RefMatte is the first large-scale challenging dataset under the task referring image matting, generated by a comprehensive image composition and expression generation engine on top of current public high-quality matting foregrounds with flexible logics and re-labelled diverse attributes. RefMatte consists of 230 object categories, 47,500 images, 118,749 expression-region entities, and 474,996 expressions, which can be further extended easily in the future.

RefMatte comes along with two settings: keyword-based and expression-based. The former one takes a high-resolution image and a keyword as input, while the latter one takes a high-resolution image and a flowery expression as input.

Additionally, we construct a real-world test set with 100 high-resolution natural images and manually annotate complex phrases to evaluate the out-of-domain generalization abilities of RIM methods, named as RefMatte-RW100.