SAW-IT-Plus Video Dataset

The SAW-IT-Plus dataset contains 11,458 videos collected in the wild, and 22 homemade videos (snake category). Videos are arranged in 8 main categories of animals (frogs - 0, snakes - 1, lizards - 2, birds - 3, small mammals < 2kg - 4, medium or large mammals > 2kg - 5, spiders - 7 and scorpions - 8). Echidnas - originally category 6 – were merged with big mammals. Some videos of crustacea and other reptiles are available but not classified. Empty videos (7,896) were added to allow for further testing of the algorithm. They are separated in 3 categories (details in Table 1). CSV files are available to detail the species for frogs, lizards, birds and small mammals for each video. Because the videos were mainly collected from real-world data; the number of videos for each animal category are unbalanced (Table 1). This folder also contains training images used to automatically detect videos containing animals in our overall dataset. More information available in the ReadMe files. The dataset was collected in Victoria, Australia, from February to October 2021 as part of the ERP22 (formerly ARI-PPD 05) grant.SAW-IT-Plus数据集包含11458条野外采集的视频，以及22条自制的蛇类类别视频。所有视频被划分为8个主要的动物类别（蛙类-0、蛇类-1、蜥蜴类-2、鸟类-3、体重小于2kg的小型哺乳类-4、体重大于2kg的中大型哺乳类-5、蜘蛛类-7以及蝎类-8）。原本属于类别6的针鼹，现已被并入大型哺乳类类别。另有部分甲壳类与其他爬行类的视频可供使用，但尚未完成分类。为支持后续算法测试，本数据集额外加入了7896条空视频，这些空视频被划分为3个子类别（详细信息见表1）。可通过CSV文件查询每条视频对应的蛙类、蜥蜴类、鸟类以及小型哺乳类的具体物种信息。由于视频主要采集自真实场景数据，各动物类别的视频数量存在不均衡现象（详见表1）。该文件夹还包含用于在整体数据集中自动检测含动物视频的训练图像。更多详细信息可参阅ReadMe文件。本数据集于2021年2月至10月期间在澳大利亚维多利亚州采集，作为ERP22（前身为ARI-PPD 05）资助项目的一部分。