[Question] Pose-only inference

[iamrobzy]

I plan on fine-tuning your StablePose model on my own dataset, but I only have pose-image pairs. I understand that the model performs better with binary pose masks. How would you recommend extracting pose masks from poses? Is it feasible to do this during inference time? Thanks.