu, Jingbo Wang, Xu Liu, Gang Yu, and Wei Jiang. 2019. An End-To-End Network for Panoptic Segmentation.
In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 6165–6174. https://doi.org/10.1109/CVPR.2019.00633
[168] Jialun Liu, Yifan Sun, Chuchu Han, Zhaopeng Dou, and Wenhui Li. 2020. Deep Representation Learning on Long-tailed Data: A Learnable
Embedding Augmentation Perspective. In CVPR. 2970–2979.
[169] Qiuhua Liu, Xuejun Liao, and Lawrence Carin. 2007. Semi-Supervised Multitask Learning. In Advances in Neu-
ral Information Processing Systems, J. Platt, D. Koller, Y. Singer, and S. Roweis (Eds.), Vol. 20. Curran Associates, Inc.
https://proceedings.neurips.cc/paper/2007/file/a34bacf839b923770b2c360eefa26748-Paper.pdf
[170] Tian-Hu Liu, Reza Ehsani, Arash Toudeshki, Xiang-Jun Zou, and Hong-Jun Wang. 2018. Detection of citrus fruit and tree trunks in natural
environments using a multi-elliptical boundary model. Computers in Industry 99 (2018), 9–16.
[171] Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexander C. Berg. 2016. SSD: Single Shot
MultiBox Detector. In Computer Vision – ECCV 2016, Bastian Leibe, Jiri Matas, Nicu Sebe, and Max Welling (Eds.). 21–37.
[172] Yang Liu, Duolin Wang, Fei He, Juexin Wang, Trupti Joshi, and Dong Xu. 2019. Phenotype Prediction and Genome-Wide Association Study Using
Deep Convolutional Neural Network of Soybean. Frontiers in Genetics 10 (2019), 1091. https://doi.org/10.3389/fgene.2019.01091
[173] Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. 2021. Swin transformer: Hierarchical vision
transformer using shifted windows. In Proceedings of the IEEE/CVF International Conference on Computer Vision. 10012–10022.
[174] Jonathan Long, Evan Shelhamer, and Trevor Darrell. 2015. Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE/CVF
conference on computer vision and pattern recognition (CVPR). 3431–3440.
[175] Jiasen Lu, Christopher Clark, Rowan Zellers, Roozbeh Mottaghi, and Aniruddha Kembhavi. 2022. Unified-IO: A Unified Model for Vision, Language,
and Multi-Modal Tasks. arXiv preprint arXiv:2206.08916 (2022).
[176] Jie-Yan Lu, Chung-Liang Chang, and Yan-Fu Kuo. 2019. Monitoring growth rate of lettuce using deep convolutional neural networks. In 2019
ASABE Annual International Meeting. American Society of Agricultural and Biological Engineers, 1.
[177] Shilei Lyu, Yawen Zhao, Ruiyao Li, Zhen Li, Renjie Fan, and Qiafeng Li. 2022. Embedded Sensing System for Recognizing Citrus Flowers Using
Cascaded Fusion YOLOv4-CF+ FPGA. Sensors 22, 3 (2022), 1255.
[178] Juncheng Ma, Keming Du, Lingxian Zhang, Feixiang Zheng, Jinxiang Chu, and Zhongfu Sun. 2017. A segmentation method for greenhouse
vegetable foliar disease spots images using color information and region growing. Computers and Electronics in Agriculture 142 (2017), 110–117.
[179] Juncheng Ma, Keming Du, Feixiang Zheng, Lingxian Zhang, Zhihong Gong, and Zhongfu Sun. 2018. A recognition method for cucumber diseases
using leaf symptom images based on deep convolutional neural network. Computers and electronics in agriculture 154 (2018), 18–24.
[180] Aravindh Mahendran and Andrea Vedaldi. 2015. Understanding deep image representations by inverting them. In Proceedings of the IEEE conference
on computer vision and pattern recognition. 5188–5196.
[181] Alireza Mehrtash, William M. Wells, Clare M. Tempany, Purang Abolmaesumi, and Tina Kapur. 2020. Confidence Calibration and Pre-
dictive Uncertainty Estimation for Deep Medical Image Segmentation. IEEE Transactions on Medical Imaging 39, 12 (2020), 3868–3878.
https://doi.org/10.1109/TMI.2020.3006437
[182] Massimo Minervini, Andreas Fischbach, Hanno Scharr, and Sotirios A Tsaftaris. 2016. Finely-grained annotated datasets for image-based plant
phenotyping. Pattern recognition letters 81 (2016), 80–89.
[183] Lj Miranda. [n. d.]. Towards data-centric machine learning: a short review. https://ljvmiranda921.github.io/notebook/2021/07/30/data-centric-ml/
[184] Christoph Molnar. 2020. Interpretable machine learning. https://christophm.github.io/interpretable-ml-book/
[185] Grégoire Montavon, Alexander Binder, Sebastian Lapuschkin, Wojciech Samek, and Klaus-Robert Müller. 2019. Layer-wise relevance propagation:
an overview. Explainable AI: interpreting, explaining and visualizing deep learning (2019), 193–209.
[186] Jednipat Moonrinta, Supawadee Chaivivatrakul, Matthew N Dailey, and Mongkol Ekpanyapong. 2010. Fruit detection, tracking, and 3D recon-
struction for crop mapping and yield estimation. In 2010 11th International Conference on Control Automation Robotics & Vision. IEEE, 1181–1186.
[187] Alexander Mordvintsev, Christopher Olah, and Mike Tyka. 2015. Inceptionism: Going deeper into neural networks. (2015).
[188] Jishnu Mukhoti, Viveka Kulharia, Amartya Sanyal, Stuart Golodetz, Philip Torr, and Puneet Dokania. 2020. Calibrating Deep Neural Networks
using Focal Loss. In Advances in Neural Information Processing Systems, H. Larochelle, M. Ranzato, R. Hadsell, M.F. Balcan, and H. Lin (Eds.), Vol. 33.
Curran Associates, Inc., 15288–15299. https://proceedings.neurips.cc/paper/2020/file/aeb7b30ef1d024a76f21a1d40e30c302-Paper.pdf
[189] Rafael Müller, Simon Kornblith, and Geoffrey E Hinton. 2019. When does label smoothing help?. In Advances in Neural Information Pro-
cessing Systems, H. Wallach, H. Larochelle, A. Beygelzimer, F. d'Alché-Buc, E. Fox, and R. Garnett (Eds.), Vol. 32. Curran Associates, Inc.
https://proceedings.neurips.cc/paper/2019/file/f1748d6b0fd9d439f71450117eba2725-Paper.pdf
[190] Joanna M Nassar, Sherjeel M Khan, Diego Rosas Villalva, Maha M Nour, Amani S Almuslem, and Muhammad M Hussain. 2018. Compliant plant
wearables for localized microclimate and plant growth monitoring. npj Flexible Electronics 2, 1 (2018), 1–12.
[191] Davy Neven, Bert De Brabandere, Marc Proesmans, and Luc Van Gool. 2019. Instance Segmentation by Jointly Optimizing Spatial
Embeddings and Clustering Bandwidth. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 8829–8837.
https://doi.org/10.1109/CVPR.2019.00904
[192] Andrew Ng. [n. d.]. A Chat with Andrew on MLOps: From Model-centric to Data-centric AI. https://www.youtube.com/watch?v=06-AZXmwHjo
[193] Canh Nguyen, Vasit Sagan, Matthew Maimaitiyiming, Maitiniyazi Maimaitijiang, Sourav Bhadra, and Misha T. Kwasniewski. 2021. Early Detection
of Plant Viral Disease Using Hyperspectral Imaging and Deep Learning. Sensors 21, 3 (2021). https://doi.org/10.3390/s21030742

25
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

[194] Xueping Ni, Changying Li, Huanyu Jiang, and Fumiomi Takeda. 2020. Deep learning image segmentation and extraction of blueberry fruit traits
associated with harvestability and yield. Horticulture research 7 (2020).
[195] Sai Vidyaranya Nuthalapati and Anirudh Tunga. 2021. Multi-domain few-shot learning and dataset for agricultural applications. In Proceedings
of the IEEE/CVF International Conference on Computer Vision. 1399–1408.
[196] Emmanuel Karlo Nyarko, Ivan Vidović, Kristijan Radočaj, and Robert Cupec. 2018. A nearest neighbor approach for fruit recognition in RGB-D
images based on detection of convex surfaces. Expert Systems with Applications 114 (2018), 454–466.
[197] Ahmad Ostovar, Ola Ringdahl, and Thomas Hellström. 2018. Adaptive image thresholding of yellow peppers for a harvesting robot. Robotics 7, 1
(2018), 11.
[198] Nobuyuki Otsu. 1979. A threshold selection method from gray-level histograms. IEEE transactions on systems, man, and cybernetics 9, 1 (1979),
62–66.
[199] Christian Payer, Darko Štern, Thomas Neff, Horst Bischof, and Martin Urschler. 2018. Instance Segmentation and Tracking with Cosine Embed-
dings and Recurrent Hourglass Networks. In International Conference on Medical Image Computing and Computer-Assisted Intervention.
[200] Tejaswini Pedapati, Avinash Balakrishnan, Karthikeyan Shanmugam, and Amit Dhurandhar. 2020. Learning global transparent models consistent
with local contrastive explanations. Advances in neural information processing systems 33 (2020), 3592–3602.
[201] Gabriel Pereyra, George Tucker, Jan Chorowski, Łukasz Kaiser, and Geoffrey Hinton. 2017. Regularizing Neural Networks by Penalizing Confident
Output Distributions. arXiv 1701.06548 (2017).
[202] Wilhelm Pfeffer. 1900. The physiology of plants: a treatise upon the metabolism and sources of energy in plants. Vol. 1. Clarendon Press.
[203] Pedro O Pinheiro and Ronan Collobert. 2015. Doll´ ar P. Learning to segment object candidates. In Proc. the 28th Int. Conf. Neural Information
Processing Systems. 1990–1998.
[204] Pedro O. Pinheiro, Tsung-Yi Lin, Ronan Collobert, and Piotr Dollár. 2016. Learning to Refine Object Segments. In Computer Vision – ECCV 2016,
Bastian Leibe, Jiri Matas, Nicu Sebe, and Max Welling (Eds.). Springer International Publishing, Cham, 75–91.
[205] Przemyslaw Prusinkiewicz. 2002. Art and science of life: designing and growing virtual plants with L-systems. In XXVI International Horticultural
Congress: Nursery Crops; Development, Evaluation, Production and Use 630. 15–28.
[206] Garima Pruthi, Frederick Liu, Satyen Kale, and Mukund Sundararajan. 2020. Estimating training data influence by tracing gradient descent.
Advances in Neural Information Processing Systems 33 (2020), 19920–19930.
[207] Zhongang Qi, Saeed Khorram, and Li Fuxin. 2021. Embedding deep networks into visual explanations. Artificial Intelligence 292 (2021), 103435.
[208] K Ragazou, A Garefalakis, E Zafeiriou, and I Passas. 2022. Agriculture 5.0: A New Strategic Management Mode for a Cut Cost and an Energy
Efficient Agriculture Sector. Energies 2022, 15, 3113.
[209] Parastoo Rahimi, Md Saiful Islam, Phelipe Magalhães Duarte, Sina Salajegheh Tazerji, Md Abdus Sobur, Mohamed E El Zowalaty, Hossam M
Ashour, and Md Tanvir Rahman. 2021. Impact of the COVID-19 pandemic on food production and animal health. Trends in Food Science &
Technology (2021).
[210] Maryam Rahnemoonfar and Clay Sheppard. 2017. Deep count: fruit counting based on deep simulated learning. Sensors 17, 4 (2017), 905.
[211] Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. 2016. You Only Look Once: Unified, Real-Time Object Detection. arXiv 1506.02640
(2016).
[212] Joseph Redmon and Ali Farhadi. 2017. YOLO9000: Better, Faster, Stronger. In Proceedings of the IEEE Conference on Computer Vision and Pattern
Recognition (CVPR).
[213] Mengye Ren and Richard S Zemel. 2017. End-to-end instance segmentation with recurrent attention. In Proceedings of the IEEE conference on
computer vision and pattern recognition. 6656–6664.
[214] Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. 2015. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks.
In Advances in Neural Information Processing Systems.
[215] Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. 2016. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks.
arXiv 1506.01497 (2016).
[216] A. Reyes-Yanes, P. Martinez, and R. Ahmad. 2020. Real-time growth rate and fresh weight estimation for little gem romaine lettuce in aquaponic
grow beds. Computers and Electronics in Agriculture 179 (2020), 105827. https://doi.org/10.1016/j.compag.2020.105827
[217] A Reyes-Yanes, Pablo Martinez, and R Ahmad. 2020. Real-time growth rate and fresh weight estimation for little gem romaine lettuce in aquaponic
grow beds. Computers and Electronics in Agriculture 179 (2020), 105827.
[218] Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. 2016. " Why should i trust you?" Explaining the predictions of any classifier. In Proceedings
of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining. 1135–1144.
[219] Joe M Roberts, Toby JA Bruce, James M Monaghan, Tom W Pope, Simon R Leather, and Andrew M Beacham. 2020. Vertical farming systems bring
new considerations for pest and disease management. Annals of Applied Biology 176, 3 (2020), 226–232.
[220] David Rolnick, Priya L. Donti, Lynn H. Kaack, Kelly Kochanski, Alexandre Lacoste, Kris Sankaran, Andrew Slavin Ross, Nikola Milojevic-Dupont,
Natasha Jaques, Anna Waldman-Brown, Alexandra Sasha Luccioni, Tegan Maharaj, Evan D. Sherwin, S. Karthik Mukkavilli, Konrad P. Kording,
Carla P. Gomes, Andrew Y. Ng, Demis Hassabis, John C. Platt, Felix Creutzig, Jennifer Chayes, and Yoshua Bengio. 2022. Tackling Climate Change
with Machine Learning. ACM Comput. Surv. 55, 2, Article 42 (feb 2022), 96 pages. https://doi.org/10.1145/3485128
[221] Eduardo Romera, José M Alvarez, Luis M Bergasa, and Roberto Arroyo. 2017. Erfnet: Efficient residual factorized convnet for real-time semantic
segmentation. IEEE Transactions on Intelligent Transportation Systems 19, 1 (2017), 263–272.
26
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

[222] Bernardino Romera-Paredes and Philip Hilaire Sean Torr. 2016. Recurrent instance segmentation. In European conference on computer vision.
Springer, 312–329.
[223] Olaf Ronneberger, Philipp Fischer, and Thomas Brox. 2015. U-Net: Convolutional Networks for Biomedical Image Segmentation. In Medical Image
Computing and Computer-Assisted Intervention – MICCAI 2015, Nassir Navab, Joachim Hornegger, William M. Wells, and Alejandro F. Frangi (Eds.).
Springer International Publishing, Cham, 234–241.
[224] Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael
Bernstein, Alexander C. Berg, and Li Fei-Fei. 2015. ImageNet Large Scale Visual Recognition Challenge. International Journal of Computer Vision
(IJCV) 115, 3 (2015), 211–252. https://doi.org/10.1007/s11263-015-0816-y
[225] Radu Bogdan Rusu. 2010. Semantic 3D object maps for everyday manipulation in human living environments. KI-Künstliche Intelligenz 24, 4
(2010), 345–348.
[226] Inkyu Sa, Zongyuan Ge, Feras Dayoub, Ben Upcroft, Tristan Perez, and Chris McCool. 2016. Deepfruits: A fruit detection system using deep
neural networks. sensors 16, 8 (2016), 1222.
[227] Inkyu Sa, Zongyuan Ge, Feras Dayoub, Ben Upcroft, Tristan Perez, and Chris McCool. 2016. DeepFruits: A Fruit Detection System Using Deep
Neural Networks. Sensors 16, 8 (2016). https://doi.org/10.3390/s16081222
[228] Verónica Saiz-Rubio and Francisco Rovira-Más. 2020. From smart farming towards agriculture 5.0: A review on crop data management. Agronomy
10, 2 (2020), 207.
[229] Amaia Salvador, Miriam Bellver, Victor Campos, Manel Baradad, Ferran Marques, Jordi Torres, and Xavier Giro-i Nieto. 2017. Recurrent Neural
Networks for Semantic Instance Segmentation. arXiv Preprint 1712.00617 (2017). https://arxiv.org/abs/1712.00617
[230] Amaia Salvador, Miriam Bellver, Victor Campos, Manel Baradad, Ferran Marques, Jordi Torres, and Xavier Giro-i Nieto. 2017. Recurrent neural
networks for semantic instance segmentation. arXiv preprint arXiv:1712.00617 (2017).
[231] Hanno Scharr, Massimo Minervini, Andrew P French, Christian Klukas, David M Kramer, Xiaoming Liu, Imanol Luengo, Jean-Michel Pape, Gerrit
Polder, Danijela Vukadinovic, et al. 2016. Leaf segmentation in plant phenotyping: a collation study. Machine vision and applications 27, 4 (2016),
585–606.
[232] Michael Gomez Selvaraj, Alejandro Vergara, Henry Ruiz, Nancy Safari, Sivalingam Elayabalan, Walter Ocimati, and Guy Blomme. 2019. AI-
powered banana diseases and pest detection. Plant Methods 15, 1 (2019), 1–11.
[233] Ramprasaath R Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. 2017. Grad-cam: Visual
explanations from deep networks via gradient-based localization. In Proceedings of the IEEE international conference on computer vision. 618–626.
[234] Woo Chaw Seng and Seyed Hadi Mirisaee. 2009. A new method for fruits recognition system. In 2009 international conference on electrical
engineering and informatics, Vol. 1. IEEE, 130–134.
[235] Jayavelu Senthilnath, Akanksha Dokania, Manasa Kandukuri, KN Ramesh, Gautham Anand, and SN Omkar. 2016. Detection of tomatoes using
spectral-spatial methods in remotely sensed RGB images captured by UAV. Biosystems engineering 146 (2016), 16–32.
[236] Pierre Sermanet, David Eigen, Xiang Zhang, Michaël Mathieu, Rob Fergus, and Yann LeCun. 2013. Overfeat: Integrated recognition, localization
and detection using convolutional networks. arXiv preprint arXiv:1312.6229 (2013).
[237] Lei Sha, Oana-Maria Camburu, and Thomas Lukasiewicz. 2021. Learning from the Best: Rationalizing Predictions by Adversarial Information
Calibration.. In AAAI. 13771–13779.
[238] Manish Sharma, Mayur Dhanaraj, Srivallabha Karnam, Dimitris G Chachlakis, Raymond Ptucha, Panos P Markopoulos, and Eli Saber. 2020.
YOLOrs: Object detection in multimodal remote sensing imagery. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
14 (2020), 1497–1508.
[239] Li Shen, Zhouchen Lin, and Qingming Huang. 2016. Relay backpropagation for effective learning of deep convolutional neural networks. In ECCV.
Springer, 467–482.
[240] Rui Shi, Tianxing Li, and Yasushi Yamaguchi. 2020. An attribution-based pruning method for real-time mango detection with YOLO network.
Computers and electronics in agriculture 169 (2020), 105214.
[241] Ruifeng Shi, Deming Zhai, Xianming Liu, Junjun Jiang, and Wen Gao. 2020. Rectified meta-learning from noisy labels for robust image-based
plant disease diagnosis. arXiv preprint arXiv:2003.07603 (2020).
[242] Shigeharu Shimamura. [n. d.]. Indoor Cultivation for the Future. https://frc.ri.cmu.edu/~ssingh/VF/Challenges_in_Vertical_Farming/Schedule_files/SHIMAMURA.pdf .
[243] Vivswan Shitole, Fuxin Li, Minsuk Kahng, Prasad Tadepalli, and Alan Fern. 2021. One explanation is not enough: structured attention graphs for
image classification. Advances in Neural Information Processing Systems 34 (2021), 11352–11363.
[244] Avanti Shrikumar, Peyton Greenside, Anna Shcherbina, and Anshul Kundaje. 2016. Not just a black box: Interpretable deep learning by propagating
activation differences. arXiv preprint arXiv:1605.01713 4 (2016).
[245] Aleksandra Sidor and Piotr Rzymski. 2020. Dietary choices and habits during COVID-19 lockdown: experience from Poland. Nutrients 12, 6 (2020),
1657.
[246] Claudênia Ferreira da Silva, Carlos Hidemi Uesugi, Luiz Eduardo Bassay Blum, Abi Soares dos Anjos Marques, and Marisa Álvares da Silva Velloso
Ferreira. 2016. Molecular detection of Erwinia psidii in guava plants under greenhouse and field conditions. Ciência Rural 46 (2016), 1528–1534.
[247] David Silver, Julian Schrittwieser, Karen Simonyan, Ioannis Antonoglou, Aja Huang, Arthur Guez, Thomas Hubert, Lucas Baker, Matthew Lai,
Adrian Bolton, Yutian Chen, Timothy Lillicrap, Fan Hui, Laurent Sifre, George van den Driessche, Thore Graepel, and Demis Hassabis. 2017.
Mastering the game of Go without human knowledge. Nature 550 (2017), 354–359.
27
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

[248] Karen Simonyan and Andrew Zisserman. 2015. Very Deep Convolutional Networks for Large-Scale Image Recognition. In International Conference
on Learning Representations.
[249] Uday Pratap Singh, Siddharth Singh Chouhan, Sukirty Jain, and Sanjeev Jain. 2019. Multilayer convolution neural network for the classification
of mango leaves infected by anthracnose disease. IEEE Access 7 (2019), 43721–43729.
[250] Daniel Smilkov, Nikhil Thorat, Been Kim, Fernanda Viégas, and Martin Wattenberg. 2017. Smoothgrad: removing noise by adding noise. arXiv
preprint arXiv:1706.03825 (2017).
[251] Samuel L Smith, Benoit Dherin, David GT Barrett, and Soham De. 2021. On the origin of implicit regularization in stochastic gradient descent.
arXiv preprint arXiv:2101.12176 (2021).
[252] Jake Snell, Kevin Swersky, and Richard Zemel. 2017. Prototypical networks for few-shot learning. Advances in neural information processing
systems 30 (2017).
[253] Jie Song, Chengchao Shen, Jie Lei, An-Xiang Zeng, Kairi Ou, Dacheng Tao, and Mingli Song. 2018. Selective zero-shot classification with augmented
attributes. In Proceedings of the European Conference on Computer Vision (ECCV). 468–483.
[254] Jie Song, Chengchao Shen, Yezhou Yang, Yang Liu, and Mingli Song. 2018. Transductive unbiased embedding for zero-shot learning. In Proceedings
of the IEEE conference on computer vision and pattern recognition. 1024–1033.
[255] Edgar P Spalding and Nathan D Miller. 2013. Image analysis is driving a renaissance in growth measurement. Current opinion in plant biology 16,
1 (2013), 100–104.
[256] Andreas Peter Steiner, Alexander Kolesnikov, Xiaohua Zhai, Ross Wightman, Jakob Uszkoreit, and Lucas Beyer. 2022. How to
train your ViT? Data, Augmentation, and Regularization in Vision Transformers. Transactions on Machine Learning Research (2022).
https://openreview.net/forum?id=4nPswr1KcP
[257] Richard N Strange and Peter R Scott. 2005. Plant disease: a threat to global food security. Annual review of phytopathology 43, 1 (2005), 83–116.
[258] Chen Sun, Abhinav Shrivastava, Saurabh Singh, and Abhinav Gupta. 2017. Revisiting Unreasonable Effectiveness of Data in Deep Learning Era.
In Proceedings of the IEEE International Conference on Computer Vision (ICCV).
[259] Jun Sun, Xiaofei He, Xiao Ge, Xiaohong Wu, Jifeng Shen, and Yingying Song. 2018. Detection of key organs in tomato based on deep migration
learning in a complex background. Agriculture 8, 12 (2018), 196.
[260] Kaiqiong Sun, Xuan Wang, Shoushuai Liu, and ChangHua Liu. 2021. Apple, peach, and pear flower detection using semantic segmentation
network and shape constraint level set. Computers and Electronics in Agriculture 185 (2021), 106150.
[261] Mukund Sundararajan, Ankur Taly, and Qiqi Yan. 2017. Axiomatic attribution for deep networks. In International conference on machine learning.
PMLR, 3319–3328.
[262] Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, and Zbigniew Wojna. 2015. Rethinking the Inception Architecture for
Computer Vision. arXiv Preprint 1512.00567 (2015).
[263] Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. 2016. Rethinking the Inception Architecture for Computer
Vision. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2818–2826. https://doi.org/10.1109/CVPR.2016.308
[264] Christian Szegedy, Wojciech Zaremba, Ilya Sutskever, Joan Bruna, Dumitru Erhan, Ian Goodfellow, and Rob Fergus. 2013. Intriguing properties
of neural networks. arXiv preprint arXiv:1312.6199 (2013).
[265] Niko Sünderhauf, Oliver Brock, Walter Scheirer, Raia Hadsell, Dieter Fox, Jürgen Leitner, Ben Upcroft, Pieter Abbeel, Wolfram Burgard, Michael
Milford, and Peter Corke. 2018. The limits and potentials of deep learning for robotics. The International Journal of Robotics Research 37, 4-5 (2018),
405–420.
[266] Mingxing Tan and Quoc Le. 2019. EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks. In Proceedings of the 36th Interna-
tional Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 97), Kamalika Chaudhuri and Ruslan Salakhutdinov (Eds.).
PMLR, 6105–6114. https://proceedings.mlr.press/v97/tan19a.html
[267] Mingxing Tan and Quoc V. Le. 2019. EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks. In ICML.
[268] Wenzhi Tang, Tingting Yan, Fei Wang, Jingxian Yang, Jian Wu, Jianlong Wang, Tianli Yue, and Zhonghong Li. 2019. Rapid fabrication of wearable
carbon nanotube/graphite strain sensor for real-time monitoring of plant growth. Carbon 147 (2019), 295–302.
[269] Mercè Teixidó, Davinia Font, Tomàs Pallejà, Marcel Tresanchez, Miquel Nogués, and Jordi Palacín. 2012. Definition of linear color models in the
RGB vector color space to detect red peaches in orchard images taken under natural illumination. Sensors 12, 6 (2012), 7701–7718.
[270] Sunil Thulasidasan, Gopinath Chennupati, Jeff A Bilmes, Tanmoy Bhattacharya, and Sarah Michalak. 2019. On Mixup Train-
ing: Improved Calibration and Predictive Uncertainty for Deep Neural Networks. In Advances in Neural Information Processing Sys-
tems, H. Wallach, H. Larochelle, A. Beygelzimer, F. d'Alché-Buc, E. Fox, and R. Garnett (Eds.), Vol. 32. Curran Associates, Inc.
https://proceedings.neurips.cc/paper/2019/file/36ad8b5f42db492827016448975cc22d-Paper.pdf
[271] Mengxiao Tian, Hao Guo, Hong Chen, Qing Wang, Chengjiang Long, and Yuhao Ma. 2019. Automated pig counting using deep learning. Computers
and Electronics in Agriculture 163 (2019), 104840. https://doi.org/10.1016/j.compag.2019.05.049
[272] Zhi Tian, Chunhua Shen, Hao Chen, and Tong He. 2019. FCOS: Fully Convolutional One-Stage Object Detection. In Proceedings of the IEEE/CVF
International Conference on Computer Vision (ICCV).
[273] Hugo Touvron, Matthieu Cord, Alexandre Sablayrolles, Gabriel Synnaeve, and Hervé Jégou. 2021. Going Deeper With Image Transformers. In
Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). 32–42.

28
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

[274] Magdalena Trojak, Ernest Skowron, Tomasz Sobala, Maciej Kocurek, and Jan Pałyga. 2022. Effects of partial replacement of red by green light in
the growth spectrum on photomorphogenesis and photosynthesis in tomato plants. Photosynthesis research 151, 3 (2022), 295–312.
[275] Jordan Ubbens, Mikolaj Cieslak, Przemyslaw Prusinkiewicz, and Ian Stavness. 2018. The use of plant models in deep learning: an application to
leaf counting in rosette plants. Plant methods 14, 1 (2018), 1–10.
[276] Nathalie van Wijkvliet. [n. d.]. No space, no problem. How Singapore is turning into an edible paradise.
https://sustainableurbandelta.com/singapore-30-by-30-food-system/. Accessed: 2022-8-15.
[277] Dong Wang and Xiaoyang Tan. 2017. Robust distance metric learning via Bayesian inference. IEEE Transactions on Image Processing 27, 3 (2017),
1542–1553.
[278] Dongyi Wang, Robert Vinson, Maxwell Holmes, Gary Seibel, Avital Bechar, Shimon Nof, and Yang Tao. 2019. Early detection of tomato spotted
wilt virus by hyperspectral imaging and outlier removal auxiliary classifier generative adversarial nets (OR-AC-GAN). Scientific reports 9, 1 (2019),
1–14.
[279] Deng-Bao Wang, Lei Feng, and Min-Ling Zhang. 2021. Rethinking Calibration of Deep Neural Networks: Do Not Be Afraid of Overconfidence. In
Advances in Neural Information Processing Systems, M. Ranzato, A. Beygelzimer, Y. Dauphin, P.S. Liang, and J. Wortman Vaughan (Eds.), Vol. 34.
Curran Associates, Inc., 11809–11820. https://proceedings.neurips.cc/paper/2021/file/61f3a6dbc9120ea78ef75544826c814e-Paper.pdf
[280] Hao Wang, Yitong Wang, Zheng Zhou, Xing Ji, Dihong Gong, Jingchao Zhou, Zhifeng Li, and Wei Liu. 2018. Cosface: Large margin cosine loss
for deep face recognition. In CVPR. 5265–5274.
[281] Peng Wang, An Yang, Rui Men, Junyang Lin, Shuai Bai, Zhikang Li, Jianxin Ma, Chang Zhou, Jingren Zhou, and Hongxia Yang. 2022. Unifying
architectures, tasks, and modalities through a simple sequence-to-sequence learning framework. arXiv preprint arXiv:2202.03052 (2022).
[282] Qilong Wang, Banggu Wu, Pengfei Zhu, Peihua Li, Wangmeng Zuo, and Qinghua Hu. 2020. ECA-Net: Efficient Channel Attention
for Deep Convolutional Neural Networks. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 11531–11539.
https://doi.org/10.1109/CVPR42600.2020.01155
[283] Xinlong Wang, Tao Kong, Chunhua Shen, Yuning Jiang, and Lei Li. 2020. SOLO: Segmenting Objects by Locations. In Computer Vision – ECCV
2020, Andrea Vedaldi, Horst Bischof, Thomas Brox, and Jan-Michael Frahm (Eds.). Springer International Publishing, Cham, 649–665.
[284] Xinlong Wang, Rufeng Zhang, Tao Kong, Lei Li, and Chunhua Shen. 2020. SOLOv2: Dynamic and Fast Instance Segmentation. In Advances
in Neural Information Processing Systems, H. Larochelle, M. Ranzato, R. Hadsell, M.F. Balcan, and H. Lin (Eds.), Vol. 33. Curran Associates, Inc.,
17721–17732. https://proceedings.neurips.cc/paper/2020/file/cd3afef9b8b89558cd56638c3631868a-Paper.pdf
[285] Yulong Wang, Hang Su, Bo Zhang, and Xiaolin Hu. 2018. Interpret neural networks by identifying critical data routing paths. In proceedings of
the IEEE conference on computer vision and pattern recognition. 8906–8914.
[286] Zhenglin Wang, Kerry Walsh, and Anand Koirala. 2019. Mango fruit load estimation using a video based MangoYOLO—Kalman filter—hungarian
algorithm method. Sensors 19, 12 (2019), 2742.
[287] Hongxin Wei, Lei Feng, Xiangyu Chen, and Bo An. 2020. Combating noisy labels by agreement: A joint training method with co-regularization.
In CVPR. 13726–13735.
[288] Xiangqin Wei, Kun Jia, Jinhui Lan, Yuwei Li, Yiliang Zeng, and Chunmei Wang. 2014. Automatic method of fruit object extraction under complex
agricultural background for vision system of fruit picking robot. Optik 125, 19 (2014), 5684–5689.
[289] Yeming Wen, Dustin Tran, and Jimmy Ba. 2020. BatchEnsemble: An Alternative Approach to Efficient Ensemble and Lifelong Learning. In ICLR.
[290] Jan Weyler, Federico Magistri, Peter Seitz, Jens Behley, and Cyrill Stachniss. 2022. In-Field Phenotyping Based on Crop Leaf and Plant Instance
Segmentation. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 2725–2734.
[291] Steven Euijong Whang, Yuji Roh, Hwanjun Song, and Jae-Gil Lee. 2021. Data Collection and Quality Challenges in Deep Learning: A Data-Centric
AI Perspective. (2021). https://doi.org/10.48550/ARXIV.2112.06409
[292] Adrian Wolny, Qin Yu, Constantin Pape, and Anna Kreshuk. 2022. Sparse Object-Level Supervision for Instance Segmentation With Pixel Embed-
dings. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 4402–4411.
[293] Adrian Wolny, Qin Yu, Constantin Pape, and Anna Kreshuk. 2022. Sparse object-level supervision for instance segmentation with pixel embed-
dings. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 4402–4411.
[294] Arissa Wongpanich, Hieu Pham, James Demmel, Mingxing Tan, Quoc Le, Yang You, and Sameer Kumar. 2021. Training EfficientNets at Super-
computer Scale: 83% ImageNet Top-1 Accuracy in One Hour. In 2021 IEEE International Parallel and Distributed Processing Symposium Workshops
(IPDPSW). 947–950.
[295] Jingui Wu, Baohua Zhang, Jun Zhou, Yingjun Xiong, Baoxing Gu, and Xiaolong Yang. 2019. Automatic recognition of ripening tomatoes by
combining multi-feature fusion with a bi-layer classification strategy for harvesting robots. Sensors 19, 3 (2019), 612.
[296] Yuli Wu, Long Chen, and Dorit Merhof. 2020. Improving Pixel Embedding Learning through Intermediate Distance Regression Supervision for
Instance Segmentation. In European Conference on Computer Vision Workshop. Springer, 213–227.
[297] Yan Wu, Jeff Donahue, David Balduzzi, Karen Simonyan, and Timothy Lillicrap. 2019. Logan: Latent optimisation for generative adversarial
networks. arXiv preprint arXiv:1912.00953 (2019).
[298] Adnelba Vitória Oliveira Xavier, Geovani Soares de Lima, Hans Raj Gheyi, André Alisson Rodrigues da Silva, Lauriane Almeida dos Anjos Soares,
and Cassiano Nogueira de Lacerda. 2022. Gas exchange, growth and quality of guava seedlings under salt stress and salicylic acid. Revista Ambiente
& Água 17 (2022).

29
Woodstock ’18, June 03–05, 2018, Woodstock, NY Luo, et al.

[299] Yongqin Xian, Zeynep Akata, Gaurav Sharma, Quynh Nguyen, Matthias Hein, and Bernt Schiele. 2016. Latent embeddings for zero-shot classifi-
cation. In Proceedings of the IEEE conference on computer vision and pattern recognition. 69–77.
[300] Jingjing Xie, Bing Xu, and Zhang Chuang. 2013. Horizontal and Vertical Ensemble with Deep Representation for Classification. arXiv 1306.2759
(2013).
[301] Saining Xie, Ross B. Girshick, Piotr Dollár, Zhuowen Tu, and Kaiming He. 2016. Aggregated Residual Transformations for Deep Neural Networks.
arXiv Preprint 1611.05431 (2016).
[302] Haotian Yan, Zhe Li, Weijian Li, Changhu Wang, Ming Wu, and Chuang Zhang. 2021. ConTNet: Why not use convolution and transformer at the
same time? arXiv Preprint 2104.13497 (2021).
[303] Biyun Yang and Yong Xu. 2021. Applications of deep-learning approaches in horticultural research: a review. Horticulture Research 8 (2021).
[304] Zhengyuan Yang, Zhe Gan, Jianfeng Wang, Xiaowei Hu, Yumao Lu, Zicheng Liu, and Lijuan Wang. 2022. An empirical study of gpt-3 for few-shot
knowledge-based vqa. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 36. 3081–3089.
[305] Chih-Kuan Yeh, Cheng-Yu Hsieh, Arun Suggala, David I Inouye, and Pradeep K Ravikumar. 2019. On the (in) fidelity and sensitivity of explanations.
Advances in Neural Information Processing Systems 32 (2019).
[306] Chih-Kuan Yeh, Joon Kim, Ian En-Hsu Yen, and Pradeep K Ravikumar. 2018. Representer point selection for explaining deep neural networks.
Advances in neural information processing systems 31 (2018).
[307] T Yeshitela, PJ Robbertse, and PJC Stassen. 2005. Effects of pruning on flowering, yield and fruit quality in mango (Mangifera indica). Australian
Journal of Experimental Agriculture 45, 10 (2005), 1325–1330.
[308] Michael Yeung, Leonardo Rundo, Yang Nan, Evis Sala, Carola-Bibiane Schönlieb, and Guang Yang. 2021. Calibrating the Dice loss to handle neural
network overconfidence for biomedical image segmentation. arXiv preprint arXiv:2111.00528 (2021).
[309] Hui Ying, Zhaojin Huang, Shu Liu, Tianjia Shao, and Kun Zhou. 2021. EmbedMask: Embedding Coupling for Instance Segmentation.. In IJCAI.
1266–1273.
[310] Mo Yu, Yang Zhang, Shiyu Chang, and Tommi Jaakkola. 2021. Understanding interlocking dynamics of cooperative rationalization. Advances in
Neural Information Processing Systems 34 (2021), 12822–12835.
[311] Yang Yu, Kailiang Zhang, Hui Liu, Li Yang, and Dongxing Zhang. 2020. Real-Time Visual Localization of the Picking Points for a Ridge-Planting
Strawberry Harvesting Robot. IEEE Access 8 (2020), 116556–116568. https://doi.org/10.1109/ACCESS.2020.3003034
[312] Yang Yu, Kailiang Zhang, Li Yang, and Dongxing Zhang. 2019. Fruit detection for strawberry harvesting robot in non-structural environment
based on Mask-RCNN. Computers and Electronics in Agriculture 163 (2019), 104846.
[313] Zhiwen Yu, Hau-San Wong, and Guihua Wen. 2011. A modified support vector machine and its application to image segmentation. Image and
Vision Computing 29, 1 (2011), 29–40.
[314] Hongbo Yuan, Jiajun Zhu, Qifan Wang, Man Cheng, and Zhenjiang Cai. 2022. An Improved DeepLab v3+ Deep Learning Network Applied to the
Segmentation of Grape Leaf Black Rot Spots. Frontiers in Plant Science 13 (2022).
[315] Kun Yuan, Shaopeng Guo, Ziwei Liu, Aojun Zhou, Fengwei Yu, and Wei Wu. 2021. Incorporating Convolution Designs Into Visual Transformers.
In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). 579–588.
[316] Ting Yuan, Lin Lv, Fan Zhang, Jun Fu, Jin Gao, Junxiong Zhang, Wei Li, Chunlong Zhang, and Wenqiang Zhang. 2020. Robust cherry tomatoes
detection algorithm in greenhouse scene based on SSD. Agriculture 10, 5 (2020), 160.
[317] Ilaria Zambon, Massimo Cecchini, Gianluca Egidi, Maria Grazia Saporito, and Andrea Colantoni. 2019. Revolution 4.0: Industry vs. agriculture in
a future development for SMEs. Processes 7, 1 (2019), 36.
[318] Chiyuan Zhang, Samy Bengio, Moritz Hardt, Benjamin Recht, and Oriol Vinyals. 2017. Understanding deep learning requires rethinking general-
ization. arXiv 1611.03530 (2017).
[319] Kaipeng Zhang, Zhanpeng Zhang, Zhifeng Li, and Yu Qiao. 2016. Joint face detection and alignment using multitask cascaded convolutional
networks. IEEE signal processing letters 23, 10 (2016), 1499–1503.
[320] Li Zhang, Guan Gui, Abdul Mateen Khattak, Minjuan Wang, Wanlin Gao, and Jingdun Jia. 2019. Multi-task cascaded convolutional networks
based intelligent fruit detection for designing automated robot. IEEE Access 7 (2019), 56028–56038.
[321] Li Zhang, Jingdun Jia, Guan Gui, Xia Hao, Wanlin Gao, and Minjuan Wang. 2018. Deep Learning Based Improved Classification System for
Designing Tomato Harvesting Robot. IEEE Access 6 (2018), 67940–67950. https://doi.org/10.1109/ACCESS.2018.2879324
[322] Li Zhang, Jingdun Jia, Guan Gui, Xia Hao, Wanlin Gao, and Minjuan Wang. 2018. Deep learning based improved classification system for designing
tomato harvesting robot. IEEE Access 6 (2018), 67940–67950.
[323] Lingxian Zhang, Zanyu Xu, Dan Xu, Juncheng Ma, Yingyi Chen, and Zetian Fu. 2020. Growth monitoring of greenhouse lettuce based on a
convolutional neural network. Horticulture research 7 (2020).
[324] Shifeng Zhang, Cheng Chi, Yongqiang Yao, Zhen Lei, and Stan Z. Li. 2020. Bridging the Gap Between Anchor-Based and Anchor-Free Detection
via Adaptive Training Sample Selection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).
[325] Shanwen Zhang, Subing Zhang, Chuanlei Zhang, Xianfeng Wang, and Yun Shi. 2019. Cucumber leaf disease identification with global pooling
dilated convolutional neural network. Computers and Electronics in Agriculture 162 (2019), 422–430.
[326] Wendong Zhang. 2021. The Case for Healthy US-China Agricultural Trade Relations despite Deglobalization Pressures. Applied Economic Per-
spectives and Policy 43, 1 (2021), 225–247.
[327] Wenwei Zhang, Jiangmiao Pang, Kai Chen, and Chen Change Loy. 2021. K-Net: Towards Unified Image Segmentation. In NeurIPS.
30
A Survey of Computer Vision Technologies In Urban and Controlled-environment Agriculture Woodstock ’18, June 03–05, 2018, Woodstock, NY

[328] Xiao Zhang and Ximing Cai. 2011. Climate change impacts on global agricultural land availability. Environmental Research Letters 6, 1 (2011),
014014.
[329] Xiangyu Zhang, Xinyu Zhou, Mengxiao Lin, and Jian Sun. 2017. ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile
Devices. arXiv Preprint 1707.01083 (2017).
[330] Yu Zhang, Peter Tiňo, Aleš Leonardis, and Ke Tang. 2021. A survey on neural network interpretability. IEEE Transactions on Emerging Topics in
Computational Intelligence (2021).
[331] Qijie Zhao, Tao Sheng, Yongtao Wang, Zhi Tang, Ying Chen, Ling Cai, and Haibin Ling. 2018. M2Det: A Single-Shot Object Detector based on
Multi-Level Feature Pyramid Network. arXiv Preprint 1811.04533 (2018).
[332] Guoqing Zheng, Ahmed Hassan Awadallah, and Susan Dumais. 2021. Meta Label Correction for Noisy Label Learning. In AAAI, Vol. 35.
[333] Yaoyao Zhong, Weihong Deng, Mei Wang, Jiani Hu, Jianteng Peng, Xunqiang Tao, and Yaohai Huang. 2019. Unequal-training for deep face
recognition with long-tailed noisy data. In CVPR. 7812–7821.
[334] Boyan Zhou, Quan Cui, Xiu-Shen Wei, and Zhao-Min Chen. 2020. BBN: Bilateral-Branch Network with Cumulative Learning for Long-Tailed
Visual Recognition. (2020), 1–8.
[335] Daquan Zhou, Bingyi Kang, Xiaojie Jin, Linjie Yang, Xiaochen Lian, Qibin Hou, and Jiashi Feng. 2021. DeepViT: Towards Deeper Vision Trans-
former. arXiv preprint arXiv:2103.11886 (2021).
[336] Xingyi Zhou, Jiacheng Zhuo, and Philipp Krähenbühl. 2019. Bottom-Up Object Detection by Grouping Extreme and Center Points. In 2019
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).
[337] Xizhou Zhu, Jinguo Zhu, Hao Li, Xiaoshi Wu, Hongsheng Li, Xiaohua Wang, and Jifeng Dai. 2022. Uni-perceiver: Pre-training unified architecture
for generic perception for zero-shot and few-shot tasks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.
16804–16815.

31