Semantic-SAM: Segment and Recognize Anything at Any Granularity
This work addresses the need for flexible and comprehensive segmentation models in computer vision, offering a novel approach that integrates multiple segmentation tasks, though it is incremental in combining existing datasets and methods.
The paper tackles the problem of universal image segmentation by introducing Semantic-SAM, a model that segments and recognizes objects at any granularity, achieving semantic-awareness and granularity-abundance through joint training on multiple datasets, with experimental results showing performance improvements in tasks like panoptic and part segmentation.
In this paper, we introduce Semantic-SAM, a universal image segmentation model to enable segment and recognize anything at any desired granularity. Our model offers two key advantages: semantic-awareness and granularity-abundance. To achieve semantic-awareness, we consolidate multiple datasets across three granularities and introduce decoupled classification for objects and parts. This allows our model to capture rich semantic information. For the multi-granularity capability, we propose a multi-choice learning scheme during training, enabling each click to generate masks at multiple levels that correspond to multiple ground-truth masks. Notably, this work represents the first attempt to jointly train a model on SA-1B, generic, and part segmentation datasets. Experimental results and visualizations demonstrate that our model successfully achieves semantic-awareness and granularity-abundance. Furthermore, combining SA-1B training with other segmentation tasks, such as panoptic and part segmentation, leads to performance improvements. We will provide code and a demo for further exploration and evaluation.