Research map / Label-efficient and robust learning / Recognition and long-tail
PetFace: A Large-Scale Dataset and Benchmark for Animal Identification
Risa Shinoda, Kaede Shiohara
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 |
|---|---|---|---|
| use ArcFace instead of softmax or center loss for animal face re-identification (ablation) | ResNet-50 on PetFace · PetFace re-identification (avg) | top-1 accuracy: 41.88 -> 51.23 | Improved |
| use center loss added to softmax (ablation) | ResNet-50 on PetFace · PetFace re-identification | top-1 accuracy: 41.88 -> 9.81 | Hurt |
| joint-train ArcFace on all families instead of per family (ablation) | ArcFace on PetFace · PetFace re-identification | top-1 accuracy: Cat 54.29 -> 70.30, Pig 29.08 -> 41.49, Chimp 43.27 -> 34.30, Parakeet 62.19 -> 44.78 | Mixed |
| joint-train ArcFace on all families instead of per family (ablation) | ArcFace on PetFace · PetFace unseen verification (avg) | AUC: 91.30 -> 92.17 | Improved |
| use triplet loss instead of ArcFace (ablation) | ResNet-50 on PetFace · PetFace unseen verification (avg) | AUC: 91.30 -> 83.48 | Hurt |
| use ResNet-50 instead of ViT-B/32 or Swin-B backbone (ablation) | ArcFace on PetFace · PetFace unseen verification (avg) | AUC: 83.35 / 83.17 -> 91.30 | Improved |
| train on PetFace instead of other datasets (ImageNet, CLIP, MegaDescriptor) | face verification model · PetFace unseen verification (avg) | AUC: 83.70 -> 92.17 | Improved |
| train on PetFace Dog instead of DogFaceNet | dog face verification · Flickr-dog test | AUC: 85.51 -> 95.54 | Improved |
| train on PetFace Chimp instead of CZoo or CTai | chimp face verification · CTai test | AUC: 66.76 -> 67.33 | Improved |
| restrict verification pairs to same dog breed | MegaDescriptor · dog verification | AUC: 93.75 -> 86.25 | Hurt |
| restrict verification pairs to same dog breed | PetFace-trained model · dog verification | AUC: 99.01 -> 98.39 | No clear effect |
Closest papers by meaning
- AnimalWeb: A Large-Scale Hierarchical Dataset of Annotated Animal Faces (CVPR 2020 · similarity 0.94)
- AnimalClue: Recognizing Animals by their Traces (ICCV 2025 · similarity 0.88)
- MammalNet: A Large-Scale Video Benchmark for Mammal Recognition and Behavior Understanding (CVPR 2023 · similarity 0.88)
- On Combining Animal Re-Identification Models to Address Small Datasets (IJCV 2026 · similarity 0.88)
- Generating the Wild: Individual-Consistent Image-to-Video Generation for Wildlife (NeurIPS 2026 · similarity 0.85)
- Animal-Bench: Benchmarking Multimodal Video Models for Animal-centric Video Understanding (NeurIPS 2024 · 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.
