How do I run keypoints model using a video?

Firstly, I am very thankful for the work done on this piece of work! It is a million times simpler to set up and run than the alternatives!

I wanted to know if I am doing the demo.py script incorrectly for videos. I have tried running it using the --video-input tag, and the videos attempted were .mp4 .mov and .avi, but when I check the output folder, the result doesn’t appear as it did for the image.

This is the command I’ve run: python3.6 demo/demo.py --config-file configs/COCO-Keypoints/keypoint_rcnn_R_50_FPN_1x.yaml --video-input ~/IMG_2172.avi --output output/test.mp4 --opts MODEL.WEIGHTS detectron2://COCO-Keypoints/keypoint_rcnn_R_50_FPN_1x/137261548/model_final_04e291.pkl

I believe it is a case where I’m missing something, but I’m not sure what.

Any advice or suggestions are welcomed. Let me know if any further information is needed. Much thanks in advance.

[ppwwyyxx]

Just changing “x264” to some other codec (e.g. “MPEG”) should work.

[malcolmgarner-movement]

Aha that wooorrrkkkssss!!!

Thanks a lot @ppwwyyxx! Much appreciated. I changed it to “mp4v” and that solved it’s issue with opencv for mp4