Research map / 3D vision / Face and shape
PIFuHD: Multi-Level Pixel-Aligned Implicit Function for High-Resolution 3D Human Digitization
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 a high-resolution fine PIFu level conditioned on coarse 3D features | Single PIFu · RenderPeople | Chamfer (cm): 1.47 -> 1.43 | Helped |
| Add predicted front and back normal maps as input guidance (ablation) | Multi-level PIFu · RenderPeople | P2S (cm): 1.41 -> 1.37 | Helped |
| Add predicted front and back normal maps as input guidance (ablation) | Multi-level PIFu · BUFF | P2S (cm): 1.63 -> 1.63 | No clear effect |
| Train coarse and fine modules alternately instead of jointly end-to-end (ablation) | Multi-level PIFu · RenderPeople | P2S (cm): 1.66 -> 1.41 | Helped |
| Train coarse and fine modules alternately instead of jointly end-to-end (ablation) | Multi-level PIFu · BUFF | Chamfer (cm): 1.81 -> 1.73 | Helped |
| Use only high-resolution fine features, without coarse 3D context (ablation) | Multi-level PIFu · RenderPeople | P2S (cm): 1.37 -> 4.15 | Hurt |
| Add a global image feature to a fine-only model (ablation) | Fine module only · RenderPeople | P2S (cm): 4.15 -> 2.92 | Helped |
| Replace global image feature conditioning with coarse pixel-aligned 3D context (ablation) | Fine module + Global image feature · RenderPeople | P2S (cm): 2.92 -> 1.37 | Helped |
| Use high-resolution fine features without holistic context (ablation) | Fine-level PIFu · RenderPeople and BUFF | P2S (cm) | Hurt |
Closest papers by meaning
- ANIM: Accurate Neural Implicit Model for Human Reconstruction from a single RGB-D Image (CVPR 2024 · similarity 0.87)
- Fine Structure-Aware Sampling: A New Sampling Training Scheme for Pixel-Aligned Implicit Models in Single-View Human Reconstruction (AAAI 2024 · similarity 0.87)
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 2024 · similarity 0.87)
- SAT-HMR: Real-Time Multi-Person 3D Mesh Estimation via Scale-Adaptive Tokens (CVPR 2025 · similarity 0.86)
- TokenHMR: Advancing Human Mesh Recovery with a Tokenized Pose Representation (CVPR 2024 · similarity 0.86)
- MultiGO: Towards Multi-level Geometry Learning for Monocular 3D Textured Human Reconstruction (CVPR 2025 · 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.