Research map / 3D vision / Face and shape
Distilling Neural Fields for Real-Time Articulated Shape Reconstruction
Read the paperCheck 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 |
|---|---|---|---|
| distill dynamic NeRF supervision into a feed-forward predictor | DASR · AMA human sequences | inference time per frame: 43000 -> 22 | Helped |
| add a temporal encoder (ablation) | DASR · AMA T_samba | CD (cm): 12.2 -> 10.6 | Helped |
| use a Conv1D temporal encoder instead of the full method's temporal encoder (ablation) | DASR · AMA sequences | CD (cm): 10.6 -> 10.8; 17.3 -> 15.0 | Mixed |
| use a Transformer temporal encoder instead of the full method's temporal encoder (ablation) | DASR · AMA sequences | CD (cm): 10.6 -> 10.6; 12.0 -> 13.2 | Mixed |
| remove frozen teacher decoders (ablation) | DASR · AMA sequences | CD (cm): 10.6 -> 11.0; 12.0 -> 11.5 | Mixed |
| extract the template using a 64³ grid instead of a 128³ grid (ablation) | DASR · AMA T_samba | F@2%: 49.6 -> 42.2 | Hurt |
| extract the template using a 256³ grid instead of a 128³ grid (ablation) | DASR · AMA T_samba | F@2%: 49.6 -> 50.7 | No clear effect |
| train with 5 videos instead of 35 (ablation) | DASR · AMA T_samba | CD (cm): 10.8 -> 11.2 | Hurt |
| train with 16 videos instead of 35 (ablation) | DASR · AMA T_samba | CD (cm): 10.8 -> 10.5 | Helped |
| train with 26 videos instead of 35 (ablation) | DASR · AMA T_samba | CD (cm): 10.8 -> 10.2 | Helped |
| train with 35 videos rather than 26, adding unrelated identities (ablation) | DASR · AMA T_samba | CD (cm): 10.2 -> 10.8 | Hurt |
Closest papers by meaning
- REACTO: Reconstructing Articulated Objects from a Single Video (CVPR 2024 · similarity 0.90)
- Learning Implicit Representation for Reconstructing Articulated Objects (ICLR 2024 · similarity 0.89)
- TFS-NeRF: Template-Free NeRF for Semantic 3D Reconstruction of Dynamic Scene (NeurIPS 2024 · similarity 0.89)
- Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses (NeurIPS 2024 · similarity 0.87)
- RigGS: Rigging of 3D Gaussians for Modeling Articulated Objects in Videos (CVPR 2025 · similarity 0.87)
- SAOR: Single-View Articulated Object Reconstruction (CVPR 2024 · 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.