Research map / 3D vision / Face and shape
Detailed Avatar Recovery From Single Image
Hao Zhu, Xinxin Zuo, Haotian Yang, Sen Wang, Xun Cao, Ruigang Yang
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 |
|---|---|---|---|
| crop windows around handles instead of full image as joint net input (ablation) | HMD joint stage · WILD | sil IoU: 68.3% -> 70.7% | Improved |
| crop windows around handles instead of full image as joint net input (ablation) | HMD joint stage · WILD | 2D joint error pixel: 9.85 -> 8.81 | Improved |
| crop windows instead of full image as anchor net input (ablation) | HMD anchor stage · WILD | sil IoU: 70.1% -> 71.3% | Improved |
| combine joint and anchor deformation instead of anchor only (ablation) | HMD · WILD | sil IoU: 71.3% -> 76.5% | Improved |
| add joint-level and anchor-level deformation to HMR initial shape (ablation) | HMD · WILD | sil IoU: 67.6% -> 76.5% | Improved |
| add photometric loss to ShadingNet (ablation) | ShadingNet · captured depth dataset | surface error mm: 1.346 -> 1.129 | Improved |
| add ground-truth silhouette as input | HMD anchor · WILD | sil IoU: 76.5 -> 82.4 | Improved |
| add random perturbation augmentation to texture training data (ablation) | texture synthesis · SURREAL | PSNR: 17.62 -> 19.65 | Improved |
| add flow network before refinement (ablation) | texture synthesis · SURREAL | PSNR: 19.48 -> 19.65 | Improved |
| add symmetry-flipped texture input to refinement (ablation) | texture synthesis · SURREAL | PSNR: 19.58 -> 19.65 | Improved |
| add vertex-level deformation with shading net after anchor stage (ablation) | HMD · SYN | 3D err full mm: 47.18 -> 44.75 | Improved |
| add vertex-level deformation after anchor stage (ablation) | HMD · RECON | 3D err visible mm: 39.73 -> 41.76 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- DeepHuman — DeepHuman: 3D Human Reconstruction From a Single Image (ICCV 2019) · better on 0 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
- Detailed Human Shape Estimation From a Single Image by Hierarchical Mesh Deformation (CVPR 2019 · similarity 0.95)
- End-to-End Recovery of Human Shape and Pose (CVPR 2018 · similarity 0.90)
- High-Fidelity Human Avatars From a Single RGB Camera (CVPR 2022 · similarity 0.89)
- FRESA: Feedforward Reconstruction of Personalized Skinned Avatars from Few Images (CVPR 2025 · similarity 0.88)
- Neural Re-Rendering of Humans from a Single Image (ECCV 2020 · similarity 0.88)
- Pose Modulated Avatars from Video (ICLR 2024 · similarity 0.88)
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.
