elirtf/opus
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

feat: CPU object detection, motion-gated recording, production deployment improvements #12

Merged
elirtf merged 1 commit from feat/adaptive-streaming into master 2026-06-17 15:25:39 +00:00
Conversation 0 Commits 1 Files changed 39 +2143 -256
elirtf commented 2026-06-17 15:25:33 +00:00
Owner
Copy link

Object detection (CPU-only, OpenCV DNN):

  • MobileNet-SSD as second stage after motion — runs once per event, not per frame
  • Global semaphore bounds concurrent inference across camera workers
  • Graceful fallback when model files absent; no new Python deps beyond opencv
  • Detected labels stored in recording_event.detected_objects (comma-separated)
  • Notification messages include object labels when available (e.g. 'Person detected on Front Door')

Continuous motion-gated recording:

  • Replace old clip-per-cooldown model with FFmpeg subprocess that lives while
    motion persists + configurable linger window
  • Proper SIGTERM/process lifecycle management for clean clip finalisation
  • Re-arm gate (motion_cooldown_seconds) prevents rapid restart

Production deployment:

  • OPUS_ROLE env var to split API/processor/recorder across separate processes
  • go2rtc API/RTSP bind addresses configurable via GO2RTC_API_LISTEN/GO2RTC_RTSP_LISTEN
    for bare-metal security hardening
  • .dockerignore cleanup (secrets, runtime data excluded from image layers)
  • Docker build --no-cache-dir, NGINX hardening, health sampler improvements

Other:

  • Prometheus metrics for motion detection pipeline
  • Recording settings UI improvements (default duration display, NAS path hints)
  • Motion detection tuning: configurable contrast enhancement, lightning rejection,
    frame-delta accumulation alpha, per-camera motion mask presets
  • Auth: session refresh endpoint, go2rtc stream validation on login
  • Worker status server: expose recording state (active/since) for frontend
Object detection (CPU-only, OpenCV DNN): - MobileNet-SSD as second stage after motion — runs once per event, not per frame - Global semaphore bounds concurrent inference across camera workers - Graceful fallback when model files absent; no new Python deps beyond opencv - Detected labels stored in recording_event.detected_objects (comma-separated) - Notification messages include object labels when available (e.g. 'Person detected on Front Door') Continuous motion-gated recording: - Replace old clip-per-cooldown model with FFmpeg subprocess that lives while motion persists + configurable linger window - Proper SIGTERM/process lifecycle management for clean clip finalisation - Re-arm gate (motion_cooldown_seconds) prevents rapid restart Production deployment: - OPUS_ROLE env var to split API/processor/recorder across separate processes - go2rtc API/RTSP bind addresses configurable via GO2RTC_API_LISTEN/GO2RTC_RTSP_LISTEN for bare-metal security hardening - .dockerignore cleanup (secrets, runtime data excluded from image layers) - Docker build --no-cache-dir, NGINX hardening, health sampler improvements Other: - Prometheus metrics for motion detection pipeline - Recording settings UI improvements (default duration display, NAS path hints) - Motion detection tuning: configurable contrast enhancement, lightning rejection, frame-delta accumulation alpha, per-camera motion mask presets - Auth: session refresh endpoint, go2rtc stream validation on login - Worker status server: expose recording state (active/since) for frontend
Object detection (CPU-only, OpenCV DNN):
- MobileNet-SSD as second stage after motion — runs once per event, not per frame
- Global semaphore bounds concurrent inference across camera workers
- Graceful fallback when model files absent; no new Python deps beyond opencv
- Detected labels stored in recording_event.detected_objects (comma-separated)
- Notification messages include object labels when available (e.g. 'Person detected on Front Door')

Continuous motion-gated recording:
- Replace old clip-per-cooldown model with FFmpeg subprocess that lives while
  motion persists + configurable linger window
- Proper SIGTERM/process lifecycle management for clean clip finalisation
- Re-arm gate (motion_cooldown_seconds) prevents rapid restart

Production deployment:
- OPUS_ROLE env var to split API/processor/recorder across separate processes
- go2rtc API/RTSP bind addresses configurable via GO2RTC_API_LISTEN/GO2RTC_RTSP_LISTEN
  for bare-metal security hardening
- .dockerignore cleanup (secrets, runtime data excluded from image layers)
- Docker build --no-cache-dir, NGINX hardening, health sampler improvements

Other:
- Prometheus metrics for motion detection pipeline
- Recording settings UI improvements (default duration display, NAS path hints)
- Motion detection tuning: configurable contrast enhancement, lightning rejection,
  frame-delta accumulation alpha, per-camera motion mask presets
- Auth: session refresh endpoint, go2rtc stream validation on login
- Worker status server: expose recording state (active/since) for frontend
elirtf merged commit d295e51aa9 into master 2026-06-17 15:25:39 +00:00
elirtf deleted branch feat/adaptive-streaming 2026-06-17 15:25:39 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
elirtf/opus!12
No description provided.