Scribd
Upload
80 views

Technical Answers For Real World Problems Digital Assignment 2

This document describes using a deep learning algorithm to detect plant diseases from images. It proposes using a ResNet-152 model with data augmentation techniques like rotation, cropping and flipping on a dataset of over 54,000 plant leaf images. The pretrained ResNet-152 model is modified by replacing the final layer to classify images into 38 disease classes. The model is trained using backpropagation and optimization to accurately detect diseases when testing new leaf images. The goal is to develop a user-friendly tool to help farmers identify plant diseases faster and more accurately than existing methods.

Uploaded by

Raj Aryan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
80 views

Technical Answers For Real World Problems Digital Assignment 2

This document describes using a deep learning algorithm to detect plant diseases from images. It proposes using a ResNet-152 model with data augmentation techniques like rotation, cropping and flipping on a dataset of over 54,000 plant leaf images. The pretrained ResNet-152 model is modified by replacing the final layer to classify images into 38 disease classes. The model is trained using backpropagation and optimization to accurately detect diseases when testing new leaf images. The goal is to develop a user-friendly tool to help farmers identify plant diseases faster and more accurately than existing methods.

Uploaded by

Raj Aryan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 7

TECHNICAL ANSWERS FOR REAL WORLD

PROBLEMS
DIGITAL ASSIGNMENT 2

NAME RAJ ARYAN PATHAK


REGNO 18BCE0852
SLOT TD2
DESCRIPTION OF THE PROBLEM AND METHOD/ALGORITHM (WITH
MATHEMATICAL FOUNDATION) PROPOSED TO PROVIDE THE SOLUTION TO THE
PROBLEM.

CONTENT

S. NUMBER TITLE
1 PROBLEM STATEMENT
2 OBJECTIVE
3 ALGORITHM
Problem Statement

1. Plants play an essential role in climate change, agriculture industry and a country's
economy. Thereby taking care of plants is very crucial.
2. Identification of the plant diseases is the key to preventing the losses in the yield and
quantity of the agricultural product but it is very difficult task in agriculture field. Wrong
decision can lead to devastating loss on the production of crop and economical value of
market.
3. Farmers try to identify and diagnose the diseases and monitor the health of plants with
their own knowledge and experience. Their naked eyes can’t observe big plantation area
and it can be expensive too.
4. Leaf disease detection demands huge amount of work, knowledge in the plant diseases, and
also require the more processing time. Identification includes finding visually observable
patterns seen on the plant Existing models are used for limited crops.
5. If the scope of project is big then accuracy is less.
6. These kinds of software don’t have user friendly interface.
Objective
Using Deep Learning for Image-Based Plant Disease Detection

ALGORITHM

DATASET
Dataset contains 54,306 images of plant leaves, which 38 class labels. The dataset is split into
two parts, training and validation.

DATA AUGMENTATION
For the training, following transformations were applied to the dataset

Random Rotation
Image is rotated horizontally

Random Resized Crop


This will extract a patch of size (224, 224) from your input image randomly. So, it might pick this
path from top left, bottom right or anywhere in between.

Random Horizontal Flip - Image is flipped horizontally

Resize - Input image is resized to be of size (256, 256)

DATA NORMALIZATION

Input images has been normalize using the formula for the all the three channels RGB

image = (image - mean) / std

For the means, it's ' [0.485, 0.456, 0.406]' and for the standard deviations ''[0.229, 0.224,
0.225]' These values will shift each color channel to be centered at 0 and range from -1 to 1.

MODEL – RESNET 152

Model used is resnet-152 for training.


In this network we use a technique called skip connections . The skip
connection skips training from a few layers and connects directly to the output.
The approach behind this network is instead of layers learn the underlying
mapping, we allow network fit the residual mapping. So, instead of say H(x),
initial mapping, let the network fit,

F(x) := H(x) – x which gives  H(x) := F(x) + x.


ResNet-152 Pre-trained Model for PyTorch with a depth of up to 152 layers---8x deeper than
VGG nets but still having lower complexity.

Each ResNet block is either 2 layer deep (Used in small networks like ResNet 18, 34) or 3 layer
deep( ResNet 50, 101, 152).

ResNet-152, both have about 60M parameters so we Freeze parameters so we don't backprop through
them.
We Define a new, untrained feed-forward network as a classifier, using ReLU activations

We create classifier ordered dictionary first and since the pretranined model has 1000 classes
in the final layer and we need only 38 classes, we replace the final layer of the pretrained model
with our untrained feed-forward network

TRAINING

BACKPROPAGATION

OPTIMIZATION

TESTING

You might also like