According to IT House's report on July 17, the Hong Kong University of Science and Technology team developed an image segmentation AI model called Semantic-SAM. Compared with the SAM model previously released by Meta, Semantic-SAM has stronger granularity and semantic functions. It can segment and recognize objects at different levels of granularity, and provide semantic labels for the segmented entities. It is reported that Semantic-SAM is developed based on the Mask DINO framework, and its model structure is mainly improved in the decoder part, while supporting general segmentation and interactive segmentation. The research team realized the optimization of multi-granularity segmentation tasks and interactive segmentation tasks by adopting decoupled object classification and part classification methods to learn the semantic information of objects and parts. Experimental results show that Semantic-SAM outperforms Meta's SAM model in terms of segmentation quality and granularity controllability.