Research map / 3D vision / Gaussian splatting
CGHair: Compact Gaussian Hair Reconstruction with Card Clustering
Haimin Luo, Srinjay Sarkar, Albert Mosella-Montoro, Francisco Vicente Carrasco, Fernando De la Torre
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 card-based clustering of appearance textures (ablation) | CGHair · HairSalon synthetic | PSNR: 28.00 -> 32.40 | Improved |
| add card-based clustering of appearance textures (ablation) | CGHair · HairSalon synthetic | appearance size (MB): 22.42 -> 0.71 | Improved |
| use shared latent features with a decoder instead of Multi-SH (ablation) | CGHair · HairSalon synthetic | appearance size (MB): 11.95 -> 0.71 | Improved |
| use shared latent features with a decoder instead of Multi-SH (ablation) | CGHair · HairSalon synthetic | PSNR: 33.50 -> 32.40 | Hurt |
| use latent features with MLP decoder instead of Single-SH (ablation) | CGHair · HairSalon synthetic | PSNR: 30.48 -> 32.40 | Improved |
| share appearance codebooks instead of unique per-strand SH | CGHair · HairSalon synthetic | compression ratio: 1.0 -> 230.6 | Improved |
| enlarge codebook size K from 10 to 90 (ablation) | CGHair · HairSalon synthetic | PSNR: 32.40 -> 32.10 | No clear effect |
| raise latent dimension D from 64 to 256 (ablation) | CGHair · HairSalon synthetic | PSNR: 32.40 -> 31.94 | Hurt |
| lower latent dimension D from 64 to 32 (ablation) | CGHair · HairSalon synthetic | PSNR: 32.40 -> 32.39 | No clear effect |
| increase number of texture clusters N_T from 16 to 96 (ablation) | CGHair · HairSalon synthetic | PSNR: 31.49 -> 32.57 | Improved |
| use one texture per hair card (400) instead of 64 clusters (ablation) | CGHair · HairSalon synthetic | PSNR: 32.40 -> 33.04 | Improved |
| add full strand generator (PERM prior with all modules) over only D_PCA (ablation) | CGHair · HairSalon synthetic strands | positional error: 0.162 -> 0.148 | Improved |
Closest papers by meaning
- Dr.Hair: Reconstructing Scalp-Connected Hair Strands without Pre-Training via Differentiable Rendering of Line Segments (CVPR 2024 · similarity 0.89)
- MonoHair: High-Fidelity Hair Modeling from a Monocular Video (CVPR 2024 · similarity 0.89)
- DiffLocks: Generating 3D Hair from a Single Image using Diffusion Models (CVPR 2025 · similarity 0.88)
- Perm: A Parametric Representation for Multi-Style 3D Hair Modeling (ICLR 2025 · similarity 0.88)
- Doubly Hierarchical Geometric Representations for Strand-based Human Hairstyle Generation (NeurIPS 2024 · similarity 0.88)
- DGH: Dynamic Gaussian Hair (NeurIPS 2025 · 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.
