Few-Shot Segmentation via Rich Prototype Generation and Recurrent Prediction Enhancement

Prototype learning and decoder construction are the keys for few-shot segmentation. However, existing methods use only a single prototype generation mode, which can not cope with intractable problem of objects various scales. Moreover, one-way forward propagation adopted by previous may cause information dilution from registered features during decoding process. In this research, we propose rich module (RPGM) recurrent prediction enhancement (RPEM) to reinforce paradigm build unified memory-augmented segmentation, respectively. Specifically, RPGM combines superpixel K-means clustering generate complementary scale relationships adapt gap between support query images. The RPEM utilizes mechanism design round-way decoder. way, provide object-aware continuously. Experiments show that our method consistently outperforms other competitors on two popular benchmarks PASCAL- $${{5}^{i}}$$ COCO- $${{20}^{i}}$$ .