Research map / 3D vision / 3D generation
RobustFusion: Robust Volumetric Performance Reconstruction Under Human-Object Interactions From Monocular RGBD Stream
Zhuo Su, Lan Xu, Dawei Zhong, Zhong Li, Fan Deng, Shuxue Quan, Lu Fang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add mask refinement for object tracking (ablation) | RobustFusion · Object sequence (sofa/chair) | object MAE (mm): 35.10 -> 11.11 | Improved |
| add interpenetration term (human-object spatial relation) (ablation) | RobustFusion · human-cart interaction (qualitative) | qualitative interpenetration | Improved |
| add pose term in human motion optimization (ablation) | RobustFusion · fast motion / reappearance (qualitative) | qualitative tracking error | Improved |
| add GMM interaction pose prior and LSTM motion prediction prior (ablation) | RobustFusion · sofa occlusion (qualitative) | qualitative pose plausibility | Improved |
| add object-aware adaptive fusion (ablation) | RobustFusion · high-speed and occluded sequences (qualitative) | qualitative geometry | Improved |
| add full human initialization (model plus motion) instead of none or skeleton-only (ablation) | RobustFusion · sequence without self-scanning (qualitative) | qualitative tracking and geometry | Improved |
| use RobustFusion instead of RobustFusion(Conf.) | monocular fusion · Captured human-object sequences (avg of 13) | average projective MAE (mm): 11.73 -> 10.02 | Improved |
| use RobustFusion instead of DoubleFusion | monocular fusion · Captured human-object sequences (avg of 13) | average projective MAE (mm): 18.57 -> 10.02 | Improved |
| use RobustFusion instead of HybridFusion (IMU) on Vicon sequence | monocular fusion · Vicon ground-truth sequence (human only) | average error (m): 0.0417 m vs 0.0546 m | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- DoubleFusion — DoubleFusion: Real-Time Capture of Human Performances With Inner Body Shapes From a Single Depth Sensor (CVPR 2018) · better on 3 of 3 shared results
- HybridFusion — HybridFusion: Real-Time Performance Capture Using a Single Depth Sensor and Sparse IMUs (ECCV 2018) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- RobustFusion: Human Volumetric Capture with Data-driven Visual Cues using a RGBD Camera (ECCV 2020 · similarity 0.94)
- Free-Viewpoint RGB-D Human Performance Capture and Rendering (ECCV 2022 · similarity 0.87)
- Visibility Aware Human-Object Interaction Tracking From Single RGB Camera (CVPR 2023 · similarity 0.87)
- CARI4D: Category Agnostic 4D Reconstruction of Human-Object Interaction (CVPR 2026 · similarity 0.87)
- NeuralHOFusion: Neural Volumetric Rendering Under Human-Object Interactions (CVPR 2022 · similarity 0.87)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
