Research map / 3D vision / Human pose estimation
Humans in 4D: Reconstructing and Tracking Humans with Transformers
Shubham Goel, Georgios Pavlakos, Jathushan Rajasegaran, Angjoo Kanazawa, Jitendra Malik
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 |
|---|---|---|---|
| replace ResNet-50 backbone with ViT-H (ablation) | HMR · LSP-Extended | PCK@0.05: 0.35 -> 0.48 | Improved |
| add AVA data to training set (ablation) | HMR 2.0 · 3DPW test | MPJPE (mm): 84.1 -> 80.2 | Improved |
| train from random ViT initialization instead of MAE plus 2D keypoint pretraining (ablation) | HMR 2.0 · 3DPW test | MPJPE (mm): 81.3 -> 98.9 | Hurt |
| add 2D keypoint pretraining after MAE (ablation) | HMR 2.0 · LSP-Extended | PCK@0.05: 0.52 -> 0.53 | Improved |
| use transformer decoder head instead of MLP IEF head (ablation) | HMR 2.0 · LSP-Extended | PCK@0.05: 0.48 -> 0.53 | Improved |
| train longer on more data (HMR 2.0b vs 2.0a) (ablation) | HMR 2.0 · 3DPW test | MPJPE (mm): 70.0 -> 81.3 | Mixed |
| use HMR 2.0 as pose engine in PHALP' tracking | PHALP' · PoseTrack val | ID switches: 623 -> 456 | Improved |
| use full 4DHumans instead of PHALP | tracking · PoseTrack val | ID switches: 541 -> 421 | Improved |
| use ViTDet detector instead of Mask R-CNN in 4DHumans | 4DHumans · PoseTrack val | HOTA: 54.3 -> 57.8 | Improved |
| use HMR 2.0 poses for pose-based action recognition | LART pose-only · AVA v2.2 | mAP: 19.6 -> 22.3 | 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.
- PyMAF — PyMAF: 3D Human Pose and Shape Regression With Pyramidal Mesh Alignment Feedback Loop (ICCV 2021) · better on 2 of 2 shared results
- CLIFF — CLIFF: Carrying Location Information in Full Frames into Human Pose and Shape Estimation (ECCV 2022) · 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
- DETRAM: End-to-end DEtection, Tracking and Recovery of HumAn Meshes (ECCV 2026 · similarity 0.90)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.87)
- MultiPly: Reconstruction of Multiple People from Monocular Video in the Wild (CVPR 2024 · similarity 0.86)
- UniCon3R: Unified Contact-aware 4D Human-Scene Reconstruction from Monocular Video (NeurIPS 2026 · similarity 0.86)
- Synergistic Global-space Camera and Human Reconstruction from Videos (CVPR 2024 · similarity 0.86)
- HumanMM: Global Human Motion Recovery from Multi-shot 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.
