Flask inference API for a custom CNN that classifies apple leaf diseases across 4 classes from the PlantVillage dataset. Accepts an image, runs it through a Keras .h5 model, and returns the predicted class with per-class confidence scores.

Frontend repo: cnn-frontend
Live demo: app-dlite-net.vercel.app

Accepts a leaf image and returns the top predicted class with confidence scores.

Request

Content-Type: multipart/form-data
Body: image (file)

Response

{
  "predicted_class": "Apple___Apple_scab",
  "confidence": 0.9812,
  "all_scores": {
    "Apple___Apple_scab": 0.9812,
    "Apple___Black_rot": 0.0103,
    ...
  }
}

Returns 200 OK if the service is running and the model is loaded.

Prerequisites: Python 3.8+, pip

git clone https://github.com/Somansh1/cnn-backend.git
cd cnn-backend
pip install -r requirements.txt

Place your trained model file in the project root (or update the path in app.py):

cnn-backend/
├── app.py
├── model.h5          # ← your trained Keras model
├── requirements.txt
├── Procfile
└── scripts/

python app.py         # starts on http://localhost:5000

Configured for Heroku via Procfile:

heroku create your-app-name
git push heroku main

The scripts/ directory contains shell helpers for environment setup.

• Frontend: React app with image upload and confidence score display — cnn-frontend
• Training notebook: Available on request

MIT