Research map / 3D vision / Face and shape
OmniFit: Multi-modal 3D Body Fitting via Scale-agnostic Dense Landmark Prediction
Zeyu CAI, Yuliang Xiu, renke wang, Zhijing Shao, Xiaoben Li, YU Siyuan, Chao Xu, Yang Liu, Baigui Sun, Jian Yang, zhenyu zhang
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 scale predictor for scale-normalized inputs (ablation) | OmniFit · 4D-DRESS (scale-normalized) | V2V (cm): 1.120 -> 0.651 | Improved |
| add scale predictor for scale-normalized inputs (ablation) | OmniFit · 4D-DRESS (scale-normalized) | MPJPE (cm): 1.172 -> 0.709 | Improved |
| add image adapter with front-view RGB (ablation) | OmniFit · 4D-DRESS | V2V (cm): 0.589 -> 0.539 | Improved |
| add image adapter with front-view RGB (ablation) | OmniFit · 4D-DRESS | MPJPE (cm): 0.662 -> 0.622 | Improved |
| use 180/120/300 landmark split instead of 120/120/360 hands/head/body (ablation) | OmniFit · 4D-DRESS | V2V (cm): 0.637 -> 0.589 | Improved |
| use 180/120/300 landmark split instead of 300/120/180 (ablation) | OmniFit · 4D-DRESS | V2V (cm): 0.655 -> 0.589 | Improved |
| increase landmarks from 100 to 600 (ablation) | OmniFit · 4D-DRESS | V2V (cm): 0.637 -> 0.589 | Improved |
| increase landmarks from 100 to 600 (ablation) | OmniFit · 4D-DRESS | GFLOPs: 23.82 -> 93.38 | Hurt |
| increase input points from 5000 to 15000 (ablation) | OmniFit · 4D-DRESS | V2V (cm): 0.589 -> 0.452 | Improved |
| use DINOv2 instead of Sapiens as image encoder (ablation) | OmniFit adapter · 4D-DRESS | V2V (cm): 0.589 -> 0.539 | Improved |
| train on unified dataset (Ours*) instead of 4D-DRESS only | OmniFit · THuman2.1 | V2V (cm): 1.768 -> 1.338 | Improved |
| train adapter on unified dataset (Adapter*) | OmniFit adapter · THuman2.1 | V2V (cm): 1.767 -> 0.835 | Improved |
Closest papers by meaning
- SHARP: Shape-Aware Reconstruction of People in Loose Clothing (IJCV 2022 · similarity 0.88)
- ETCH: Generalizing Body Fitting to Clothed Humans via Equivariant Tightness (ICCV 2025 · similarity 0.88)
- PHD: Personalized 3D Human Body Fitting with Point Diffusion (ICCV 2025 · similarity 0.88)
- ETCH-X: Robustify Expressive Body Fitting to Clothed Humans with Composable Synthetic Data (ECCV 2026 · similarity 0.87)
- Unite the People: Closing the Loop Between 3D and 2D Human Representations (CVPR 2017 · similarity 0.87)
- 3DPeople: Modeling the Geometry of Dressed Humans (ICCV 2019 · similarity 0.87)
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.
