Detect wheat diseases instantly — from a photo

A lightweight, accurate MobileNetV2-based model that identifies wheat leaf diseases and provides treatment suggestions and cost estimates. Built for farmers and extension workers.

Fast Inference

Optimized for low latency

Practical Advice

Dosage & cost estimates

Mobile Ready

Lightweight model for edge later

Project Overview

This project automates early detection of wheat diseases using computer vision. Farmers can upload a photo of a leaf and receive an instant diagnosis and recommended treatments with dosage and cost estimates. The backend uses PyTorch and FastAPI, while the frontend is a lightweight HTML app for easy access.

How it helps

  • Faster diagnosis
  • Lower cost due to targeted treatment
  • Easy access via smartphones

Technical Summary

Transfer-learning on MobileNetV2 with standard preprocessing and augmentations. Training done on Colab/GPU. Model exported as .pth and loaded by FastAPI for inference. Treatment mapping provides pesticide suggestions.

Pipeline

  1. Image upload
  2. Preprocessing
  3. Model prediction
  4. Treatment mapping

Team

Abhishek Kahate
ML & Backend
Role: model training, API
Tanhvi Shanware
Frontend & UX
Role: UI, deployment
Krish Giri
Data & Testing
Role: dataset curation