Leaf Disease Classification Project

Project Description

This project focuses on leaf disease classification using deep learning. The model is trained on an augmented dataset of leaf images, with preprocessing steps such as dataset balancing, image transformation, and augmentation. The goal is to predict the disease category of a given leaf image.

Setup and Run Instructions

Prerequisites

To install the required dependencies, run:

pip install -r requirements.txt

1. Check Image Distribution

Run the following command to visualize the current image distribution:

python src/visualization/distribution.py -data <images folder>

This will display the distribution of images across different classes, allowing you to check if the dataset is balanced.

2. Balance the Dataset

To balance the dataset, use the following command. This will automatically augment and balance the dataset, which may take a few minutes:

python src/data/balance_dataset.py -src <images folder> -dst balanced_images/ -aug src/data/augment.py

3. Verify New Distribution

Run the distribution check again to see the updated balanced dataset:

python src/visualization/distribution.py -data balanced_images/

4. Image Transformations (Optional)

You can test image transformations (such as rotation, flipping, etc.) to understand how they work with the data. Run this command:

python src/data/transform.py <image>

This is mainly for visualization; the model doesn’t require this step.

5. Train the Model

Train the model using the augmented and balanced dataset:

python src/model/train.py balanced_images

6. Make Predictions

To test the model's predictions, use the following command:

python src/model/predict.py trained_model.zip [-image <image to predict> | -dir <folder of images to predict>]

This will show the model’s prediction for the specified image.