Spaces:
Sleeping
Sleeping
| import gradio as gr | |
| def create_demo(): | |
| title = "π¬ EndoSight AI - Development Preview" | |
| description = """ | |
| ## Advanced Gastrointestinal Polyp Detection & Analysis System | |
| **Performance Metrics:** | |
| - **88% mAP\\@0.5** detection accuracy | |
| - **75.6% Dice Score** segmentation accuracy | |
| - **35+ FPS** real-time processing | |
| - **YOLOv8 + U-Net** dual architecture with intelligent stabilization | |
| ### π₯ Clinical Capabilities | |
| - Real-time polyp detection and localization with smooth tracking | |
| - Precise boundary segmentation with fluid heatmap visualization | |
| - Intelligent measurement system with anti-fluctuation technology | |
| - Automated size classification and risk assessment | |
| - Optimized for endoscopy workflow integration | |
| ### π― Technical Highlights | |
| - **Multi-modal AI**: Combined object detection and segmentation | |
| - **Clinical-grade performance**: Sub-second processing times | |
| - **Intelligent measurement**: Size, area, and morphometric analysis with temporal stabilization | |
| - **Adaptive visualization**: Real-time heatmaps that adjust to camera movement | |
| - **Real-time inference**: GPU-accelerated deployment ready | |
| --- | |
| ## π₯ Live System Demonstration | |
| **Real-time EndoSight AI in action** - showcasing automated polyp detection, intelligent segmentation, stabilized measurement analysis, and clinical classification on endoscopy footage. | |
| """ | |
| footer = """ | |
| --- | |
| ## π Development Status & Impact | |
| β **Model Training**: Complete with validation metrics achieved | |
| π **Clinical Integration**: Active collaboration with gastroenterology clinic | |
| π **Performance**: Production-ready accuracy and speed benchmarks | |
| π― **Target**: Characterization of neoplastic versus non-neoplastic lesions | |
| ### π€ Research Collaboration | |
| **Alumnus of**: Universidad Central de Venezuela | |
| **Research Focus**: AI-Assisted Endoscopy & Computer-Aided Diagnosis | |
| ### π Professional Contact | |
| **Technical Lead**: Daniel Cavadia | |
| **Email**: [dan.cavadia@gmail.com](mailto:dan.cavadia@gmail.com) | |
| --- | |
| βοΈ **Medical Disclaimer**: *Research prototype under clinical validation - not intended for diagnostic use.* | |
| """ | |
| # Create simple interface using Gradio 3.34.0 syntax | |
| with gr.Blocks(title="EndoSight AI") as demo: | |
| gr.Markdown(f"# {title}") | |
| gr.Markdown(description) | |
| # Simple video display (no fancy features in older version) | |
| video = gr.Video( | |
| value="demo_video.mp4", | |
| label="EndoSight AI Real-time Detection & Segmentation with Intelligent Measurements" | |
| ) | |
| gr.Markdown(""" | |
| ### π What You're Seeing: | |
| - **Yellow bounding boxes**: Real-time polyp detection with smooth tracking | |
| - **Fluid heatmaps**: Dynamic segmentation visualization that adapts to movement | |
| - **Bottom-left panels**: Stabilized measurement displays with clinical classifications | |
| - **Size categories**: Diminutive (<5mm), Small (5-9mm), Large (β₯10mm) classifications | |
| - **Risk assessment**: Low/Moderate/High risk indicators based on polyp size | |
| - **Measurement accuracy**: Diameter and area calculations with error margins | |
| - **Anti-fluctuation**: Intelligent system prevents measurement swings during camera movement | |
| - **Processing metrics**: Real-time FPS and confidence indicators | |
| - **Multi-polyp detection**: Simultaneous analysis capability with unique tracking IDs | |
| """) | |
| gr.Markdown(footer) | |
| return demo | |
| if __name__ == "__main__": | |
| demo = create_demo() | |
| demo.launch() | |