---
title: Plant Disease Detection
emoji: 🌿
colorFrom: green
colorTo: green
sdk: gradio
sdk_version: 5.29.0
app_file: app.py
pinned: false
license: mit
tags:
  - plant-disease
  - agriculture
  - computer-vision
  - tensorflow
  - grad-cam
---

# 🌿 Plant Disease Detection

[![Hugging Face Spaces](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https://huggingface.co/spaces/WaqarMughal/Plant_Disease_Detection)

AI-powered plant disease diagnosis using deep learning. Upload leaf images and get instant diagnosis with treatment recommendations through a conversational chatbot interface.

Try the live demo here: **[Plant Disease Detection on Hugging Face](https://huggingface.co/spaces/WaqarMughal/Plant_Disease_Detection)**

## ✨ Features

- **🤖 AI Diagnosis**: ResNet-18 model with **99.17% accuracy** across 38 disease classes
- **🔍 Grad-CAM Visualisation**: See exactly where the AI is looking on the leaf (heatmaps)
- **💊 Treatment Recommendations**: Get actionable advice for each detected disease
- **📊 Confidence Scores**: Know how certain the model is about each prediction
- **💬 Chatbot Interface**: Natural, conversational interaction powered by Gradio 5
- **🖼️ Batch Processing**: Upload multiple images at once
- **⚡ Fast Inference**: Optimized Keras 3 processing on CPU

## 🎯 Model Details

| Property | Value |
|----------|-------|
| **Architecture** | Custom ResNet-18 (trained from scratch) |
| **Training Data** | 54,305 images from PlantVillage dataset |
| **Classes** | 38 disease + healthy classes |
| **Accuracy** | 99.17% on test set |
| **Preprocessing** | CLAHE enhancement in LAB color space |

## 🌱 Supported Crops

- **Fruits**: Apple, Cherry, Peach, Raspberry, Strawberry, Grape, Orange/Citrus, Blueberry
- **Vegetables**: Potato, Tomato, Pepper (Bell), Squash
- **Grains**: Corn (Maize), Soybean

## 🦠 Disease Categories Detected

| Type | Diseases |
|------|----------|
| **Fungal** | Scab, Black Rot, Powdery Mildew, Early/Late Blight, Leaf Spots, Leaf Mold, Rust |
| **Bacterial** | Bacterial Spot, Bacterial Speck |
| **Viral** | Mosaic Virus, Yellow Leaf Curl Virus |
| **Pests** | Spider Mites (Two-spotted) |
| **Healthy** | No disease detected |

## 📖 How to Use

1. **Upload** a clear photo of a plant leaf (drag & drop or click to browse)
2. **Click** "Analyse Leaves" (or auto-runs on upload)
3. **Review** the diagnosis, confidence score, and treatment recommendations
4. **Check** the Grad-CAM overlay to see AI attention areas

## 💡 Tips for Best Results

- Use well-lit, close-up photos of a single leaf
- Focus on the affected area of the leaf
- Ensure the leaf fills most of the frame
- Avoid blurry or dark images
- Try example images from the gallery below the input

## 💻 Running Locally

If you would like to run the Plant Disease Detection app on your own machine, you can easily set it up with the following steps:

1. **Clone the repository**:
   ```bash
   git clone https://huggingface.co/spaces/WaqarMughal/Plant_Disease_Detection
   cd Plant_Disease_Detection
   ```
2. **Install the required dependencies**:
   ```bash
   pip install -r requirements.txt
   ```
3. **Launch the app**:
   ```bash
   python app.py
   ```
4. **Open your browser** and navigate to `http://localhost:7860`

## 📁 Files

| File | Description |
|------|-------------|
| `app.py` | Main Gradio chatbot application |
| `model_best_0.9917_ep040_acc0.9917.keras` | Trained model weights (99.17% accuracy) |
| `normalization_stats.json` | Training dataset normalization statistics |
| `requirements.txt` | Python dependencies |
| `examples/` | Sample leaf images for testing |

## 📊 Output Format

For each uploaded image, you receive:

1. **Diagnosis**: Disease name with severity indicator
2. **Confidence**: Percentage with visual progress bar
3. **Pathogen**: Scientific name of causative agent
4. **Symptoms**: Key visual indicators to look for
5. **Treatment**: Actionable management recommendations
6. **Grad-CAM**: Visual heatmap showing AI attention areas

## 📄 License

MIT License - Feel free to use, modify, and deploy.

## 🙏 Credits

- **Dataset**: PlantVillage (Hughes & Salathé, 2015)
- **Framework**: TensorFlow/Keras, Gradio
- **Visualization**: Grad-CAM (Selvaraju et al., 2017)
- **Author**: Waqar Ul Hasan

## 🔗 Links

- [PlantVillage Dataset Paper](https://arxiv.org/abs/1511.08060)
- [Grad-CAM Paper](https://arxiv.org/abs/1610.02391)
- [Gradio Documentation](https://gradio.app/docs/)