Research map / Label-efficient and robust learning / Recognition and long-tail
Automated Visual Fin Identification of Individual Great White Sharks
Benjamin Hughes, Tilo Burghardt
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 |
|---|---|---|---|
| combine DoG_N descriptors across multiple scales (ablation) | DoG_N LNBNN · Great white fin ID (1-shot) | AP: 0.72 -> 0.76 | Improved |
| use contour DoG_N encoding instead of SIFT texture features | LNBNN identification · Great white fin ID (2 training images) | mAP: 0.35 -> 0.83 | Improved |
| replace DoG_N with normal encoding | LNBNN identification · Great white fin ID (1-shot) | AP: 0.76 -> 0.72 | Hurt |
| add fin-space scoring vector with random forest | DoG_N and normal baselines · Great white fin ID (1-shot) | AP: 0.76 -> 0.81 | Improved |
| combine both descriptor types in fin space (ablation) | fin space random forest · Great white fin ID | precision | Improved |
| combine OpponentSIFT and normal features for candidate classification (ablation) | fin detection · Fin detection (high visibility), t=0.9 | AP_det: 0.73 -> 0.86 | Improved |
| combine OpponentSIFT and normal features (ablation) | fin detection · Fin detection (high visibility), t=0.85 | AP_det: 0.85 -> 0.95 | Improved |
| use only normal features instead of combined (ablation) | fin detection · Fin detection (high visibility), t=0.7 | AP_det: 0.98 -> 0.98 | No clear effect |
| use only 1 image per class vs adding second image (one per fin side) with scale combination | DoG_N combined · Great white fin ID | mAP: 0.79 -> 0.83 | Improved |
| restrict to high-quality fin detections | DoG_N combined · Great white fin ID (217 detections) | mAP: 0.79 -> 0.84 | Improved |
| use finest scale sigma=1 instead of combined (ablation) | DoG_N · Great white fin ID (1-shot) | AP: 0.76 -> 0.49 | Hurt |
| use coarsest scale sigma=8 instead of combined (ablation) | Norm · Great white fin ID (1-shot) | AP: 0.72 -> 0.33 | Hurt |
Closest papers by meaning
- Fantastic Animals and Where to Find Them: Segment Any Marine Animal with Dual SAM (CVPR 2024 · similarity 0.84)
- Fish-Vista: A Multi-Purpose Dataset for Understanding & Identification of Traits from Images (CVPR 2025 · similarity 0.84)
- A Kendall Shape Space Approach to 3D Shape Estimation from 2D Landmarks (ECCV 2022 · similarity 0.83)
- CHIRP dataset: towards long-term, individual-level, behavioral monitoring of bird populations in the wild (CVPR 2026 · similarity 0.83)
- Generating the Wild: Individual-Consistent Image-to-Video Generation for Wildlife (NeurIPS 2026 · similarity 0.83)
- RealBirdID: Benchmarking Bird Species Identification in the Era of MLLMs (CVPR 2026 · similarity 0.80)
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.
