Reproduce “Occlusion-Aware Temporally Consistent Amodal Completion for 3D Human-Object Interaction Reconstruction” 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
Hyungjun Doh, Dong In Lee, Seunggeun Chi, Pin-Hao Huang, Kwonjoon Lee, Sangpil Kim, Karthik Ramani
The OTA-3DHOI framework reconstructs photo-realistic, animatable 3D human-object interactions from monocular video. It addresses occlusions and temporal inconsistencies by integrating a template-free occlusion identification strategy with a temporally consistent amodal completion pipeline.
Key insights
- The synergistic combination of bidirectional temporal feature warping and temporal fusion attention is crucial for achieving both high-fidelity content completion and stable temporal coherence in dynamic video sequences.
- Utilizing a template-free occlusion identification method, based on inferred 3D geometry and 2D segmentation, yields more accurate mask generation and enhances generalizability compared to template-dependent approaches.
- Integrating a broader temporal context (e.g., a larger temporal window size) through feature warping consistently improves the accuracy, semantic plausibility, and consistency of amodal completion results.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.