Plant Disease Detection System

This project is a comprehensive Deep Learning solution designed to detect and classify plant diseases from leaf images. It consists of a training pipeline using TensorFlow/Keras and a deployment API for serving predictions.

📂 Project Overview

The project is divided into two main components:

1. Model Training (Train_plant_disease_model.ipynb): A Jupyter Notebook that handles data ingestion, augmentation, and training of a Convolutional Neural Network (CNN).
2. Inference API (plant-api Repo): A backend service that serves the trained model to make real-time predictions.

🧠 Part 1: Model Training

The model is trained using the New Plant Diseases Dataset.

Dataset Details

• Source: Kaggle - New Plant Diseases Dataset
• Classes: The model is trained to recognize 38 unique classes, including healthy and diseased states of Apple, Corn, Grape, Potato, Tomato, and more.
• Image Size: Images are resized to 128x128 pixels.

Training Pipeline

1. Data Loading & Preprocessing:
   • Dataset is downloaded using opendatasets.
   • Images are loaded into training and validation splits with a batch size of 32.
2. Data Augmentation:
   • To improve generalization, the following augmentations are applied: RandomFlip, RandomRotation (0.1), and RandomContrast (0.1).
3. Model Architecture:
   • A custom CNN architecture is used, consisting of 5 blocks of Conv2D, BatchNormalization, ReLU, and MaxPooling2D layers.
   • The head of the network includes a Flatten layer, a Dense layer (1024 units), and a final Dense output layer with Softmax activation for 38-class classification.
4. Artifacts:
   • Model: Saved as plant_disease_classification.keras.
   • Class Labels: Saved as class.json.

---

🚀 Part 2: API Deployment

The deployment code is hosted in the plant-api repository.

Prerequisites

• Python 3.x
• TensorFlow
• FastAPI (or the framework defined in the repo)
• Pillow, NumPy

Setup & Installation

1. Clone the API repository:

   git clone [https://github.com/biswajr/plant-api.git](https://github.com/biswajr/plant-api.git)
   cd plant-api

2. Prepare Model Artifacts:

   • Copy the trained model plant_disease_classification.keras from the training output to the API directory.
   • Copy the class.json file to the API directory.

3. Install Dependencies:

   pip install -r requirements.txt

4. Run the Server: (Assuming FastAPI/Uvicorn)

   uvicorn main:app --reload
   or
   python -m uvicorn main:app --reload --host 0.0.0.0 --port 3000

API Usage

Endpoint: /predict Method: POST

Request: Upload an image file via multipart/form-data.

Example using cURL:

curl -X 'POST' \
  'http://localhost:8000/predict' \
  -H 'accept: application/json' \
  -H 'Content-Type: multipart/form-data' \
  -F 'file=@test_leaf.jpg;type=image/jpeg'

🤝 Contributing

Feel free to open issues or submit pull requests if you have suggestions for improving the model architecture or API performance.

📜 License

This project is open-source. Please check the specific license files in the repository for details.