Research map / 3D vision / Human pose estimation
DGPose: Deep Generative Models for Human Body Analysis
Rodrigo de, Arnab Ghosh, Thalaiyasingam Ajanthan, Ondřej Mikšík, Adnane Boukhayma, N. Siddharth, Philip H. S. Torr
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 |
|---|---|---|---|
| use Conditional-DGPose instead of ClothNet-body | pose-conditioned generation · ChictopiaPlus | PSNR: 16.89 -> 21.33 | Improved |
| use Conditional-DGPose instead of ClothNet-body | pose reconstruction · ChictopiaPlus | PCK@0.5: 70.89% -> 95.14% | Improved |
| use Conditional-DGPose (generative, unpaired) instead of PG2 (paired image-to-image) | pose-guided generation · DeepFashion | PSNR / SSIM: PSNR 18.96 -> 18.38; SSIM 0.83 -> 0.79 | Hurt |
| add extended joints (28 joints) to the 2D vector pose representation (ablation) | Conditional-DGPose · Validation set | L1 reconstruction error: 14.52 -> 13.91 | Improved |
| use heatmaps instead of 2D vector (14 joints) (ablation) | Conditional-DGPose · Validation set | L1 reconstruction error: 14.52 -> 13.55 | Improved |
| add 9 rigid-part and 1 whole-body heatmaps to the joint heatmaps (ablation) | Conditional-DGPose · Validation set | L1 reconstruction error: 13.55 -> 13.41 | Improved |
| reduce label supervision in Semi-DGPose from 100% to 25% (ablation) | Semi-DGPose · Human3.6M | PSNR: 22.27 -> 20.06 | Hurt |
| reduce label supervision in Semi-DGPose from 100% to 25% (ablation) | Semi-DGPose · Human3.6M | PCK@0.5: 93.85% -> 88.35% | Hurt |
| add semi-supervised pose estimation task to the generative model (Semi-DGPose vs Conditional-DGPose) | DGPose · DeepFashion | PSNR / SSIM: PSNR 18.38 -> 16.84; SSIM 0.79 -> 0.76 | Hurt |
| set regression loss weight alpha to 100 (ablation) | Semi-DGPose · Human3.6M | PCK | Improved |
Closest papers by meaning
- A Generative Model of People in Clothing (ICCV 2017 · similarity 0.88)
- Disentangled Pose and Appearance Guidance for Multi-Pose Generation (CVPR 2025 · similarity 0.87)
- StructLDM: Structured Latent Diffusion for 3D Human Generation (ECCV 2024 · similarity 0.86)
- Unsupervised Person Image Generation With Semantic Parsing Transformation (CVPR 2019 · similarity 0.86)
- PoseGen: Learning to Generate 3D Human Pose Dataset with NeRF (AAAI 2024 · similarity 0.85)
- ChatPose: Chatting about 3D Human Pose (CVPR 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.
