Scribd
Upload
2 views16 pages

H12_Final PPT

The document presents a mid-term evaluation of a project focused on image-based plant disease detection using deep learning techniques, specifically convolutional neural networks (CNN). It outlines the motivation for the project, methods used, experimentation results, and key findings that emphasize the advantages of deep learning over traditional methods. The study highlights the development of models that improve the accuracy of plant disease diagnosis through advanced algorithms and a comprehensive dataset.

Uploaded by

ipsit9009
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
2 views16 pages

H12_Final PPT

The document presents a mid-term evaluation of a project focused on image-based plant disease detection using deep learning techniques, specifically convolutional neural networks (CNN). It outlines the motivation for the project, methods used, experimentation results, and key findings that emphasize the advantages of deep learning over traditional methods. The study highlights the development of models that improve the accuracy of plant disease diagnosis through advanced algorithms and a comprehensive dataset.

Uploaded by

ipsit9009
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

SDP MID-TERM EVALUATION

Image-BASED PLANT Disease Detection Using


Deep learning Technique

Supervised By: Dr. Prativa Das

Group No H12: Department of Computer Sc. and Engineering


Faculty of Engineering & Technology (ITER)
Rishika - 1941012021 Siksha ‘O’ Anusandhan (Deemed to be) University
Bhubaneswar, Odisha
Priyanshu Jain - 1941012142
Bidusi Manasini Mishra - 1941012082 1
Presentation Outline
▪ Introduction
▪ Overview
▪ Motivation
▪ Literature Survey
▪ Model Diagram
▪ Methods used
▪ Experimentation and Results
▪ System Specifications
▪ Datasets Description
▪ Experimental outcomes (Tables and Figures)
▪ Summary (Key Findings)
▪ Bibliography
2
Introduction
▪ Overview :
▪ Plant diseases can harm the leaf at any point during crop cultivation, causing
significant losses in crop productivity and financial value. Therefore, identifying
leaf disease is important in the farming business.
▪ The goal of this model is to emphasize the advantages that deep learning and
convolutional neural networks (CNN) provide towards detection of plant diseases.

▪ Motivations :
▪ The model's ability to more swiftly diagnose plant illnesses with conventional
approaches is the driving force behind this endeavor.

3
Literature Survey

Author Deep Learning Dataset Results


Model
Sethy et al.(2020) Multiple CNN with Rice disease 98.38%(f1-score)
[1] SVM
Zeng et al.(2020) [2] Self Attention CNN MK-D2 95.33%(Testing acc.)
AES-CD9214 98%(Testing acc.)
J Chen et al.(2020) INC-VGGN PlantVillage 97.57%(Training acc.)
[3] 91.83%(Validation acc.)
Oyewola et al.(2021) Deep residual CNN Cassava 96.75%(Testing acc.)
[4]

4
Model Diagram

Train / Performa
Splitting CNN
Dataset Validate nce
the data Model
model metrics

ResNet50, F1-score,
70-20-10 Densnet121 Accuracy

Fig 1 : Basic Overview of the plant


diseases detection model
5
Methods used
▪ Keras Framework : Keras is a high-level, deep learning API developed by Google for
implementing neural networks.

▪ Python Programming

▪ ResNet : Residual Network (ResNet) is a deep learning model used for computer vision
applications. It is a Convolutional Neural Network (CNN) architecture designed to
support hundreds or thousands of convolutional layers.

▪ DenseNet : Densely Connected Convolutional Networks (DenseNet) is a feed-forward


convolutional neural network (CNN) architecture that links each layer to every other
layer.

6
Experimentation and Results
▪ Two CNN models were proposed from scratch via training, validating and testing.
There are pre-trained techniques ( i.e: Resnet50 or DenseNet121 ) are also used to
make the model more accurate and provide better results.

▪ Model 1 consists of 2 deep layers, Model 2 consists of 3 deep layers of CNN.

MODEL 1 MODEL 2 ResNet50 DenseNet121


Training Loss 08.41 09.18 13.65 50.83
Training Accuracy 97.42 97.21 95.81 83.98
Validation Loss 38.81 48.55 14.50 46.50
Validation Accuracy 89.08 88.57 95.21 85.20
System Specifications
▪ Processor :
▪ Intel(R) Core(TM) i5-9300H CPU @ 2.40GHz 2.40 GHz

▪ RAM :
▪ 16.0 GB

▪ System Type :
▪ 64-bit operating system, x64-based processor

▪ OS Specifications :
▪ Windows 11 Home Single Language
Dataset Description
▪ This dataset is recreated using offline
augmentation from the original dataset.
This dataset consists of about 87K RGB
images of healthy and diseased crop
leaves which is categorized into 38
different classes. The total dataset is
divided into 70/20/10 ratio of training,
validation and testing set preserving the
directory structure. A new directory
containing 33 test images is created later
for prediction purpose.
▪ Dataset : PlantVillage
▪ GitHub :
https://github.com/spMohanty/PlantVillag
e-Dataset fFig 2 : Dataset Details
Experimental Outcomes
Model Epoch Loss Accuracy Val_loss Val_accuracy
1 1/5 0.8474 0.7485 0.5214 0.8444
2/5 0.2425 0.9207 0.3179 0.8968
3/5 0.1366 0.9549 0.2988 0.9052
4/5 0.0981 0.9683 0.2308 0.9328
5/5 0.0841 0.9742 0.3881 0.8908
2 1/5 1.1779 0.6596 0.6280 0.7991
2/5 0.3083 0.9008 0.4195 0.8721
3/5 0.1617 0.9489 0.6616 0.8312
4/5 0.1082 0.9666 0.6559 0.8521
5/5 0.0918 0.9721 0.4855 0.8857
Experimental Outcomes

Model Epoch Loss Accuracy Val_loss Val_accuracy


ResNet50 1/5 0.2031 0.9403 0.1851 0.9433
2/5 0.1781 0.9459 0.1695 0.9486
3/5 0.1566 0.9519 0.1572 0.9512
4/5 0.1446 0.9554 0.1438 0.9566
5/5 0.1365 0.9581 0.1450 0.9521
DenseNet121 1/5 1.6669 0.5496 0.8998 0.7159
2/5 0.7887 0.7519 0.6805 0.7845
3/5 0.6276 0.8041 0.5925 0.8156
4/5 0.5611 0.8224 0.5368 0.8257
5/5 0.5083 0.8398 0.4650 0.8520
Experimental Outcomes

GRAPH 1 : MODEL 1

GRAPH 2 : MODEL 2
Experimental Outcomes

GRAPH 3 : RenNet50 MODEL

GRAPH 4 : DenseNet121 MODEL


Summary
▪ The model will offer many methods for categorizing diseases that are used for plant
leaf disease detection, as well as an algorithm for image segmentation that may be
used for later automatic detection and classification of plant leaf diseases.

▪ In conclusion, as artificial intelligence technology advanced, the focus of research on


the detection of plant diseases and pests based on machine vision shifted from
traditional image processing and machine learning techniques to deep learning
techniques, which resolved challenging issues that could not be resolved by
conventional techniques.

14
Bibliography
● [1] P. K. Sethy, N. K. Barpanda, A. K. Rath, and S. K. Behera, “Deep feature based
Rice leaf disease identification using support vector machine”, Aug. 2020

● [2] W. Zeng and M. Li, “Crop leaf disease recognition based on self-attention
convolutional neural network”, May 2020

● [3] J. Chen, J. Chen, D. Zhang, Y. Sun, and Y. A. Nanehkaran, “Using deep transfer
learning for image-based plant disease identification”, Jun 2020.

● [4] D. O. Oyewola, E. G. Dada, S. Misra, and R. Dama, “Detecting cassava mosaic


disease using a deep residual convolutional neural network with distinct block
processing” Mar 2021
16

You might also like