README.md

metadata

title: Image Gradio
emoji: 🌍
colorFrom: yellow
colorTo: blue
sdk: gradio
sdk_version: 5.47.2
app_file: app.py
pinned: false
license: mit
short_description: Stop sign image predictor interface

Stop Sign Detector

A computer vision application that detects whether an image contains a stop sign using AutoGluon's MultiModalPredictor.

Overview

This application uses deep learning to classify images into two categories:

• Stop Sign: Image contains a stop sign
• Not a Stop Sign: Image does not contain a stop sign

The model analyzes uploaded images in real-time and provides confidence scores for each class.

Features

• Image Upload: Upload images from your device or capture via webcam
• Real-time Classification: Instant predictions as soon as you upload an image
• Confidence Scores: See probability distribution across both classes
• Example Images: Pre-loaded examples to test the model
• Multiple Input Sources: Upload files or use your webcam

How to Use

1. Upload an Image:
   • Click the image area to upload from your device
   • Or click "webcam" to capture a photo in real-time
2. View Results:
   • The model will automatically analyze the image
   • See the predicted class and confidence percentages
3. Try Examples:
   • Click on the example images to see how the model performs

Model Details

• Framework: AutoGluon MultiModalPredictor
• Task: Binary Image Classification
• Model Repository: samder03/2025-24679-image-autogluon-predictor
• Input: RGB images (any size, automatically preprocessed)
• Output: Binary classification with probability scores

Classes

Class ID	Label	Description
0	Not a Stop Sign	Image does not contain a stop sign
1	Stop Sign	Image contains a stop sign

Technical Architecture

The application:

1. Accepts images via Gradio interface (upload or webcam)
2. Saves the image temporarily to disk
3. Loads the image into a pandas DataFrame (AutoGluon format)
4. Runs inference using the MultiModalPredictor
5. Returns probability scores for both classes

Use Cases

• Traffic Sign Recognition: Component for autonomous vehicle systems
• Road Safety Analysis: Automated traffic sign inventory and monitoring
• Educational Tool: Demonstrating computer vision and deep learning
• Dataset Validation: Quickly verify stop sign annotations in datasets

Limitations

• Model is specifically trained for stop signs only
• Performance may vary with:
  • Image quality and resolution
  • Lighting conditions
  • Viewing angles and partial occlusions
  • International stop sign variations
• Not intended for real-time safety-critical applications without further validation

Performance Considerations

• First prediction may take longer due to model loading
• Subsequent predictions are faster (model cached in memory)
• Image preprocessing is automatic

Requirements

gradio
autogluon.multimodal
pandas
Pillow
huggingface_hub