SeaTurtleID2022

SeaTurtleID2022是一个大规模的野外海龟重识别数据集，由捷克技术大学创建。该数据集包含438个独特个体的8729张照片，记录时间跨度长达13年，是目前最长的动物重识别数据集。数据集中的每张照片都包含多种标注，如身份、遭遇时间戳和身体部位分割掩码。该数据集不仅支持传统的随机分割，还提供了两种现实且生态驱动的分割方式：时间感知闭集和时间感知开集，以更真实地模拟动物重识别的实际场景。SeaTurtleID2022数据集主要用于动物重识别算法的基准测试，也可用于评估和测试包括对象检测、实例分割、语义分割、3D重建和概念漂移分析等基本问题。

SeaTurtleID2022 is a large-scale wild sea turtle re-identification dataset developed by the Czech Technical University. It contains 8,729 images of 438 unique individuals, with a recording time span of 13 years, making it the longest-duration animal re-identification dataset currently available. Each image in the dataset is accompanied by multiple annotations including identity, encounter timestamp, and body part segmentation masks. In addition to supporting standard random data splits, the dataset provides two realistic, ecology-driven partitioning strategies: time-aware closed-set and time-aware open-set, which more authentically simulate real-world scenarios for animal re-identification. The SeaTurtleID2022 dataset is primarily used for benchmarking animal re-identification algorithms, and can also be employed to evaluate and test fundamental tasks such as object detection, instance segmentation, semantic segmentation, 3D reconstruction, and concept drift analysis.

搜集汇总

背景与挑战

背景概述

SeaTurtleID2022是一个包含438只不同海龟8,729张照片的数据集，主要用于通过海龟独特鳞片模式进行个体识别的计算机视觉研究。该数据集具有13年的收集周期，提供丰富的注释信息，但使用受到严格限制，禁止商业和生态研究用途。