imageomics/phylo-fish
收藏资源简介:
Phylo-Fish是一个机器学习就绪的数据集,包含38种硬骨鱼(Teleost fishes),每个物种平均有200张图像。这个半平衡子集来源于大湖区入侵物种网络项目(GLIN)的一个更大的鱼类图像集合。数据集中的图像来自五个鱼类学研究收藏,这些收藏参与了GLIN项目,原始数据通过Fish-AIR数据库获取。图像经过预处理,被调整大小并适当填充为256×256像素分辨率。数据集被分为训练集和验证集,比例分别为80%和20%,总共包含4,140张训练图像和1,294张测试图像。数据包括图像文件以及元数据文件(metadata.csv),其中包含图像路径、文件名、科学名称、属、科、物种、分割信息、许可证、来源等字段。该数据集旨在支持生物学中的性状发现研究,特别是通过系统发育引导的神经网络来发现进化性状。
Phylo-Fish is a machine learning-ready dataset encompassing 38 species of Teleost fishes, with an average of 200 images per species. This semi-balanced subset is derived from a larger collection of fish images from the Great Lakes Invasive Species Network (GLIN) project. The images in the dataset originate from five ichthyological research collections that participated in the GLIN project, with raw data obtained via the Fish-AIR database. All images have been preprocessed: resized and appropriately padded to a resolution of 256×256 pixels. The dataset is split into training and validation subsets at a ratio of 80% and 20%, respectively, containing a total of 4,140 training images and 1,294 test images. The dataset includes both image files and a metadata file (metadata.csv), which contains fields such as image path, filename, scientific name, genus, family, species, segmentation information, license, and data source. This dataset is designed to support trait discovery research in biology, particularly for identifying evolutionary traits via phylogeny-guided neural networks.




