I have been able to successfully get the output of the 2d skeleton pose estimation by running the following command:
~/isaac/sdk$ bazel run packages/skeleton_pose_estimation/apps/openpose:openpose_inference
After that, I was wondering if it’s possible to run this for my own input images. I have generated synthetic data from Isaac Sim on Omniverse and I have some png files that I like to test the pose estimation on it.
when I tried this earlier the humans were not recognized properly and something else was being estimated wrongly. Is there a way to solve this? I thought the model was pre-trained and could already estimate the human poses in different pictures other than the sample image provided.