24 lines
1.1 KiB
Plaintext
24 lines
1.1 KiB
Plaintext
> there is some delay on the image shown and the detection, like when I put the logo in from of the camera the detection shows before the image. dont correct that now
|
|
but for to know
|
|
|
|
● Good observation! That's a classic real-time streaming issue where the analysis pipeline is running ahead of the video display pipeline.
|
|
|
|
What's happening:
|
|
- Analysis pipeline: Celery processes video segments immediately when they're created
|
|
- Display pipeline: HLS player has buffering/latency for smooth playback
|
|
- Result: AI detections appear before the video shows that frame
|
|
|
|
Common causes:
|
|
- HLS inherent latency (2-6 seconds typical)
|
|
- Browser video buffering
|
|
- Network buffering in HLS.js
|
|
|
|
Potential solutions (for later):
|
|
- Sync detections with video timestamp
|
|
- Delay analysis results to match video latency
|
|
- Use lower-latency streaming (WebRTC instead of HLS)
|
|
- Add timestamp correlation between segments and detections
|
|
|
|
For now, at least we know the AI pipeline is working fast! The detection accuracy and real-time processing are the main goals achieved. The timing sync is a polish
|
|
issue we can address later. 📡⏰
|