paTRICK-swk / D3DP

[ICCV2023] The PyTorch implementation for "Diffusion-Based 3D Human Pose Estimation with Multi-Hypothesis Aggregation"
MIT License
158 stars 8 forks source link

Video handling for video frames less than receptive field #13

Open anas-zafar opened 1 year ago

anas-zafar commented 1 year ago

Thanks @paTRICK-swk for the amazing work. I am having this issue when I try to run videos in the wild with frames less than the receptive field (243) the 2D and 3D predictions go out of sync. I found you are handling it by duplicating the last frame this is causing the desync. Could please guide me how should I handle this?