Research map / 3D vision / Human pose estimation
APT-36K: A Large-scale Benchmark for Animal Pose Estimation and Tracking
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 |
|---|---|---|---|
| Pre-train on MS COCO human-pose annotations instead of ImageNet-1K (ablation) | SimpleBaseline with ResNet-101 on APT-36K · APT-36K val, SF track | AP: 69.6 -> 73.5 | Helped |
| Pre-train on AP-10K animal-pose annotations instead of ImageNet-1K (ablation) | HRNet-w48 on APT-36K · APT-36K val, SF track | AP: +2.3 AP | Helped |
| Use MS COCO human-pose pre-training rather than AP-10K animal-pose pre-training (ablation) | CNN-based pose estimators on APT-36K · APT-36K val, SF track | AP: 2.2 vs 1.7 | Helped |
| Use MS COCO human-pose pre-training instead of ImageNet-1K pre-training (ablation) | ViTPose on APT-36K · APT-36K val, SF track | AP: 77.4 -> 78.3 | Helped |
| Use AP-10K animal-pose pre-training instead of ImageNet-1K pre-training (ablation) | ViTPose on APT-36K · APT-36K val, SF track | AP: 77.4 -> 78.2 | Helped |
| Use ViTTrack instead of SiamRPN++ for tracking | ViTPose on APT-36K · APT-36K test, APT track | AP: 74.2 -> 75.8 | Helped |
| Initialize and freeze the tracker encoder using the APT-36K-trained ViTPose encoder (ablation) | ViTTrack · APT-36K test, APT track with ViTPose | AP: 75.8 -> 75.5 | Hurt |
| Train without instances from the test animal family (ablation) | HRNet-w32 · APT-36K IS track, Cercopithecidae | AP: 29.6 vs over 76.0 | Hurt |
| Add 40 labeled examples from the target family rather than use zero-shot transfer (ablation) | HRNet-w32 · APT-36K few-shot, Cercopithecidae | AP: 35.6 -> 42.1 | Helped |
| Use optical-flow-based tracking rather than single-object tracking | HRNet-w32 pose estimator · APT-36K test, APT track | AP: 71.3 vs at least 73.0 | Hurt |
Closest papers by meaning
- Animal-Bench: Benchmarking Multimodal Video Models for Animal-centric Video Understanding (NeurIPS 2024 · similarity 0.89)
- Web-Scale Collection of Video Data for 4D Animal Reconstruction (NeurIPS 2025 · similarity 0.89)
- Pose Splatter: A 3D Gaussian Splatting Model for Quantifying Animal Pose and Appearance (NeurIPS 2025 · similarity 0.87)
- BuckTales: A multi-UAV dataset for multi-object tracking and re-identification of wild antelopes (NeurIPS 2024 · similarity 0.87)
- Probabilistic Prompt Distribution Learning for Animal Pose Estimation (CVPR 2025 · similarity 0.87)
- MammAlps: A Multi-view Video Behavior Monitoring Dataset of Wild Mammals in the Swiss Alps (CVPR 2025 · 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.