Related Experiment Video
Updated: May 9, 2025
09:41
Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
Published on: April 21, 2023
1.4K
TexHOI: Reconstructing Textures of 3D Unknown Objects in Monocular Hand-Object Interaction Scenes.
Summary
This study introduces a new method for 3D object texture reconstruction from videos, accurately handling shadows and illumination changes caused by hand-object interactions. It improves texture fidelity by modeling hand visibility and its impact on object appearance.
Area of Science:
- Computer Vision
- Computer Graphics
- 3D Reconstruction
Background:
- Reconstructing high-fidelity 3D object textures from monocular video is difficult due to shadows, indirect illumination, and occlusions from hand-object interactions.
- Accurate object pose estimation is crucial but challenging with occluding hands.
Purpose of the Study:
- To develop a novel approach for high-fidelity 3D object texture reconstruction from monocular frame sequences.
- To address challenges posed by shadows, indirect illumination, and hand-object interactions.
- To accurately estimate object poses and refine surface albedo and illumination.
Main Methods:
- Predicts hand's impact on environmental visibility and indirect illumination on object albedo.
- Learns geometry and low-fidelity texture via composite rendering of radiance fields.
- Optimizes hand and object poses, refines physics-based rendering parameters for precise albedo and illumination.
Main Results:
- Achieves state-of-the-art performance in 3D object texture reconstruction.
- Successfully accounts for hand-object interactions in texture reconstruction.
- Produces precise albedo, accurate hand illumination, and shadow regions.
Conclusions:
- The proposed method significantly advances 3D object texture reconstruction.
- It is the first approach to explicitly model hand-object interactions for this task.
- Offers improved fidelity and accuracy in reconstructing dynamic scenes.
