The Herbarium Challenge 2019 Dataset

The Herbarium Challenge 2019 Dataset是由纽约植物园创建的一个包含46,469张数字化的植物标本图像的数据集，专注于开花植物家族Melastomataceae。该数据集通过专家鉴定和标注，每张标本图像都附有详细的元数据。数据集的创建旨在通过自动化技术加速植物物种的识别，特别是在未被描述的物种中。该数据集面临的挑战包括类别不平衡、专家依赖的小特征识别、同种个体间的形态变异以及不同物种间的视觉相似性。该数据集的应用领域主要集中在植物学研究，特别是加速新物种的发现和描述。

The Herbarium Challenge 2019 Dataset was developed by the New York Botanical Garden. It contains 46,469 digitized herbarium specimen images, focusing on the flowering plant family Melastomataceae. Each specimen image has been expertly verified and annotated, accompanied by detailed metadata. The dataset was created to accelerate plant species identification via automated technologies, particularly for undescribed species. The challenges associated with this dataset include class imbalance, recognition of expert-dependent fine morphological features, morphological variation among conspecific individuals, and visual similarity between different species. Its main application fields lie in botanical research, especially accelerating the discovery and description of new plant species.