Reproduce “HMPDM: A Diffusion Model for Driving Video Prediction with Historical Motion Priors” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ke Li, Tianjia Yang, Kaidi Liang, Xianbiao Hu, Ruwen Qin
HMPDM predicts future driving scenes from RGB-only video by keeping past frames as context, encoding their motion at multiple spatial scales, and conditioning each denoising step on an earlier prediction. On Cityscapes’ full test set, with two past and 28 future frames at 128×128, it reports FVD 77.0 versus STDiff’s 107.3; its SSIM and LPIPS are slightly worse.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.