Research map / Video / Video-language understanding
PanAf20K: A Large Video Dataset for Wild Ape Detection and Behaviour Recognition
Otto Brookes, Majid Mirmehdi, Colleen R. Stephens, Samuel Angedakin, Katherine Corogenes, Dervla Dowd, Paula Dieguez, Thurston Cleveland Hicks, Sorrel C. Z. Jones, Kevin C. Lee, Vera Leinert, Juan Lapuente and 15 more
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 |
|---|---|---|---|
| add logit adjustment loss | TimeSformer · PanAf20K multi-label | mAP: 47.24 -> 49.39 | Improved |
| add logit adjustment loss (tail classes) | TimeSformer · PanAf20K multi-label | tail mAP: 13.29 -> 18.20 | Improved |
| add focal loss | 3D ResNet-50 · PanAf20K multi-label | mAP: 46.03 -> 47.93 | Improved |
| add focal loss | MViTV2 · PanAf20K multi-label | mAP: 45.71 -> 45.78 | No clear effect |
| add weight balancing | X3D · PanAf20K multi-label | mAP: 46.06 -> 45.64 | Hurt |
| add weight balancing | MViTV2 · PanAf20K multi-label | mAP: 45.71 -> 45.36 | Hurt |
| add long-tail losses (focal, logit adjustment, weight balancing) | all models · PanAf20K tail classes | tail mAP | Improved |
| use in-domain camera-trap pretraining (MegaDetector) instead of COCO pretraining | detector · PanAf500 detection | mAP: 88.0 | Improved |
| increase behaviour threshold from t=16 to t=128 | X3D · PanAf500 action recognition | top-1: 80.00 -> 74.24 | Hurt |
| increase behaviour threshold from t=16 to t=128 | I3D · PanAf500 action recognition | top-1: 79.29 -> 67.45 | Hurt |
Closest papers by meaning
- The PanAf-FGBG Dataset: Understanding the Impact of Backgrounds in Wildlife Behaviour Recognition (CVPR 2025 · similarity 0.93)
- From Forest to Zoo: Great Ape Behavior Recognition with ChimpBehave (IJCV 2025 · similarity 0.91)
- MammAlps: A Multi-view Video Behavior Monitoring Dataset of Wild Mammals in the Swiss Alps (CVPR 2025 · similarity 0.88)
- BuckTales: A multi-UAV dataset for multi-object tracking and re-identification of wild antelopes (NeurIPS 2024 · similarity 0.87)
- Animal-Bench: Benchmarking Multimodal Video Models for Animal-centric Video Understanding (NeurIPS 2024 · similarity 0.86)
- WildBox: A Dataset and Benchmark for Aerial Monocular 3D Detection of African Savanna Wildlife (NeurIPS 2026 · similarity 0.85)
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.
