HARRISON

HARRISON数据集由韩国科学技术院电气工程学院创建，专注于社交媒体中真实世界图像的标签推荐。该数据集包含57,383张来自Instagram的照片，每张照片平均附带4.5个标签。数据集的构建涉及从流行标签中筛选并收集图片，通过去除非英文标签和重复标签，最终形成高质量的数据集。HARRISON数据集的应用领域主要集中在图像理解，旨在通过视觉信息推荐合适的社交媒体标签，解决图像内容与标签匹配的问题。

The HARRISON dataset was created by the School of Electrical Engineering, Korea Advanced Institute of Science and Technology, focusing on tag recommendation for real-world images on social media. This dataset contains 57,383 photos sourced from Instagram, with an average of 4.5 tags attached to each photo. The dataset construction process involves screening and collecting images based on popular tags, followed by removing non-English tags and duplicate tags to ultimately form a high-quality dataset. The main application scenarios of the HARRISON dataset lie in the field of image understanding, where it aims to recommend appropriate social media tags using visual information to address the problem of matching image content with suitable tags.