Voxel51/BIOSCAN-30k

BIOSCAN-30k数据集是BIOSCAN-5M数据集的一个子集，包含3万个样本，涵盖了从昆虫等节肢动物的高分辨率图像到DNA条形码序列、分类标签、地理位置信息和标本大小数据等多种模态的信息。这个数据集旨在推进机器学习在生物多样性监测方面的应用，并支持物种分类、新物种发现、生物多样性监测和多模态学习等任务。数据集由国际联合体策划，并由加拿大政府的新前沿研究基金、CFI大型科学倡议、Walder基金会和NVIDIA学术赠款资助。

The BIOSCAN-30k dataset is a subset of the BIOSCAN-5M dataset, containing 30,000 samples that cover a variety of modalities from high-resolution images of insects and other arthropods to DNA barcode sequences, taxonomic labels, geographical information, and specimen size data. This dataset is designed to advance machine learning applications in biodiversity monitoring and supports tasks such as taxonomic classification, novel species discovery, biodiversity monitoring, and multimodal learning. The dataset is curated by an international consortium and funded by the Government of Canadas New Frontiers in Research Fund, CFI Major Science Initiatives, the Walder Foundation, and the NVIDIA Academic Grant.