I found a bug that the number of the key-points of the HRNet-based pose estimation that you provide in this repo is 17, which is not equal to the openpose setting (18) and your provided trained model is based on the openpose-based data. This bug makes the one-by-one video test in this repo incorrect and the overall performance is unknown for HRNet-based settings.
Could you please provide a pre-trained model that based on the HRNet-based skeleton estimation?
I found a bug that the number of the key-points of the HRNet-based pose estimation that you provide in this repo is 17, which is not equal to the openpose setting (18) and your provided trained model is based on the openpose-based data. This bug makes the one-by-one video test in this repo incorrect and the overall performance is unknown for HRNet-based settings.
Could you please provide a pre-trained model that based on the HRNet-based skeleton estimation?