🍅 Tomato Leaf Disease Classification

A deep learning system to detect and classify tomato diseases from leaf spot images using a Convolutional Neural Network (CNN) model. The model achieves 93% accuracy and is deployed via FastAPI and accessible through web and mobile apps.

---

✅ Key Features

• Classifies common tomato diseases via visible leaf spots.
• CNN trained on annotated tomato leaf image dataset.
• Supports both web and mobile interfaces (React / React Native).
• Fast inference using TensorFlow Lite.
• API hosted via FastAPI + Docker + ECS.
• Images are preprocessed and fed through a deployed model which returns prediction and confidence score.

---

🧠 Model Overview

Architecture	Type	Accuracy
CNN	Custom	93%

• Input: Tomato leaf images.
• Output: Predicted disease class and confidence score.
• Optimized for lightweight deployment (TFLite).

---

🧰 Tech Stack

• Backend: FastAPI, TensorFlow, TensorFlow Lite
• Frontend: React, React Native
• DevOps: Docker, Amazon ECS, GCP Cloud Storage
• Libraries: NumPy, PIL, Uvicorn

---

🔍 API Endpoints

Route	Method	Description
/ping	GET	Health check
/predict	POST	Upload an image and get prediction

Request Example

POST /predict
Body: multipart/form-data with an image file

Response Example

{
  "class": "Tomato___Late_blight",
  "confidence": 0.9321
}

---

🖼 Sample Classes

• Tomato___Early_blight
• Tomato___Late_blight
• Tomato___healthy

---

🚀 How to Run Locally

1. Clone Repository

   git clone https://github.com/SteveParadox/Agric_ai.git
   cd Agric_ai/Api

2. Install Dependencies

   pip install -r requirements.txt

3. Run FastAPI App

   uvicorn api:app --reload

4. Access API

   • Visit http://localhost:8000/docs to use Swagger UI

---

📱 Mobile/Web Integration

• React Web App: Upload tomato leaf images from browser.
• React Native App: Capture image via phone camera and send to API.
• Designed to support farmers with easy, quick disease identification.

---

🐳 Docker Support

docker build -t tomato-disease-api .
docker run -p 8000:8000 tomato-disease-api

---

☁️ Cloud Deployment

• Model Hosting: TensorFlow Serving on GCP
• API Server: Docker container on Amazon ECS

---

📁 Repo Structure

├── Api/
│   ├── api.py
│   ├── model_tf_serving.py
│   ├── requirements.txt
├── models/
│   ├── model.h5
│   └── model.tflite
├── Training/
│   └── Tomato_Disease_Prediction.ipynb
├── README.md

---

🧪 Future Improvements

• Add support for more crop types.
• Integrate pest detection.
• Add local language support for UI.
• SMS or WhatsApp bot support for farmers without smartphones.

---

📜 License

MIT License © SteveParadox