Given a textual query, Product Seeking aims to localize the image-level product from the gallery (Product Retrieval) and object-level product on an image (Product Grounding).
To research the Product Seeking with great practical application value, we collect two large-scale benchmark datasets TMPS and TLPS from Taobao Mall and Taobao Live domains with about 474k image-title pairs and 101k frame-description pairs respectively, and the locations of object-level products in images are manually annotated.
**Now, you can download dataset