Research map / 3D vision / Human pose estimation
MixSTE: Seq2seq Mixed Spatio-Temporal Encoder for 3D Human Pose Estimation in Video
Jinlu Zhang, Zhigang Tu, Jianyu Yang, Yujin Chen, Junsong Yuan
Check 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 |
|---|---|---|---|
| use MixSTE instead of PoseFormer | Human3.6M with CPN keypoints, Protocol 1 · Human3.6M Protocol 1 (CPN) | MPJPE (mm): 44.3 -> 40.9 | Improved |
| use MixSTE with ground-truth 2D keypoints instead of Wang et al. | Human3.6M Protocol 1 · Human3.6M Protocol 1 (ground-truth 2D keypoints) | MPJPE (mm): 25.6 -> 21.6 | Improved |
| use MixSTE instead of the strongest listed baseline | MPI-INF-3DHP test set · MPI-INF-3DHP test set | PCK: 88.6 -> 94.4 | Improved |
| fine-tune MixSTE from Human3.6M | MixSTE on HumanEva · HumanEva Protocol 1 | MPJPE (mm): 28.5 -> 16.1 | Improved |
| add alternating spatial and temporal transformer blocks (ablation) | Seq2seq baseline on Human3.6M · Human3.6M | MPJPE (mm): 51.7 -> 45.5 | Improved |
| add joint separation in the temporal transformer (ablation) | Seq2seq with alternating design on Human3.6M · Human3.6M | MPJPE (mm): 45.5 -> 41.7 | Improved |
| add joint separation in the temporal transformer (ablation) | Seq2seq with alternating design on Human3.6M · Human3.6M | FLOPs (M): 186405 -> 645 | Improved |
| replace MPJPE loss with WMPJPE loss (ablation) | MixSTE on Human3.6M · Human3.6M | MPJPE (mm): 41.7 -> 41.3 | Improved |
| add TCLoss to WMPJPE loss (ablation) | MixSTE on Human3.6M · Human3.6M | MPJVE: 4.6 -> 3.6 | Improved |
| add MPJVE loss to WMPJPE loss (ablation) | MixSTE on Human3.6M · Human3.6M | MPJVE: 4.6 -> 2.6 | Improved |
| add the proposed T-Loss to WMPJPE loss (ablation) | MixSTE on Human3.6M · Human3.6M | MPJVE: 4.6 -> 2.3 | Improved |
| increase depth from 4 to 10 with channel 64 and T=27 (ablation) | MixSTE on Human3.6M · Human3.6M | MPJPE (mm): 54.3 -> 51.1 | Improved |
Later papers that compared with this method
10 later accepted papers report MixSTE as a baseline in their results tables; 8 of them report a better number on most of the results they share with it.
- SemGAN: A Semantic and Hierarchical Adversarial Network for 3D Human Pose Estimation (ECCV 2026 · better on 4 of 4 shared results)
- H 2 OT: Hierarchical Hourglass Tokenizer for Efficient Video Pose Transformers (TPAMI 2025 · better on 4 of 4 shared results)
- HiPART: Hierarchical Pose AutoRegressive Transformer for Occluded 3D Human Pose Estimation (CVPR 2025 · better on 1 of 2 shared results)
- PoseAnchor: Robust Root Position Estimation for 3D Human Pose Estimation (ICCV 2025 · better on 2 of 2 shared results)
- 3D Human Pose Estimation via Non-Causal Retentive Networks (ECCV 2024 · better on 5 of 5 shared results)
- Hourglass Tokenizer for Efficient Transformer-Based 3D Human Pose Estimation (CVPR 2024 · better on 3 of 4 shared results)
- ManiPose: Manifold-Constrained Multi-Hypothesis 3D Human Pose Estimation (NeurIPS 2024 · better on 1 of 1 shared result)
- MotionBERT: A Unified Perspective on Learning Human Motion Representations (ICCV 2023 · better on 3 of 3 shared results)
2 more not listed.
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
- MixSynthFormer: A Transformer Encoder-like Structure with Mixed Synthetic Self-attention for Efficient Human Pose Estimation (ICCV 2023 · similarity 0.89)
- SVTformer: Spatial-View-Temporal Transformer for Multi-View 3D Human Pose Estimation (AAAI 2025 · similarity 0.87)
- DiffPose: SpatioTemporal Diffusion Model for Video-Based Human Pose Estimation (ICCV 2023 · similarity 0.87)
- TCPFormer: Learning Temporal Correlation with Implicit Pose Proxy for 3D Human Pose Estimation (AAAI 2025 · similarity 0.86)
- Video-Based Human Pose Regression via Decoupled Space-Time Aggregation (CVPR 2024 · similarity 0.85)
- MoGenTS: Motion Generation based on Spatial-Temporal Joint Modeling (NeurIPS 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.
