Research map / 3D vision / Human pose estimation
3D-MuPPET: 3D Multi-Pigeon Pose Estimation and Tracking
Urs Waldmann, Alex Hoi Hang Chan, Hemal Naik, Máté Nagy, Iain D. Couzin, Oliver Deußen, Bastian Goldlüecke, Fumihiro Kano
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 ViTPose* instead of KeypointRCNN as 2D pose module | 3D-MuPPET · 3D-POP 2D | PCK05 (%): 82.4 -> 91.1 | Improved |
| use ViTPose* instead of KeypointRCNN as 2D pose module | 3D-MuPPET · 3D-POP 2D | median error (px): 5.7 -> 4.4 | Improved |
| use ViTPose* instead of KeypointRCNN as 2D pose module | 3D-MuPPET · 3D-POP 2D | mean speed (fps): 7.5 -> 2.1 | Hurt |
| use 3D-ViTPose* instead of 3D-KP-RCNN | 3D-MuPPET · 3D-POP 3D | PCK05 (%): 53.2 -> 71.0 | Improved |
| use 3D-ViTPose* instead of 3D-KP-RCNN | 3D-MuPPET · 3D-POP 3D | mean speed (fps): 1.76 -> 0.51 | Hurt |
| use DLC* instead of KeypointRCNN | 3D-MuPPET · 3D-POP 3D | PCK10 (%): 85.4 -> 90.9 | Improved |
| compare 3D-MuPPET (3D-ViTPose*) against LToHP, which uses 3D ground truth | 3D-ViTPose* · 3D-POP 3D | RMSE (mm): 24.0 -> 14.8 | Hurt |
| compare 3D-MuPPET (3D-ViTPose*) against LToHP | 3D-ViTPose* · 3D-POP 3D | mean speed (fps): 0.38 -> 0.51 | Improved |
| train on single pigeon, test on 5 pigeons instead of 1 (ablation) | KeypointRCNN · 3D-POP 2D | PCK10 (%): 98.7 -> 83.6 | Hurt |
| train on single pigeon, test on 10 pigeons instead of 1 (ablation) | KeypointRCNN · 3D-POP 3D | median error (mm): 6.9 -> 246.7 | Hurt |
| fine-tune DLC from indoor weights on wild data instead of indoor-only training (ablation) | Wild-DLC · Wild-MuPPET 3D | PCK10 (%): 74.4 -> 81.3 | Improved |
| train DLC from scratch on wild data instead of fine-tuning (ablation) | DLC-Fine-tuned · Wild-MuPPET 3D | median error (mm): 11.4 -> 12.7 | No clear effect |
Closest papers by meaning
- 3D-POP - An Automated Annotation Approach to Facilitate Markerless 2D-3D Tracking of Freely Moving Birds With Marker-Based Motion Capture (CVPR 2023 · similarity 0.91)
- 3D Bird Reconstruction: a Dataset, Model, and Shape Recovery from a Single View (ECCV 2020 · similarity 0.89)
- Animal3D: A Comprehensive Dataset of 3D Animal Pose and Shape (ICCV 2023 · similarity 0.89)
- Pose Splatter: A 3D Gaussian Splatting Model for Quantifying Animal Pose and Appearance (NeurIPS 2025 · similarity 0.88)
- BEAST3D: animal behavioral analysis and neural encoding from multi-view video via Gaussian splatting (NeurIPS 2026 · similarity 0.87)
- SAM 3D Animal: Promptable Animal 3D Reconstruction from Images in the Wild (NeurIPS 2026 · 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.
