Abstract: Real-world robotic grasping can be done robustly if a complete 3D Point Cloud Data (PCD) of an object is available. However, in practice, PCDs are often incomplete when objects are viewed ...
Abstract: While methods that regress 3D human meshes from images have progressed rapidly, the estimated body shapes often do not capture the true human shape. This is problematic since, for many ...
This repository contains the model of the ICCV'2021 paper "Interacting Two-Hand 3D Pose and Shape Reconstruction from Single Color Image". We propose a novel deep learning framework to reconstruct 3D ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback