Skip to content

FACE_DETECTION_REASON.md

Latest commit

 

History

History
40 lines (23 loc) · 1.27 KB

FACE_DETECTION_REASON.md

File metadata and controls

40 lines (23 loc) · 1.27 KB
Raw

Face Detection + Song Matching Issue (Reason Log)

Date: 2026-03-05

1) Console message from vision_wasm_internal.js

The logs below are warnings/info from MediaPipe/TensorFlow Lite, not a crash:

  • INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
  • Feedback manager requires a model with a single signature inference...

These lines usually appear with the current face-landmarker model and do not block detection.

2) Actual issue causing "No matching song found for this mood"

Backend route order in Backend/src/app.js had a wildcard route before API routes.

Because of that, requests like:

  • GET /api/song/getsong?mood=happy

can be intercepted by the wildcard fallback and return public/index.html (HTML) instead of JSON.
Frontend then receives non-song data and ends up with an empty song list, showing:

  • No matching song found for this mood.

3) Additional backend data issue found

Backend/src/model/song.model.js used:

  • default: 'nutral'

This is a typo and not aligned with enum values. It should be Neutral.

4) Fix applied

  • Move API routes before fallback route.
  • Restrict fallback to GET and place it at the end.
  • Fix mood default typo in song schema.
  • Add frontend response-shape validation in song fetch hook for clearer errors.