Research map / 3D vision / Face and shape
Hierarchically Structured Neural Bones for Reconstructing Animatable Objects from Casual Videos
Subin Jeon, In Cho, Minsu Kim, Woong Oh Cho, Seon Joo Kim
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 |
|---|---|---|---|
| use the hierarchy (3 depths, 24 bones) instead of 24 bones in a single depth (ablation) | hierarchical neural bones · Samba reconstruction | Chamfer distance: 7.16 -> 6.15 | Improved |
| use the hierarchy instead of a single depth with 24 bones (ablation) | hierarchical neural bones · Swing reconstruction | Chamfer distance: 8.83 -> 7.11 | Improved |
| use bone occupancy regularization instead of Sinkhorn divergence (single depth, 24 bones) (ablation) | hierarchical neural bones · Samba reconstruction | Chamfer distance: 7.17 -> 7.16 | Improved |
| add bone regularization on a single-depth model (ablation) | hierarchical neural bones · Samba reconstruction (1 depth, 24 bones) | Chamfer distance: 6.84 -> 7.16 | Hurt |
| use hierarchical bones instead of BANMo with fewer videos (4 vs 8) | Samba reconstruction · Samba, number of input videos | Chamfer distance: BANMo 8 vids 7.22 vs Ours 4 vids 7.05 | Improved |
| use hierarchical bones instead of skeleton-based RAC on Swing | reconstruction · Swing reconstruction | F-score F2: 70.33 -> 65.88 | Hurt |
| use hierarchical bones instead of BANMo for retargeting at 50 steps | reanimation · Samba retargeting | Chamfer distance: 2.75 -> 2.15 | Improved |
| use tree-structured bones for manipulation instead of unstructured BANMo bones | manipulation · User study (12 participants), average | time to reach target pose: 5m 5s -> 3m 50s | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BANMo — BANMo: Building Animatable 3D Neural Models From Many Casual Videos (CVPR 2022) · better on 8 of 8 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- BANMo: Building Animatable 3D Neural Models From Many Casual Videos (CVPR 2022 · similarity 0.90)
- REACTO: Reconstructing Articulated Objects from a Single Video (CVPR 2024 · similarity 0.89)
- CASA: Category-agnostic Skeletal Animal Reconstruction (NeurIPS 2022 · similarity 0.89)
- NPC: Neural Point Characters from Video (ICCV 2023 · similarity 0.89)
- Animatable Implicit Neural Representations for Creating Realistic Avatars From Videos (TPAMI 2024 · similarity 0.88)
- Learning Implicit Representation for Reconstructing Articulated Objects (ICLR 2024 · similarity 0.88)
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.
