|
|
--- |
|
|
title: Crack Detection System |
|
|
emoji: 🔍 |
|
|
colorFrom: red |
|
|
colorTo: yellow |
|
|
sdk: streamlit |
|
|
sdk_version: 1.32.0 |
|
|
app_file: crack-detection.py |
|
|
pinned: false |
|
|
license: apache-2.0 |
|
|
--- |
|
|
|
|
|
# 🔍 Crack Detection System |
|
|
|
|
|
An AI-powered crack detection system using ResNet50 deep learning model. This application can analyze images and detect structural cracks with high accuracy. |
|
|
|
|
|
## Features |
|
|
|
|
|
- **High Accuracy**: ~98% accuracy on test dataset |
|
|
- **ResNet50 Model**: Pre-trained on ImageNet and fine-tuned for crack detection |
|
|
- **Real-time Detection**: Upload images and get instant predictions |
|
|
- **Visual Feedback**: Clear visualization of results with confidence scores |
|
|
|
|
|
## Model Details |
|
|
|
|
|
- **Architecture**: ResNet50 with custom classification head |
|
|
- **Training Dataset**: 40,000 images of cracked and non-cracked surfaces |
|
|
- **Performance Metrics**: |
|
|
- Accuracy: ~98% |
|
|
- AUC: ~99.9% |
|
|
|
|
|
## Classes |
|
|
|
|
|
- **negative**: No crack detected (Class 0) |
|
|
- **positive**: Crack detected (Class 1) |
|
|
|
|
|
## How to Use |
|
|
|
|
|
1. Upload an image (JPG, JPEG, PNG, or BMP format) |
|
|
2. The system will analyze the image |
|
|
3. View the prediction result with confidence score |
|
|
4. Check the debug info in the sidebar for detailed prediction values |
|
|
|
|
|
## Technical Stack |
|
|
|
|
|
- **Framework**: Streamlit |
|
|
- **Deep Learning**: TensorFlow/Keras |
|
|
- **Model**: ResNet50 |
|
|
- **Image Processing**: PIL/Pillow, NumPy |
|
|
|
|
|
## Model Performance |
|
|
|
|
|
The model includes performance visualizations: |
|
|
- Confusion Matrix |
|
|
- ROC Curve |
|
|
- Sample Predictions |
|
|
|
|
|
--- |
|
|
|
|
|
Built with ❤️ using Streamlit and TensorFlow |
|
|
|
