Self-supervised Learning on Camera Trap Footage Yields a Strong Universal Face Embedder

2025/07/14 by Vladimir Iashin, H.S. Lee, Iashin, Vladimir +5 · 1 citation
Computer Science · #Artificial Intelligence (cs.AI) #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Face and Expression Recognition #Face recognition and analysis #Machine Learning (cs.LG) #Video Surveillance and Tracking Methods

paper · pdf · doi:10.48550/arxiv.2507.10552

openalex publication_date 2025/07/14 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Camera traps are revolutionising wildlife monitoring by capturing vast amounts of visual data; however, the manual identification of individual animals remains a significant bottleneck. This study introduces a fully self-supervised approach to learning robust chimpanzee face embeddings from unlabeled camera-trap footage. Leveraging the DINOv2 framework, we train Vision Transformers on automatically mined face crops, eliminating the need for identity labels. Our method demonstrates strong open-set re-identification performance, surpassing supervised baselines on challenging benchmarks such as Bossou, despite utilising no labelled data during training. This work underscores the potential of self-supervised learning in biodiversity monitoring and paves the way for scalable, non-invasive population studies.

Citations

• PetFace: A Large-Scale Dataset and Benchmark for Animal Identification
• PanAf20K: A Large Video Dataset for Wild Ape Detection and Behaviour Recognition
• The Faiss library
• WildlifeDatasets: An open-source toolkit for animal re-identification
• DINOv2: Learning Robust Visual Features without Supervision
• ViC-MAE: Self-Supervised Representation Learning from Images and Video with Contrastive Masked Autoencoders
• iBOT: Image BERT Pre-Training with Online Tokenizer
• Masked Autoencoders Are Scalable Vision Learners
• ByteTrack: Multi-Object Tracking by Associating Every Detection Box
• ByteTrack: Multi-object Tracking by Associating Every Detection Box
• YOLOX: Exceeding YOLO Series in 2021
• Emerging Properties in Self-Supervised Vision Transformers
• An Empirical Study of Training Self-Supervised Vision Transformers
• A Simple Framework for Contrastive Learning of Visual Representations
• Face Recognition: Primates in the Wild
• CosFace: Large Margin Cosine Loss for Deep Face Recognition
• MS-Celeb-1M: A Dataset and Benchmark for Large-Scale Face Recognition
• Deep Residual Learning for Image Recognition
• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows

Cited by

• PriVi: Towards A General-Purpose Video Model For Primate Behavior In The Wild

Related

• Automatic Individual Identification of Patterned Solitary Species Based on Unlabeled Video Data
• Count, Crop and Recognise: Fine-Grained Recognition in the Wild
• Wildlife Target Re-Identification Using Self-supervised Learning in Non-Urban Settings
• Past, Present, and Future Approaches Using Computer Vision for Animal Re-Identification from Camera Trap Data
• Domain-Adaptive Pretraining Improves Primate Behavior Recognition