Tomato Leaf Disease Detector

A web app that uses deep learning to detect tomato leaf diseases instantly — running 100% in your browser.

Upload a photo of a tomato leaf and get an instant diagnosis with confidence scoring, disease details, and actionable care recommendations. No server, no signup, no installation — just open and use.

The Problem

Tomato diseases cost farmers billions in crop losses every year. Early and accurate detection is critical — but access to plant pathologists is limited, especially in rural areas. Most existing tools require internet connectivity and server-side processing.

The Solution

A client-side AI web application that runs entirely in the browser. The deep learning model is downloaded once (~16 MB) and all inference happens locally on the user's device using WebAssembly — meaning it works offline, is privacy-preserving (images never leave your device), and requires zero backend infrastructure.

Features

AI Disease Detection: Identifies 5 tomato leaf conditions with confidence percentage.
100% Client-Side: Runs entirely in the browser via ONNX Runtime Web — no server needed.
Smart Image Validation: Pre-screens uploads to reject non-leaf images before running the model.
Care Recommendations: Provides disease descriptions and step-by-step treatment advice.
Responsive Design: Works seamlessly on desktop, tablet, and mobile.
Offline Capable: After first visit, works without internet connection.
Privacy First: Images are processed locally — nothing is uploaded to any server.

Detectable Conditions

Early Blight (Fungal): Dark, target-like spots with concentric rings, starting on older lower leaves.
Late Blight (Oomycete): Large water-soaked lesions with white mold underneath; spreads fast in cool, wet weather.
Septoria Leaf Spot (Fungal): Numerous tiny circular spots with dark borders and gray centers on lower foliage.
Tomato Mosaic Virus (Viral): Mosaic-like light/dark mottling, leaf curling, and stunted plant growth.
Healthy: No disease detected — normal, vibrant green leaf.

How It Works

Upload — Drag & drop or click to select a leaf image (JPG, PNG, JPEG).
Validate — The app checks if the image looks like a leaf using color analysis.
Predict — The EfficientNet-B0 model processes the image and outputs probabilities.
Results — You see the predicted disease, confidence score, and care recommendations.

Everything happens on your device. No data is sent anywhere.

Built With

PyTorch: Model training with transfer learning (EfficientNet-B0).
ONNX Runtime Web: Client-side model inference via WebAssembly.
HTML / CSS / JavaScript: Frontend UI — no frameworks, pure vanilla.
Lucide Icons: Lightweight icon system.
Google Fonts (Outfit): Modern typography.

Design Highlights

Glassmorphism dark-mode interface with frosted glass effects.
Animated progress bar during model loading.
Micro-animations on hover, upload, and result display.
Responsive grid layout that adapts to any screen size.
Color-coded severity badges (green for healthy, red for infection, amber for warning).

The Model

Architecture: EfficientNet-B0 — a lightweight but powerful convolutional neural network.
Training: Transfer learning from ImageNet, fine-tuned on 6,000 labeled tomato leaf images.
Dataset: 5 classes × 1,200 images each (1,000 train / 100 val / 100 test per class).
Export: Converted from PyTorch (.pth) → ONNX (.onnx) for browser compatibility.

Tomato Leaf Disease Predictor

Loading AI Model...

Leaf Image Upload

Analysis Results

Early Blight

Disease Overview

Action Plan & Care Tips

Unrelated Image