A newer version of the Gradio SDK is available:
6.2.0
metadata
title: Road Sign Recognition System
emoji: 🚦
colorFrom: red
colorTo: yellow
sdk: gradio
sdk_version: 4.44.0
app_file: app.py
pinned: false
license: mit
🚦 Road Sign Recognition System
Overview
An intelligent computer vision application that identifies and classifies road signs from images using AutoGluon's MultiModal Predictor. This system helps with driver education, navigation assistance, and road safety analysis.
Model Details
- Model Type: AutoGluon MultiModal Predictor
- Architecture: Deep learning with vision transformer backbone
- Source Model: its-zion-18/sign-image-autogluon-predictor
- Task: Multi-class image classification for road sign detection
- Training: AutoGluon automatic model selection and hyperparameter tuning
Limitations
- Not Enough Classes Currently only have 2 classes (Stop and Speed limit)
Features
Input Methods
- File Upload: Support for JPG, PNG, and other common image formats
- Webcam Capture: Real-time photo capture for immediate classification
- Clipboard Paste: Quick image input from clipboard
- Example Images: Pre-loaded road sign samples for testing
Output Information
- Primary Classification: Most likely sign type with confidence score
- Top Predictions: Alternative classifications with probabilities
- Confidence Visualization: Progress bars showing probability distribution
- Detailed Analysis: Interpretation of confidence levels
How to Use
- Upload or Capture: Provide an image of a road sign
- Automatic Processing: Model analyzes the sign immediately
- View Results: See the classification and confidence scores
- Try Examples: Test with provided sample images
Best Practices
- Use clear, well-focused images
- Ensure the entire sign is visible
- Avoid excessive angles or distortion
- Good lighting improves accuracy
- One sign per image works best
Technical Specifications
- Framework: AutoGluon for automated machine learning
- Backend: PyTorch for deep learning operations
- Deployment: Hugging Face Spaces
- Interface: Gradio for user interaction
Applications
- Driver education and training
- Navigation system enhancement
- Municipal sign inventory
- Road safety assessment
- Autonomous vehicle development
Limitations
- Optimized for standard road signs
- Single sign detection per image
- May vary with weather conditions
- Region-specific signs may not be recognized
Privacy
- No image storage after processing
- No personal data collection
- Server-side processing only
Created for CMU 24-679 Course Assignment