Research map / 3D vision / Depth and stereo
Self-Supervised Monocular 4D Scene Reconstruction for Egocentric Videos
Chengbo Yuan, Geng Chen, Li Yi, Yang Gao
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 |
|---|---|---|---|
| set shape regularization weight alpha to 4 instead of 1 (ablation) | EgoMono4D · HOI4D | F1 (%): 18.2 -> 26.9 | Improved |
| add confidence mask regularization loss (ablation) | EgoMono4D · HOI4D | F1 (%): 23.9 -> 26.9 | Improved |
| use larger UniDepth backbone instead of UniDepth-S (ablation) | EgoMono4D · HOI4D | F1 (%): 23.8 -> 26.9 | Improved |
| add video adaptor blocks (ablation) | EgoMono4D · HOI4D | F1 (%): 25.8 -> 26.9 | Improved |
| use 3D shape regularization against UniDepth instead of relative-scale midas depth loss (ablation) | EgoMono4D · HOI4D | F1 (%): 26.1 -> 26.9 | Improved |
| use 3D flow/track loss instead of 2D pixel-space flow loss (ablation) | EgoMono4D · HOI4D | F1 (%): 26.7 -> 26.9 | No clear effect |
| add camera intrinsic consistency loss (ablation) | EgoMono4D · HOI4D | F1 (%): 25.2 -> 26.9 | Improved |
| use inference window Nw equal to training window of 4 (ablation) | EgoMono4D · HOI4D | Chamfer Distance (mm): 17.9 -> 5.9 | Improved |
| increase window overlap No from 1 to 2 or 3 (ablation) | EgoMono4D · HOI4D | F1 (%): 27.9 -> 27.9 | No clear effect |
| lower video frame rate to fps/12 (ablation) | EgoMono4D · POV-Surgery | Chamfer Distance (mm): 25.5 -> 28.5 | Hurt |
| self-supervised EgoMono4D vs supervised DUSt3R-family baselines on egocentric zero-shot reconstruction | EgoMono4D · POV-Surgery (zero-shot) | F1 (%): 9.7 -> 13.5 | Improved |
| end-to-end self-supervised training vs modularized no-training version | EgoMono4D · HOI4D | Chamfer Distance (mm): 8.9 -> 5.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DUSt3R — DUSt3R: Geometric 3D Vision Made Easy (CVPR 2024) · better on 4 of 4 shared results
- MonSt3R — MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion (ICLR 2025) · better on 1 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
- ReViV: Reconstructing the Viewer and the View in 4D from Monocular Egocentric Video (ECCV 2026 · similarity 0.90)
- EgoMo3R: Joint Egocentric Motion and Scene Reconstruction (NeurIPS 2026 · similarity 0.90)
- Ego-1K - A Large-Scale Multiview Video Dataset for Egocentric Vision (CVPR 2026 · similarity 0.90)
- Layered Motion Fusion: Lifting Motion Segmentation to 3D in Egocentric Videos (CVPR 2025 · similarity 0.87)
- EgoVid-5M: A Large-Scale Video-Action Dataset for Egocentric Videos Generation (NeurIPS 2025 · similarity 0.87)
- 3D Human Pose Perception from Egocentric Stereo Videos (CVPR 2024 · similarity 0.87)
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.
