examples(through-wall): ESP32 sensor auto-detection + WiFlow analysis tools

- wiflow_browser.html: auto-detect live ESP32 nodes from the /ws/sensing stream and lock
  them as the model schema (NODE_IDS/CSI_DIM dynamic), persisted + restorable
- wiflow_ab.py: leakage-controlled A/B (chronological/random/blocked-gap/grouped-bucket,
  multi-seed) — the honest CSI→pose evaluation harness
- wiflow_capture.py / wiflow_train.py / wiflow_infer.py: camera-paired capture + train + infer
- pose.html: live WiFi-inferred skeleton viewer; serve.py: static server
- gitignore the regenerable 1.5MB model.npz artifact

Co-Authored-By: claude-flow <ruv@ruv.net>
This commit is contained in:
ruv
2026-06-16 17:00:57 -04:00
parent a784546918
commit 42c764652d
7 changed files with 747 additions and 6 deletions
+3
View File
@@ -277,3 +277,6 @@ aether-arena/staging/
# MM-Fi benchmark dataset archives — large data, fetch separately, never commit
assets/MM-Fi/E0*.zip
assets/MM-Fi/*.zip
# through-wall demo: regenerable trained model artifact
examples/through-wall/model/