BGLab/BioTrove

BioTrove是一个大型的、精心策划的图像数据集，旨在通过AI技术促进生物多样性研究。该数据集包含超过161.9M的图像，覆盖多个生物分类群，如爬行动物、植物、软体动物等，涉及366.6K物种。此外，还提供了三个新的基准数据集：BioTrove-Balanced、BioTrove-Unseen和BioTrove-LifeStages，分别用于平衡物种分布、评估模型在未见物种上的泛化能力和识别物种在不同发育阶段的能力。数据集还包括详细的元数据和软件工具，便于用户下载、访问和操作数据。

BioTrove is a large, curated image dataset designed to enable AI for biodiversity research. It comprises over 161.9M images across several taxonomic groups, including reptiles, plants, mollusks, and more, spanning 366.6K species. Additionally, it introduces three new benchmark datasets: BioTrove-Balanced, BioTrove-Unseen, and BioTrove-LifeStages, aimed at balanced species distribution, evaluating model generalization on unseen species, and recognizing species across various developmental stages, respectively. The dataset also includes comprehensive metadata and a collection of software tools for easy access, download, and manipulation of the data.

搜集汇总

背景与挑战

背景概述

BioTrove是一个包含161.9M张图像的大型生物多样性数据集，涵盖366.6K种物种，涉及多个生物分类群。数据集提供详细元数据和多个基准测试数据集，支持AI在生物多样性研究中的应用。