SCNet: Enhancing Few-Shot Semantic Segmentation by Self-Contrastive Background Prototypes

Open in new window