Research map / 3D vision / Face and shape
HairNet: Single-View Hair Reconstruction using Convolutional Neural Networks
Yi Zhou, Liwen Hu, Jun Xing, Weikai Chen, Han-Wei Kung, Xin Tong, Hao Li
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 visibility-adaptive weights to the loss (ablation) | HairNet · Synthetic test set (100 hair models x 4 views) | visible-part position error: 0.024 -> 0.017 | Improved |
| add visibility-adaptive weights to the loss (ablation) | HairNet · Synthetic test set (100 hair models x 4 views) | collision error: 3.5x10^-7 -> 2.26x10^-7 | Improved |
| add visibility-adaptive weights to the loss (ablation) | HairNet · Synthetic test set (100 hair models x 4 views) | invisible-part position error: 0.026 -> 0.027 | No clear effect |
| add collision loss (ablation) | HairNet · Synthetic test set (100 hair models x 4 views) | collision error: 3.26x10^-6 -> 2.26x10^-7 | Improved |
| add collision loss (ablation) | HairNet · Synthetic test set (100 hair models x 4 views) | visible-part position error: 0.019 -> 0.017 | Improved |
| add curvature loss (ablation) | HairNet · Synthetic test set (100 hair models x 4 views) | visible-part position error: 0.020 -> 0.017 | Improved |
| add curvature loss (ablation) | HairNet · Synthetic test set (100 hair models x 4 views) | invisible-part position error: 0.029 -> 0.027 | Improved |
| replace Chai et al. retrieval-and-refinement with HairNet | Chai et al. · Synthetic test set (100 hair models x 4 views) | invisible-part position error: 0.040 -> 0.027 | Improved |
| replace Chai et al. retrieval-and-refinement with HairNet | Chai et al. · Time and space at 9K strands | total time: 53 s -> 0.04 - 0.08 s | Improved |
| replace Chai et al. database with HairNet network | Chai et al. · Time and space at 9K strands | total storage: 1 TiB -> 70 MiB | Improved |
| replace nearest-neighbor retrieval by HairNet | NN · Synthetic test set (100 hair models x 4 views) | visible-part position error: 0.033 -> 0.017 | Improved |
| replace nearest-neighbor retrieval by HairNet | NN · Synthetic test set (100 hair models x 4 views) | collision error: 0 -> 2.26x10^-7 | Hurt |
Closest papers by meaning
- Doubly Hierarchical Geometric Representations for Strand-based Human Hairstyle Generation (NeurIPS 2024 · similarity 0.92)
- DiffLocks: Generating 3D Hair from a Single Image using Diffusion Models (CVPR 2025 · similarity 0.90)
- Perm: A Parametric Representation for Multi-Style 3D Hair Modeling (ICLR 2025 · similarity 0.90)
- Text-Conditioned Generative Model of 3D Strand-based Human Hairstyles (CVPR 2024 · similarity 0.89)
- HairFastGAN: Realistic and Robust Hair Transfer with a Fast Encoder-Based Approach (NeurIPS 2024 · similarity 0.89)
- MonoHair: High-Fidelity Hair Modeling from a Monocular Video (CVPR 2024 · similarity 0.89)
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.
