How jj-quesada/bct works
Main Components
A. Input Processing Layer
• Video Input Component (from assets/videos)
B. Core Processing Pipeline
1. Detection System
• Ball Handler Detection Module
• BPDv1 Detection Module
2. Segmentation System
• SegmentV1 Module
• SegmentV2 Module
3. Homography Processing Module
C. Model Management
• YOLO Base Models Repository
• Trained Models Storage
• Training Results Storage
┌─────────────────────────────────────────────┐
• Input Layer │
• ┌─────────────┐ ┌────────────┐ │
• ┴─────────────┴───────────┴────────────┴──┘
• ↓ ↓
┌─────────────────────────────────────────────┐
• Processing Pipeline │
• ┌─────────────┐ ┌────────────────────┐ │
• Detection │←→ │ Homography │ │
• System │ │ Processing │ │
• ┘ └────────────────────┘ │
• ↕ ↕ │
• ┌─────────────┐ │
• Segmentation │ │
• System │ │
• ┘ │
• ┘
• ↑ ↑
┌─────────────────────────────────────────────┐
• Model Repository │
• ┌─────────────┐ ┌────────────┐ │
• YOLO Models │ │Training │ │
• Results │ │
• ┴─────────────┴───────────┴────────────┴──┘
Detailed Component Specifications
A. For each detection model:
• Input resolution
• Specify YOLO version (YOLOv11n)
• Connection to training results
B. For the segmentation models:
• Both versions (segmentv1 and segmentv2)
• Connection to base segmentation model
C. For the homography component:
• Transformation flow
Technical Details to Highlight
• YOLO model architecture (YOLOv11n base)
• Multiple detection and segmentation model versions
• Training pipeline flow
• Video processing pipeline
• Court mapping and transformation process