Research map / Detection and segmentation / Weakly supervised detection and segmentation
Dynamic Curriculum Learning for Great Ape Detection in the Wild
Xinyu Yang, Tilo Burghardt, Majid Mirmehdi
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 linear-increase unlabelled sample policy instead of constant use of all unlabelled data (ablation) | Dynamic curriculum detector · Extended PanAfrican (PLD 50% labelled, fold1, ablation) | mAP: 60.46 -> 61.62 | Improved |
| add warm-up and cool-down phases to linear increase sample policy (ablation) | Dynamic curriculum detector · Extended PanAfrican (PLD 50% labelled, fold1, ablation) | mAP: 61.62 -> 64.73 | Improved |
| use linear decrease of unlabelled samples instead of increase (ablation) | Dynamic curriculum detector · Extended PanAfrican (PLD 50% labelled, fold1, ablation) | mAP: 64.73 -> 53.47 | Hurt |
| use dynamic linear 0.1->1 loss weight instead of constant 0.5 (ablation) | Dynamic curriculum detector · Extended PanAfrican (PLD 50% labelled, fold1, ablation) | mAP: 63.98 -> 64.73 | Improved |
| use constant loss weight 2.0 instead of 0.5 (ablation) | Dynamic curriculum detector · Extended PanAfrican (PLD 50% labelled, fold1, ablation) | mAP: 63.98 -> 54.68 | Hurt |
| add arctan increasing confidence threshold 0.1->0.6 instead of constant 0.6 (ablation) | Dynamic curriculum detector · Extended PanAfrican (PLD 50% labelled, fold1, ablation) | mAP: 62.62 -> 64.73 | Improved |
| use very low constant threshold 0.05 instead of 0.6 (ablation) | Dynamic curriculum detector · Extended PanAfrican (PLD 50% labelled, fold1, ablation) | mAP: 62.62 -> 51.73 | Hurt |
| add weak and strong augmentation policy (ablation) | Dynamic curriculum detector · Extended PanAfrican (PLD 50% labelled, fold1, ablation) | mAP: 55.41 -> 64.73 | Improved |
| add dynamic cosine teacher momentum 0.998->0.9998 instead of constant 0.999 (ablation) | Dynamic curriculum detector · Extended PanAfrican (PLD 50% labelled, fold1, ablation) | mAP: 60.07 -> 64.73 | Improved |
| use SWAV self-supervised init instead of random init (ablation) | Dynamic curriculum detector · Extended PanAfrican (PLD 50% labelled, fold1, ablation) | mAP: 0.17 -> 64.73 | Improved |
| use SWAV self-supervised init instead of supervised ImageNet init (ablation) | Dynamic curriculum detector · Extended PanAfrican (PLD 50% labelled, fold1, ablation) | mAP: 61.55 -> 64.73 | Improved |
| apply full method vs UbTeacher at 10% labelled | Dynamic curriculum detector · Extended PanAfrican (PLD 10% labelled) | mAP: 44.03 -> 45.96 | Improved |
Closest papers by meaning
- PanAf20K: A Large Video Dataset for Wild Ape Detection and Behaviour Recognition (IJCV 2024 · similarity 0.90)
- Focus on the Positives: Self-Supervised Learning for Biodiversity Monitoring (ICCV 2021 · similarity 0.87)
- From Forest to Zoo: Great Ape Behavior Recognition with ChimpBehave (IJCV 2025 · similarity 0.86)
- The PanAf-FGBG Dataset: Understanding the Impact of Backgrounds in Wildlife Behaviour Recognition (CVPR 2025 · similarity 0.85)
- MammAlps: A Multi-view Video Behavior Monitoring Dataset of Wild Mammals in the Swiss Alps (CVPR 2025 · similarity 0.85)
- Generating the Wild: Individual-Consistent Image-to-Video Generation for Wildlife (NeurIPS 2026 · similarity 0.84)
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.
