LoTE-Animal

LoTE-Animal是一个历经 12 年收集的大规模濒危动物数据集，数据集中包含了环境的巨大变化，例如生态季节、天气条件、时期、观点和栖息地场景。数据集里对 11 种濒危动物进行行为标注，包括用于动作识别任务的 10K 视频序列、用于目标检测、实例分割和姿势估计任务的 28K 图像。
发表于ICCV2023，https://lote-animal.github.io/

LoTE-Animal is a large-scale endangered animal dataset compiled over a 12-year collection period. It encompasses substantial environmental variations, including ecological seasons, weather conditions, time periods, viewpoints, and habitat scenarios. The dataset features behavioral annotations for 11 endangered animal species, comprising 10K video sequences for action recognition tasks, as well as 28K images tailored for object detection, instance segmentation, and pose estimation tasks. Published in ICCV 2023, its official project website is https://lote-animal.github.io/

搜集汇总

背景与挑战

背景概述

LoTE-Animal是一个历经12年收集的大规模濒危动物数据集，专注于11种濒危动物的行为理解，包含环境变化如生态季节和天气条件。数据集提供10K视频序列用于动作识别任务，以及28K图像用于目标检测、实例分割和姿势估计任务，发表于ICCV2023，支持多模态计算机视觉研究。