Recent non-local self-attention methods have proven to be effective in capturing long-range dependencies for semantic segmentation. These methods usually form a similarity map of (\mathbb{R}^{C\times ...
Some results have been hidden because they may be inaccessible to you