Research map / 3D vision / Human pose estimation
3D Human Pose, Shape and Texture from Low-Resolution Images and Videos
Xiangyu Xu, Hao Chen, Francesc Moreno-Noguer, László A. Jeni, Fernando De la Torre
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 self-supervision loss to baseline network (ablation) | Ba · 3DPW (32x32) | MPJPE: 143.63 -> 128.88 | Improved |
| add self-supervision loss to resolution-aware network (ablation) | RA · 3DPW (32x32) | MPJPE: 135.29 -> 124.23 | Improved |
| replace baseline network with resolution-aware network (basic loss) (ablation) | Ba · 3DPW (32x32) | MPJPE: 143.63 -> 135.29 | Improved |
| add contrastive feature loss on top of RA+SS (ablation) | RA+SS · 3DPW (32x32) | MPJPE: 124.23 -> 117.12 | Improved |
| use MSE feature loss instead of contrastive loss (ablation) | RA+SS · 3DPW (32x32) | MPJPE: 117.12 -> 124.85 | Hurt |
| use cosine distance instead of full contrastive loss (ablation) | RA+SS · 3DPW (32x32) | MPJPE: 117.12 -> 125.08 | Hurt |
| add progressive training (ablation) | RSC-Net · 3DPW (32x32) | MPJPE: 127.05 -> 117.12 | Improved |
| use original symmetric self-supervision instead of directional loss (ablation) | RSC-Net · 3DPW (32x32) | MPJPE: 117.12 -> 156.25 | Hurt |
| use hierarchical supervision from all higher resolutions instead of only the highest (ablation) | RSC-Net · 3DPW (32x32) | MPJPE: 122.01 -> 117.12 | Improved |
| add global context module and upsampling layer to texture network (ablation) | Texture network · Market-1501 | ReID loss: 60.21 -> 59.89 | Improved |
| fuse global features by concatenation instead of summation (ablation) | Texture network · Market-1501 | ReID loss: 59.89 -> 60.25 | Hurt |
| add temporal recurrent module to single-frame model | RSC-Net · 3DPW video (32x32) | MPJPE: 141.87 -> 123.14 | Improved |
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.
- SPIN — SPIN: Structure-Preserving Inner Offset Network for Scene Text Recognition (AAAI 2021) · better on 6 of 6 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
- 3D Human Shape and Pose from a Single Low-Resolution Image with Self-Supervised Learning (ECCV 2020 · similarity 0.98)
- Thin-Slicing Network: A Deep Structured Model for Pose Estimation in Videos (CVPR 2017 · similarity 0.87)
- Deep High-Resolution Representation Learning for Human Pose Estimation (CVPR 2019 · similarity 0.87)
- Super-Resolution 3D Human Shape from a Single Low-Resolution Image (ECCV 2022 · similarity 0.87)
- Self-Supervised 3D Human Pose Estimation via Part Guided Novel Image Synthesis (CVPR 2020 · similarity 0.86)
- Learning to Estimate 3D Human Pose and Shape From a Single Color Image (CVPR 2018 · 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.
