How to solve significant frame-to-frame jitter

[Maxwell-Zhao]

I notice that single-image based approaches often suffer from significant frame-to-frame jitter. I'm curious about how you mitigated this issue in your research.

[TairanHe]

what is the single-image based approach that you are referring to?

[Maxwell-Zhao]

I notice that approaches that only use single image to get mocap results, such as HybrIK. They estimate SMPL parameters without considering the previous and next frames, often suffer from significant frame-to-frame jitter. The lack of temporal consistency leads to incoherent predictions between frames, causing the final SMPL mesh to exhibit noticeable shaking or jitter. How to solve this in this project?