Sample PY Code for Running YOLOv5 Blob on PreRecorded Video in DepthAI MyriadX?

m9rx

Hi. Similar to this sample PY code for mobilenet [ https://docs.luxonis.com/projects/api/en/latest/samples/MobileNet/video_mobilenet/ ] — Is there sample PY code for running YOLO on pre-recorded video in DepthAI MyriadX?

Preferably with depth, as we also recorded left and right cam frames simultaneous with RGB frames.
But without depth would also be useful. Plus:

Preferably YOLOv5 blob, which we converted from YOLOv5 Pytorch model following your guide and tested successfully on live DepthAI left, RGB and right camera inputs.
But even tiny YOLOv3 and others would be good.

Just want to see how this should be done with YOLO. Thanks!

erik

Hello m9rx ,
For only rgb, the video_mobilenet demo is the one I would suggest modifying - simply replacing MobileNetDetectionNetwork with YoloDetectionNetwork.
Have you recorded left/right/rgb with depthai record app? If so, you should use the replay.py demo as a starting point and add the spatial yolo network to the pipeline.
Thanks, Erik

Roopesh

m9rx , I Assume u might have done a Custom trained yolov5, by any chance was there a problem with SetAchorMask..for..side10 from your side, i face this (if you had come across and solved this issue to get the Depth info ) please let me know. Thanks.