Transformer-based methods are recently popular in vision tasks because of their capability to model global dependencies alone. However, it limits the performance of networks due to the lack of modeling local context and global-local correlations of multi-scale features. In this paper, we present MIS...
No comments yet
Be the first to share your thoughts!