Tomato leaf Disease Detection with Vision Transformer (ViT)

This project aims to develop a tomato leaf disease detection system using a Vision Transformer (ViT) model. The system classifies plant diseases based on images using a custom deep learning model built on top of a pre-trained Vision Transformer.

Table of Contents

• Introduction
• Installation
• Dataset
• Data Preprocessing
• Model Architecture
• Training
• Validation
• Evaluation
• Results
• Conclusion

Introduction

Plant diseases can significantly impact agricultural productivity. Early detection and diagnosis are crucial for effective management. This project uses a Vision Transformer (ViT) model to classify plant diseases from images, leveraging state-of-the-art deep learning techniques for high accuracy.

Installation

Ensure you have the following dependencies installed:

• Python 3.7+
• PyTorch
• torchvision
• transformers
• scikit-learn
• matplotlib
• seaborn

You can install the necessary packages using pip:

pip install torch torchvision transformers scikit-learn matplotlib seaborn

Dataset

The dataset used in this project is the "New Plant Diseases Dataset (Augmented)". It contains images of various plant diseases across different categories.

Data Preprocessing

The images are preprocessed with the following steps:

• Training Data:

  • Resize to 224x224 pixels
  • Random horizontal flip
  • Random rotation (up to 10 degrees)
  • Color jitter (brightness, contrast, saturation, hue)
  • Normalize using ImageNet mean and standard deviation

• Validation Data:

  • Resize to 224x224 pixels
  • Normalize using ImageNet mean and standard deviation

Model Architecture

The model consists of a pre-trained Vision Transformer (ViT) base model with a custom fully connected (FC) layer for classification.

• Vision Transformer (ViT):

  • Configuration: google/vit-base-patch16-224
  • Output hidden size: 768

• Custom Model:

  • Fully connected layer with 10 output classes (assuming 10 disease categories)

Training

The training process involves fine-tuning the last few layers of the ViT model along with the custom fully connected layer. Key steps include:

• Loss function: CrossEntropyLoss
• Optimizer: Adam with differential learning rates for different parts of the model
• Number of epochs: 20

The model is trained and validated in each epoch, and the best model (with highest validation accuracy) is saved.

Validation

Validation is performed at the end of each epoch to evaluate the model's performance on unseen data. Key metrics include validation loss and accuracy.

Evaluation

After training, the best model is loaded and evaluated on the validation dataset. Evaluation metrics include:

• Confusion matrix
• Classification report (precision, recall, F1-score for each class)
• Overall accuracy

Results

The results of the training process are visualized using plots for:

• Training and validation loss
• Training and validation accuracy

Additionally, a confusion matrix is plotted to analyze the model's performance across different classes.

Conclusion

This project demonstrates the application of Vision Transformers for plant disease detection. The fine-tuned ViT model achieves high accuracy in classifying various plant diseases, providing a valuable tool for agricultural diagnostics.

---