I have came across https://docs.openvinotoolkit.org/latest/_intel_models_human_pose_estimation_0001_description_human_pose_estimation_0001.html where openvino provides a pre-trained model based on openpose. It is working faster compared to when I directly test on CPU version of openpose. I currently use https://github.com/seancook/docker-openpose-cpu to run openpose on docker in CPU. The pre-trained openvino model only detects 18 points, while I want to detect 25 points at full resolution as in the cpu docker version. Hence, I wanted to build a model just like the pre-trained model. Since you might have already gone through this process once, to create the pre-trained model, is it possible for you share the steps so that I can try create a new one?