Research map / 3D vision / Depth and stereo
OcclusionFusion: Occlusion-Aware Motion Estimation for Real-Time Dynamic 3D Reconstruction
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the OcclusionFusion full-motion estimator instead of ARAP deformation | Occluded-node motion estimation · DeformingThings4D animal motion subset (100 animations) | EPE (mm): 4.39 -> 3.75 | Helped |
| Use OcclusionFusion reconstruction instead of the strongest listed offline baseline | Dynamic reconstruction · DeepDeform non-rigid reconstruction benchmark | Geometry error (cm): 0.403 -> 0.386 | Helped |
| Use OcclusionFusion reconstruction instead of the strongest listed offline baseline | Dynamic reconstruction · DeepDeform non-rigid reconstruction benchmark | Deformation error (cm): 2.872 -> 2.800 | Helped |
| Add temporal motion through the LSTM (ablation) | OcclusionFusion motion estimator · DeformingThings4D animal motion subset (100 animations) | EPE (mm): 4.02 -> 3.75 | Helped |
| Use per-node Gaussian log-likelihood loss instead of MSE (ablation) | OcclusionFusion motion estimator · DeformingThings4D animal motion subset (100 animations) | EPE (mm): 4.07 -> 3.75 | Helped |
| Add optimization-based ARAP post-processing (ablation) | OcclusionFusion motion estimator · DeformingThings4D animal motion subset (100 animations) | EPE (mm): 3.75 -> 3.28 | Helped |
| Add the complete-node motion constraint E_motion (ablation) | OcclusionFusion reconstruction · DeepDeform non-rigid reconstruction benchmark | Geometry error (cm): 0.406 -> 0.386 | Helped |
| Add the complete-node motion constraint E_motion (ablation) | OcclusionFusion reconstruction · DeepDeform non-rigid reconstruction benchmark | Deformation error (cm): 3.154 -> 2.800 | Helped |
| Use per-node confidence to down-weight uncertain motion (ablation) | OcclusionFusion reconstruction · DeepDeform sequences | geometry error | Helped |
| Use RGB-D optical flow rather than RGB-only RAFT flow | OcclusionFusion reconstruction · Top sequence in the DeepDeform results | average geometry error: 3.79mm -> 3.45mm | Helped |
| Add Gaussian noise to RAFT-RGBD optical flow | OcclusionFusion reconstruction · Top sequence in the DeepDeform results | average geometry error: 3.45mm -> 3.41mm | No clear effect |
Closest papers by meaning
- WALT3D: Generating Realistic Training Data from Time-Lapse Imagery for Reconstructing Dynamic Objects Under Occlusion (CVPR 2024 · similarity 0.86)
- OAMaskFlow: Occlusion-Aware Motion Mask for Scene Flow (AAAI 2025 · similarity 0.85)
- STCOcc: Sparse Spatial-Temporal Cascade Renovation for 3D Occupancy and Scene Flow Prediction (CVPR 2025 · similarity 0.85)
- O^2-Recon: Completing 3D Reconstruction of Occluded Objects in the Scene with a Pre-trained 2D Diffusion Model (AAAI 2024 · similarity 0.85)
- RoHM: Robust Human Motion Reconstruction via Diffusion (CVPR 2024 · similarity 0.85)
- OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation (CVPR 2024 · similarity 0.85)
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.