非城市环境中基于自监督学习的野生动物目标重识别
Wildlife Target Re-Identification Using Self-supervised Learning in Non-Urban Settings
AI总结：
本文研究利用相机陷阱时间图像对进行自监督学习的野生动物重识别方法，证明自监督特征在数据有限时更鲁棒，并在所有下游任务中优于监督特征。
摘要：
野生动物重识别旨在跨不同观测匹配同一物种的个体。当前最先进的模型依赖类别标签来训练用于个体分类的监督模型。这种对标注数据的依赖推动了众多大规模野生动物数据集的构建。本研究探讨了自监督学习（Self-Supervised Learning，SSL）在野生动物重识别中的应用。我们利用相机陷阱数据中的时间图像对，在无监督条件下自动提取同一个体的两种不同视图。这些图像对从可能无穷无尽的视频数据流中训练出一个自监督模型。我们在开放世界场景以及各种野生动物下游任务的迁移学习中，将学习到的表示与监督特征进行了评估。实验结果分析表明，即使在数据有限的情况下，自监督模型也更具鲁棒性。此外，自监督特征在所有下游任务中均优于监督特征。代码可在 https://github.com/pxpana/SSLWildlife 获取。
英文摘要：
Wildlife re-identification aims to match individuals of the same species across different observations. Current state-of-the-art (SOTA) models rely on class labels to train supervised models for individual classification. This dependence on annotated data has driven the curation of numerous large-scale wildlife datasets. This study investigates self-supervised learning Self-Supervised Learning (SSL) for wildlife re-identification. We automatically extract two distinct views of an individual using temporal image pairs from camera trap data without supervision. The image pairs train a self-supervised model from a potentially endless stream of video data. We evaluate the learnt representations against supervised features on open-world scenarios and transfer learning in various wildlife downstream tasks. The analysis of the experimental results shows that self-supervised models are more robust even with limited data. Moreover, self-supervised features outperform supervision across all downstream tasks. The code is available here https://github.com/pxpana/SSLWildlife.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
