Potential RGB frame loss

yishu_corpex

Hi,

I was working on a test script to get the image timestamp and sequence number using the functions getTimeStamp() and getSequenceNum() on an OAK-D S2 PoE camera. The the color camera is set to 4K resolution. The entire script ran at about 4.6 FPS. I observed here that the sequence numbers are quite off from each other as shown below:

Image Timestamp:  6 days, 21:42:48.038959
Image Sequence Number:  3055
4.638218924021366
Width:  3840  Height:  2160
Image Timestamp:  6 days, 21:42:48.271477
Image Sequence Number:  3062
4.638218924021366
Width:  3840  Height:  2160
Image Timestamp:  6 days, 21:42:48.470779
Image Sequence Number:  3068
4.638218924021366

Additionally I also saw if I update the resolution to 12MP and also add code to get the disparity image then I see the disparity image sequence number is consistent but the RGB image sequence number is not.

RGB Image Sequence Number: 1983
Depth Image Sequence Number: 134

RGB Image Sequence Number: 1997
Depth Image Sequence Number: 135

RGB Image Sequence Number: 2011
Depth Image Sequence Number: 136

RGB Image Sequence Number: 2029
Depth Image Sequence Number: 137

Is it that we are losing the RGB frames here, whereas the depth frames sequence number seems to be consistent?

Thanks
Yishu

erik

Hi @yishu_corpex ,
very likely you are over the bandwidth limit (docs here: https://docs.luxonis.com/software/depthai/optimizing/#Optimizing%20FPS%20%26%20Latency-Bandwidth), so it's dropping color frames, which are the largest. I'd suggest encoding high-res frames to reduce bandwidth and avoid frame-dropping.
Thanks, Erik