FoodX-251
收藏资源简介:
FoodX-251数据集由SRI国际创建,包含251个细粒度食品类别,总计158k张图像,主要从网络收集。数据集分为118k训练图像、12k验证图像和28k测试图像,所有图像均有人工验证的标签。数据集创建过程中,首先从Food-101数据集的101个食品类别中提取并筛选出251个类别,然后通过网络图像搜索下载相应图像,并进行去重和人工验证。该数据集主要应用于食品分类,旨在解决食品识别中的细粒度分类问题,为计算机视觉研究提供了一个独特的数据集。
The FoodX-251 dataset, created by SRI International, consists of 251 fine-grained food categories with a total of 158,000 images primarily collected from the web. The dataset is split into 118k training images, 12k validation images, and 28k test images, and all images have manually verified labels. During the dataset creation process, 251 categories were first extracted and screened from the 101 food categories of the Food-101 dataset, then corresponding images were downloaded via web image searches, followed by deduplication and manual verification. This dataset is mainly used for food classification tasks, aiming to solve the fine-grained classification problem in food recognition, providing a unique benchmark dataset for computer vision research.




