2 Commits

Author	SHA1	Message	Date
ruv	a1c411a6ba	feat(training): two-checkerboard camera↔room calibration for ADR-079 labels (ADR-152 §2.1.3) ... Defends the camera-supervised pipeline against PerceptAlign's "coordinate overfitting": MediaPipe keypoints were emitted in raw camera coordinates with no shared frame and no transceiver-geometry metadata — the exact label shape that memorizes deployment layout and collapses cross-layout. - scripts/calibrate-camera-room.py + calibration_lib.py: OpenCV two-checkerboard calibration → versioned bundle JSON (intrinsics, camera→room extrinsics, checkerboard spec, transceiver geometry, sha256 calibration_id). Intrinsics resolve from file > cache > multi-view computation > loud-warning 2-view fallback. - collect-ground-truth.py --calibration <bundle>: every sample gains keypoints_room (unit bearing rays from the camera center in the room frame — documented projective alignment; raw image coords preserved so training chooses), camera_origin_room, calibration_id, and the transceiver geometry stamp. Without the flag, output is byte-identical to before (tested) + a one-line ADR-152 warning. Design finding (recorded for ADR-152): a single planar checkerboard's corner grid is centrosymmetric — the reversed corner ordering fits a ghost camera pose with IDENTICAL reprojection error, so per-board flip disambiguation is mathematically ill-posed. solve_two_board_extrinsics solves the joint wall+floor set over all 4 flip combinations, where the minimum is unique — an independent reason the TWO-checkerboard method is required, beyond what PerceptAlign states. 15 headless pytest tests green (synthetic corners: extrinsics recovery incl. ghost resolution, bundle round-trip + hash stability, ray transforms w/ distortion + cross-resolution, no-calibration byte identity). Co-Authored-By: RuFlo <ruv@ruv.net>	2026-06-10 16:00:32 -04:00
ruv	e3522ddcda	feat: camera ground-truth training pipeline (ADR-079, #362) ... Add 4 scripts for camera-supervised WiFlow pose training: - collect-ground-truth.py: synchronized webcam + CSI capture via MediaPipe PoseLandmarker (17 COCO keypoints at 30fps) - align-ground-truth.js: time-align camera keypoints with CSI windows using binary search, confidence-weighted averaging - train-wiflow-supervised.js: 3-phase supervised training (contrastive pretrain → supervised keypoint regression → bone-constrained refinement) with curriculum learning and CSI augmentation - eval-wiflow.js: PCK@10/20/50, MPJPE, per-joint breakdown, baseline proxy mode for benchmarking Baseline benchmark (proxy poses, no camera supervision): PCK@10: 11.8% | PCK@20: 35.3% | PCK@50: 94.1% | MPJPE: 0.067 Camera pipeline validated over Tailscale to Mac Mini M4 Pro (1920x1080, 14/17 keypoints visible, MediaPipe confidence 0.94-1.0). Target after camera-supervised training: PCK@20 > 50% Closes #362 Co-Authored-By: claude-flow <ruv@ruv.net>	2026-04-06 14:07:25 -04:00