DCPose
DCPose copied to clipboard
A question about model output
Why not let the model directly output 15 joints corresponding to Posetrack dataset, but output 17 joints corresponding to COCO dataset, and then convert to posetrack?