ruv
ba82fcfc37
feat: camera-free 17-keypoint pose training (10 sensor signals)
Multi-modal pipeline using PIR, BME280, reed switch, vibration,
RSSI triangulation, subcarrier asymmetry — no camera needed.
Phases: multi-modal collection → weak label generation → enhanced
contrastive → 5-keypoint pose proxy → 17-keypoint interpolation
→ self-refinement (3 rounds) → LoRA + TurboQuant + EWC
Validated: 2,360 frames, 100% presence, 0 skeleton violations,
82.8 KB model (8 KB at 4-bit), 114.8s training
Co-Authored-By: claude-flow <ruv@ruv.net>
2026-04-02 23:05:07 -04:00
..
2026-03-14 13:39:51 -04:00
2026-04-02 22:04:57 -04:00
2026-04-02 22:40:48 -04:00
2026-03-14 13:39:51 -04:00
2026-04-02 22:04:57 -04:00
2026-03-03 11:35:57 -05:00
2026-04-02 22:04:57 -04:00
2026-03-15 11:49:29 -04:00
2026-03-01 15:51:38 -05:00
2026-03-14 13:39:51 -04:00
2026-03-14 13:39:51 -04:00
2026-04-02 22:42:32 -04:00
2026-03-15 16:10:10 -04:00
2026-03-10 08:40:47 -04:00
2026-04-02 22:04:16 -04:00
2026-04-02 22:04:16 -04:00
2026-03-15 11:49:29 -04:00
2026-03-14 13:39:51 -04:00
2026-03-14 13:39:51 -04:00
2026-03-14 13:39:51 -04:00
2026-03-14 13:39:51 -04:00
2026-03-14 13:39:51 -04:00
2026-04-02 19:32:18 -04:00
2026-04-02 19:32:18 -04:00
2026-03-14 13:39:51 -04:00
2026-04-02 23:05:07 -04:00
2026-04-02 22:40:48 -04:00
2026-04-02 22:04:57 -04:00
2026-03-14 13:39:51 -04:00
2026-03-15 11:49:29 -04:00