SAM2-UNet: Segment Anything 2 Makes Strong Encoder for Natural and Medical Image Segmentation
This work provides a simple and effective framework for improving segmentation performance across natural and medical domains, though it is incremental as it builds on existing foundation models and U-shaped designs.
The paper tackles the problem of versatile image segmentation by proposing SAM2-UNet, which uses the Segment Anything Model 2 as an encoder in a U-shaped architecture, and demonstrates that it outperforms specialized state-of-the-art methods on tasks like camouflaged object detection and polyp segmentation.
Image segmentation plays an important role in vision understanding. Recently, the emerging vision foundation models continuously achieved superior performance on various tasks. Following such success, in this paper, we prove that the Segment Anything Model 2 (SAM2) can be a strong encoder for U-shaped segmentation models. We propose a simple but effective framework, termed SAM2-UNet, for versatile image segmentation. Specifically, SAM2-UNet adopts the Hiera backbone of SAM2 as the encoder, while the decoder uses the classic U-shaped design. Additionally, adapters are inserted into the encoder to allow parameter-efficient fine-tuning. Preliminary experiments on various downstream tasks, such as camouflaged object detection, salient object detection, marine animal segmentation, mirror detection, and polyp segmentation, demonstrate that our SAM2-UNet can simply beat existing specialized state-of-the-art methods without bells and whistles. Project page: \url{https://github.com/WZH0120/SAM2-UNet}.