0

I am working on pose estimation for hands and body only using OpenPose (https://github.com/CMU-Perceptual-Computing-Lab/openpose).

I have thousands of videos in my dataset but I can not be able to run OpenPose on each video individually. Because I think I run out of gpu memory which I have GeForce RTX 2080 Ti. I have reduced the frame count of each video to 16 but no luck.

Is there any efficient way that I can extract pose coordinates of hands and body estimations through OpenPose for 20 000 videos?

-- Edit --

I am simply using the following command for each video;

./build/examples/openpose/openpose.bin --image_dir *some_vid-image_location*  --hand --write_json *destination_pose_location*

-- Edit --

Thank you very much.

meandbobbymcgee
  • 341
  • 2
  • 5
  • 12

0 Answers0