Research map / 3D vision / Depth and stereo
A Neural Network for Detailed Human Depth Estimation From a Single Image
Sicong Tang, Feitong Tan, Kelvin Cheng, Zhaoyang Li, Siyu Zhu, Ping Tan
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 skeleton and segmentation cues (ablation) | Depth-Net (Base + Detail) · Human depth test set (fused GT) | accuracy 1.25cm: 25.74 -> 29.24 | Improved |
| add skeleton and segmentation cues (ablation) | Depth-Net (Base + Detail) · Human depth test set (fused GT) | MAE: 4.382 -> 3.282 | Improved |
| use two-branch base/detail separation (ablation) | Depth-Net (Base + Detail) · Human depth test set (fused GT) | accuracy 5.0cm: 72.97 -> 75.52 | Improved |
| use two-branch base/detail separation (ablation) | Depth-Net (Base + Detail) · Human depth test set (fused GT) | MAE: 3.480 -> 3.282 | Improved |
| add stage 2 joint training after stage 1 (ablation) | Depth-Net (Base + Detail) · Human depth test set (fused GT) | accuracy 1.25cm: 26.64 -> 29.24 | Improved |
| add stage 2 joint training after stage 1 (ablation) | Depth-Net (Base + Detail) · Human depth test set (fused GT) | MAE: 3.592 -> 3.282 | Improved |
| add stage 1 pre-training before stage 2 (ablation) | Depth-Net (Base + Detail) · Human depth test set (fused GT) | accuracy 1.25cm: 27.89 -> 29.24 | Improved |
| add stage 1 pre-training before stage 2 (ablation) | Depth-Net (Base + Detail) · Human depth test set (fused GT) | MAE: 3.332 -> 3.282 | Improved |
| use truncated L1 instead of Huber loss on composed shape (ablation) | Depth-Net (Base + Detail) · Human depth test set (fused GT) | accuracy 5.0cm: 74.23 -> 75.52 | Improved |
| use truncated L1 instead of Huber loss on composed shape (ablation) | Depth-Net (Base + Detail) · Human depth test set (fused GT) | MAE: 3.410 -> 3.282 | Improved |
| add iterative normal-based depth refinement (ablation) | Ours · Human depth test set (fused GT) | accuracy 1.25cm: 29.24 -> 30.06 | Improved |
| add iterative normal-based depth refinement (ablation) | Ours · Human depth test set (fused GT) | MAE: 3.282 -> 3.208 | Improved |
Closest papers by meaning
- Depth from a Single Image by Harmonizing Overcomplete Local Network Predictions (NeurIPS 2016 · similarity 0.88)
- Single Image Depth Estimation Trained via Depth From Defocus Cues (CVPR 2019 · similarity 0.88)
- Unsupervised CNN for Single View Depth Estimation: Geometry to the Rescue (ECCV 2016 · similarity 0.87)
- Surface Normals in the Wild (ICCV 2017 · similarity 0.87)
- Refining 3D Human Texture Estimation From a Single Image (TPAMI 2024 · similarity 0.87)
- ANIM: Accurate Neural Implicit Model for Human Reconstruction from a single RGB-D Image (CVPR 2024 · 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.
