Herbarium specimen records from GBIF used for Digi-Leap

We searched the GBIF database on September 11, 2023 for all specimens meeting the following criteria; members of Tracheophyta, collected in the United States, record containing a specimen image, preserved specimen type, english language and not cultivated. This search resulted in 4,091,778 records of which we randomly selected 2,000 to use for Digi-Leap development and training. All GBIF records have the prefix "DL_GBIF" and include a rights.txt file. There are 3 separate scoring files. One for each model; rule-based, GPT and ensemble.我们于2023年9月11日检索了GBIF数据库，筛选符合以下全部条件的标本：隶属于维管植物门（Tracheophyta）、采集于美国境内、记录包含标本图像、为保藏标本类型、标注语言为英语且非人工栽培个体。本次检索共获取4091778条记录，我们从中随机抽取2000条用于Digi-Leap的开发与训练。 所有GBIF记录均带有前缀"DL_GBIF"，且包含一个rights.txt文件。 本次数据集包含3个独立的评分文件，分别对应三类模型：基于规则的模型、GPT模型与集成学习模型。