Research map / 3D vision / Human pose estimation
PosePAL: Efficient Animal Pose Labeling Using Point Trackers
Zhuoyang Pan, Boxiao Pan, Guandao Yang, Adam W. Harley, Leonidas Guibas
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 |
|---|---|---|---|
| fine-tune only the query-point embedding instead of the feature extractor FNet (ablation) | CoTracker3 · DAVIS-Animals | delta_avg: 54.08 -> 67.53 | Improved |
| fine-tune only the query-point embedding instead of all components (ablation) | CoTracker3 · DAVIS-Animals | delta_avg: 66.26 -> 67.53 | Improved |
| fine-tune only the query-point embedding instead of all components (ablation) | CoTracker3 · DAVIS-Animals | fine-tuning time (min): 10.1 -> 3.1 | Improved |
| fine-tune query embedding with supervision instead of using CoTracker3 out of the box | CoTracker3 · DAVIS-Animals | delta_avg: 49.59 -> 67.53 | Improved |
| fine-tune query embedding with supervision instead of using CoTracker3 out of the box | CoTracker3 · DeepFly3D | delta_avg: 50.26 -> 70.80 | Improved |
| annotate keypoints evenly across frames instead of all keypoints in the last frame (ablation) | Ours · DeepFly3D | delta_avg: 44.86 -> 54.39 | Improved |
| annotate keypoints evenly across frames instead of randomly (ablation) | Ours · DeepFly3D | delta_avg: 44.61 -> 54.39 | Improved |
| supervise with manual annotations instead of ViTPose++-H pseudo-labels (ablation) | Ours · DAVIS-Animals | delta_avg: 51.17 -> 67.53 | Improved |
| add Kalman filter post-processing (ablation) | Ours · DAVIS-Animals | delta_avg: 67.53 -> 67.82 | Improved |
| add Kalman filter post-processing (ablation) | Ours · DAVIS-Animals | jitter J: 7.04 -> 6.97 | Improved |
| add Kalman filter post-processing (ablation) | SuperAnimal +sup. · DAVIS-Animals | jitter J: 10.36 -> 7.57 | Improved |
| add Kalman filter post-processing (ablation) | DINO-Tracker +sup. · DAVIS-Animals | delta_avg: 63.33 -> 61.91 | Hurt |
Closest papers by meaning
- APT-36K: A Large-scale Benchmark for Animal Pose Estimation and Tracking (NeurIPS 2022 · similarity 0.91)
- ScarceNet: Animal Pose Estimation With Scarce Annotations (CVPR 2023 · similarity 0.90)
- Deep Graph Pose: a semi-supervised deep graphical model for improved animal pose tracking (NeurIPS 2020 · similarity 0.89)
- Animal Pose Tracking: 3D Multimodal Dataset and Token-based Pose Optimization (IJCV 2022 · similarity 0.89)
- AP-10K: A Benchmark for Animal Pose Estimation in the Wild (NeurIPS 2021 · similarity 0.89)
- Pose Splatter: A 3D Gaussian Splatting Model for Quantifying Animal Pose and Appearance (NeurIPS 2025 · 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.
