"CAP-PKFashion"

"Vision-Language dataset for multimodal modeling in e-commerce. The dataset consists of 132,159 fashion products with 528,534 images. There are 12 primary attributes and structured attributes extracted from the description using our pipeline. There are 184 categories of fashion products in this dataset. This dataset includes 92,816 outfits. This dataset can be used for kinds of fashion tasks, including cross-modal retrieval , category recognition , fashion image generation, outfit generation, and outfit compatibility. All the data are collected from public resources and processed by our researchers. There are more tasks and applications that can be performed on this dataset about vision-language in other fields, as we provide more detailed features for each product."