Research map / 3D vision / Human pose estimation
EgoM2P: Egocentric Multimodal Multitask Pretraining
Gen Li, Yutong Chen, Yiqian Wu, Kaifeng Zhao, Marc Pollefeys, Siyu Tang
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 EgoGen synthetic data to training (ablation) | EgoM2P · EgoExo4D camera tracking | ATE: 0.028 -> 0.017 | Improved |
| add EgoGen synthetic data to training (ablation) | EgoM2P · ADT camera tracking (unseen) | RRE: 0.593 -> 0.490 | Improved |
| add EgoGen synthetic data to training (ablation) | EgoM2P · H2O video depth | Abs Rel: 0.062 -> 0.055 | Improved |
| add EgoGen synthetic data to training (ablation) | EgoM2P · HOI4D video depth (unseen) | delta 1.25: 97.1 -> 98.0 | Improved |
| increase max visible input/target tokens from 1024 to 2048 (ablation) | EgoM2P pretraining · Pretraining validation | avg loss: 5.80 -> 4.93 | Improved |
| sample datasets proportional to size instead of uniform or log-size (ablation) | EgoM2P pretraining · validation loss across datasets | validation loss | Improved |
| post-train on unseen ADT training set | EgoM2P · ADT camera tracking (unseen) | ATE: 0.032 -> 0.026 | Improved |
| post-train on unseen HOI4D training set | EgoM2P · HOI4D video depth (unseen) | Abs Rel: 0.061 -> 0.041 | Improved |
| post-train on unseen ASE training set | EgoM2P · ASE depth-to-RGB synthesis (unseen) | FVD/10^3: 1.336 -> 0.525 | Improved |
| EgoM2P versus test-time-optimization camera trackers on speed | EgoM2P · Camera tracking runtime per sequence | time: 2.7s -> 0.18s | Improved |
Later papers that compared with this method
1 later accepted paper reports EgoM2P as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- ReViV: Reconstructing the Viewer and the View in 4D from Monocular Egocentric Video (ECCV 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DROID-SLAM — DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras (NeurIPS 2021) · better on 4 of 4 shared results
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
- EgoSim: An Egocentric Multi-view Simulator and Real Dataset for Body-worn Cameras during Motion and Activity (NeurIPS 2024 · similarity 0.91)
- Test-Time Adaptation for Combating Missing Modalities in Egocentric Videos (ICLR 2025 · similarity 0.90)
- Gaze Beyond the Frame: Forecasting Egocentric 3D Visual Span (NeurIPS 2025 · similarity 0.90)
- EgoGen: An Egocentric Synthetic Data Generator (CVPR 2024 · similarity 0.90)
- EgoVid-5M: A Large-Scale Video-Action Dataset for Egocentric Videos Generation (NeurIPS 2025 · similarity 0.89)
- Action Scene Graphs for Long-Form Understanding of Egocentric Videos (CVPR 2024 · similarity 0.89)
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.
