Abstract:Automated animal face identification plays a crucial role in the monitoring of behaviors, conducting of surveys, and finding of lost animals. Despite the advancements in human face identification, the lack of datasets and benchmarks in the animal domain has impeded progress. In this paper, we introduce the PetFace dataset, a comprehensive resource for animal face identification encompassing 257,484 unique individuals across 13 animal families and 319 breed categories, including both experimental and pet animals. This large-scale collection of individuals facilitates the investigation of unseen animal face verification, an area that has not been sufficiently explored in existing datasets due to the limited number of individuals. Moreover, PetFace also has fine-grained annotations such as sex, breed, color, and pattern. We provide multiple benchmarks including re-identification for seen individuals and verification for unseen individuals. The models trained on our dataset outperform those trained on prior datasets, even for detailed breed variations and unseen animal families. Our result also indicates that there is some room to improve the performance of integrated identification on multiple animal families. We hope the PetFace dataset will facilitate animal face identification and encourage the development of non-invasive animal automatic identification methods.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve several key problems in **Automated Animal Face Identification**: 1. **Lack of large - scale datasets and benchmark tests**: - Although significant progress has been made in the field of human face recognition, in the field of animal face recognition, research progress has been hindered due to the lack of large - scale datasets and benchmark tests. Existing public datasets usually contain a limited number of individuals (less than 100), which makes it difficult to train models with wide applicability and high recognition. 2. **Unexplored identification of unknown individuals**: - Existing datasets perform poorly when dealing with unseen individuals because the number of individuals in these datasets is limited and cannot fully evaluate the ability to recognize new individuals. Therefore, the verification of unknown individuals is an area that has not been fully explored. 3. **Recognition of multi - species and fine - grained attributes**: - Existing datasets usually focus on a single animal species and lack annotations for fine - grained attributes such as gender, breed, and color. This limits the comprehensive recognition research of multiple animal families and breeds. To solve these problems, the author introduced a large - scale animal face recognition dataset named **PetFace**. PetFace contains 257,484 unique individuals from 13 animal families and 319 breeds, with a total of more than 1 million images. In addition, PetFace also provides fine - grained annotation information, such as gender, breed, color, and pattern. Through this dataset, the author hopes to promote research on animal face recognition and encourage the development of non - invasive automatic animal identification methods, thereby reducing the dependence on traditional invasive identification means (such as ear tags, tattoos, etc.) and improving animal welfare. ### Main contributions - **Established a new large - scale animal face recognition dataset PetFace**, which contains 257,484 individuals, covering 13 animal families and 319 breeds, and provides fine - grained annotations. - **Set up two benchmark tests**: one is to identify known (seen) individuals, and the other is to identify unknown (unseen) individuals. - **Demonstrated the generalization ability of the model on unseen individuals and unseen animal categories**. Through these contributions, the PetFace dataset is expected to promote the development of animal face recognition technology and provide a solid foundation for future related research.

PetFace: A Large-Scale Dataset and Benchmark for Animal Identification

AnimalWeb: A Large-Scale Hierarchical Dataset of Annotated Animal Faces

Open-set Pets Facial Recognition Using Deep Learning and Statistical Learning

Multispecies Animal Re-ID Using a Large Community-Curated Dataset

CatFLW: Cat Facial Landmarks in the Wild Dataset

Automated Detection of Cat Facial Landmarks

OpenAnimals: Revisiting Person Re-Identification for Animals Towards Better Generalization

Distinguishing Individual Red Pandas from Their Faces

Primate Face Identification in the Wild

Animal Kingdom: A Large and Diverse Dataset for Animal Behavior Understanding

dopanim: A Dataset of Doppelganger Animals with Noisy Annotations from Multiple Humans

AP-10K: A Benchmark for Animal Pose Estimation in the Wild

Large-scale Datasets: Faces with Partial Occlusions and Pose Variations in the Wild

Pixel-Face: A Large-Scale, High-Resolution Benchmark for 3D Face Reconstruction

WildlifeDatasets: An open-source toolkit for animal re-identification

Addressing the Elephant in the Room: Robust Animal Re-Identification with Unsupervised Part-Based Feature Alignment

DogFLW: Dog Facial Landmarks in the Wild Dataset

Potency of Individual Identification of Japanese Macaques (Macaca fuscata) Using a Face Recognition System and a Limited Number of Learning Images

A Parallel Attention Network for Cattle Face Recognition

OpenAnimalTracks: A Dataset for Animal Track Recognition