Views
No views yet
RD_REPORT.md — Full R&D report (50+ pages) covering:enhanced/ — Prototype implementation:config.py — Full configuration system with Edge/Production/Research presetsdetector.py — YOLOv11 / D-FINE / RT-DETR detector with mock fallbacktracker.py — OC-SORT multi-object tracker (replaces ByteTrack)feature_extractor.py — DINOv2 embedding extractorreid_manager.py — EMA gallery for card re-entry detectionreference_matcher.py — SuperPoint + LightGlue reference matchingquality_scorer.py — Card quality assessmentpipeline.py — End-to-end pipeline orchestrationscripts/ — Evaluation and data generation:evaluate_detection.py — mAP, precision/recall evaluationevaluate_tracking.py — MOTA, IDF1, ID-switch metricsevaluate_reid.py — ROC/AUC, CMC, EER evaluationgenerate_synthetic_cards.py — Privacy-safe synthetic ID card generatortests/ — Integration tests:test_system.py — 68 tests covering all components| Component | Model | HF Model | Latency |
|---|---|---|---|
| Detector | D-FINE-S | Fine-tune from D-FINE GitHub | 3.5ms (T4) |
| Tracker | BoT-SORT-ReID | Custom + DINOv2 | ~5ms |
| Re-ID | DINOv2-base | facebook/dinov2-base | ~15ms |
| Reference Match | SuperPoint + LightGlue | ETH-CVG/lightglue_superpoint | sub-10ms |
| Total | ~25ms |
1from enhanced.config import PipelineConfig
2from enhanced.pipeline import IDCardPipeline
3
4config = PipelineConfig.production_config()
5pipeline = IDCardPipeline(config)
6
7# Process a frame
8result = pipeline.process_frame(frame)
9for card in result.cards:
10 print(f"Card {card.track_id}: {card.quality_label} (conf={card.confidence:.2f})")1pip install opencv-python-headless numpy faker
2python tests/test_system.py
3# Expected: 68/68 passed