1 paper
Yanqi Ge, Ye Huang, Wen Li +1
We introduced SSR, which utilizes SAM (segment-anything) as a strong regularizer during training, to greatly enhance the robustness of the image encoder for handling various domain…