
HandDiff is a novel diffusion-based model that iteratively denoises accurate 3D hand poses from depth images and point clouds, leveraging joint-wise conditions and local detail features to achieve state-of-the-art performance on challenging hand pose benchmarks.


coremsg

handdiff-a-diffusion-based-model-for-accurate-3d-hand-pose-estimation-from-depth-images-and-point-clouds


HandDiff: A Diffusion-based Model for Accurate 3D Hand Pose Estimation from Depth Images and Point Clouds
