GeoDE

GeoDE是由普林斯顿大学等机构创建的地理多样性评估数据集，包含61,940张来自40个类别和6个世界地区的图像，无个人识别信息。该数据集通过众包方式收集，旨在解决现有数据集的地理偏见问题，提高模型在不同地理区域的识别性能。GeoDE不仅用于评估模型，还可作为训练数据集，帮助改进现有模型，特别是在处理非西方地区的常见家庭物品识别时。

GeoDE is a geographic diversity evaluation dataset developed by Princeton University and other institutions. It contains 61,940 images covering 40 categories across 6 global regions, with no personally identifiable information (PII) included. Collected through crowdsourcing, this dataset is designed to address the geographic bias in existing datasets and improve the recognition performance of models across different geographic regions. Beyond being used for model evaluation, GeoDE can also serve as a training dataset to help enhance existing models, particularly in the recognition of common household items from non-Western regions.