Spaces:

---

fcakyon

/

sahi-yolo11

like 33

Running on Zero

This pull request introduces several significant enhancements to the SAHI + YOLOv11 Gradio demonstration. Key changes include:

• Dynamic Model Selection: Added a dropdown menu allowing users to select between various YOLOv11 models (yolo11n.pt, yolo11s.pt, yolo11m.pt, yolo11l.pt, yolo11x.pt), enabling flexible experimentation with different model scales.
• Intuitive UI Controls: Converted confidence_threshold, overlap_width_ratio, overlap_height_ratio, and postprocess_match_threshold from numerical inputs to user-friendly sliders, improving interactive control.
• Max Detections Control: Introduced a new slider for max_detections, allowing users to limit the number of displayed bounding box predictions for both standard and sliced inference.
• Code Refactoring and Robustness:
  • Transitioned the Gradio interface from gr.Interface to gr.Blocks for more structured UI management.
  • Implemented a dedicated load_yolo_model function to handle dynamic model loading, centralizing model initialization.
  • Added Google-style docstrings to load_yolo_model and sahi_yolo_inference for enhanced code readability and maintainability.
• MCP Compatibility: Configured the application launch with mcp_server=True for seamless integration into Multi-Container Pod (MCP) environments.