Research map / 3D vision / Face and shape
BANMo: Building Animatable 3D Neural Models From Many Casual Videos
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 |
|---|---|---|---|
| Use BANMo instead of ViSER | 3D reconstruction · AMA-swing | Chamfer distance (cm): 15.7 -> 9.1 | Helped |
| Use BANMo instead of ViSER | 3D reconstruction · Eagle | F@2% (%): 20.6 -> 56.7 | Helped |
| Use BANMo instead of ViSER | 3D reconstruction · Hands | Chamfer distance (cm): 16.8 -> 7.5 | Helped |
| Add canonical embeddings and their registration losses (ablation) | BANMo · Multiple videos | registration quality | Helped |
| Add measured optical-flow reconstruction loss (ablation) | BANMo · Multiple videos | registration quality | Helped |
| Use neural blend skinning instead of MLP-SE(3) (ablation) | BANMo · Eagle | deformation regularity | Helped |
| Use neural blend skinning instead of MLP-translation (ablation) | BANMo · Eagle | geometric registration | Helped |
| Use 16 bones instead of 25 (ablation) | BANMo · AMA swing and samba | Chamfer distance (cm): 9.08 -> 9.02 | Helped |
| Add the in-surface loss (ablation) | BANMo · AMA swing and samba | F@2% (%): 54.8 -> 57.0 | Helped |
| Use embeddings pre-trained for quadruped animals on humans (ablation) | BANMo · AMA swing and samba | F@2% (%): 57.0 -> 54.2 | Hurt |
| Increase initial pose error to 90° (ablation) | BANMo · AMA swing and samba | Chamfer distance (cm): 9.08 -> 9.91 | Hurt |
| Increase view coverage from 90° to 180° (ablation) | BANMo · AMA swing and samba | F@2% (%): 54.3 -> 59.8 | Helped |
Closest papers by meaning
- Distilling Neural Fields for Real-Time Articulated Shape Reconstruction (CVPR 2023 · similarity 0.93)
- REACTO: Reconstructing Articulated Objects from a Single Video (CVPR 2024 · similarity 0.92)
- A-NeRF: Articulated Neural Radiance Fields for Learning Human Shape, Appearance, and Pose (NeurIPS 2021 · similarity 0.90)
- Learning Implicit Representation for Reconstructing Articulated Objects (ICLR 2024 · similarity 0.90)
- SAOR: Single-View Articulated Object Reconstruction (CVPR 2024 · similarity 0.89)
- S3O: A Dual-Phase Approach for Reconstructing Dynamic Shape and Skeleton of Articulated Objects from Single Monocular Video (ICML 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.