PixelRec
收藏资源简介:
PixelRec是一个大规模的以图像为中心的推荐数据集,由西湖大学创建。该数据集包含约2亿用户-图像交互、3000万用户和40万高质量封面图像。PixelRec通过提供对原始图像像素的直接访问,使推荐模型能够直接从这些像素中学习项目表示。数据集主要用于研究图像内容驱动的推荐模型,特别是在冷启动和跨平台推荐场景中,PixelNet模型展示了其优势。PixelRec的发布旨在推动基于图像像素内容的研究,为推荐系统领域提供一个重要的资源和测试平台。
PixelRec is a large-scale image-centric recommendation dataset developed by Westlake University. It contains approximately 200 million user-image interactions, 30 million user profiles, and 400,000 high-quality cover images. By providing direct access to the raw pixel data of images, PixelRec enables recommendation models to directly learn item representations from these pixels. The dataset is primarily used for research on image content-driven recommendation models, and the PixelNet model has demonstrated its outstanding advantages in cold-start and cross-platform recommendation scenarios. The release of PixelRec aims to advance research on recommendation methods based on image pixel content, providing a vital resource and testbed for the recommender system field.

- 1An Image Dataset for Benchmarking Recommender Systems with Raw Pixels西湖大学 · 2023年



