Multi-Scale Feature Aggregation by Cross-Scale Pixel-to-Region Relation Operation for Semantic Segmentation

Exploiting multi-scale features has shown great potential in tackling semantic segmentation problems. The aggregation is commonly done with sum or concatenation (concat) followed by convolutional (conv) layers. However, it fully passes down the high-level context to following hierarchy without considering their interrelation. In this work, we aim enable low-level feature aggregate complementary from adjacent maps a cross-scale pixel-to-region relation operation. We leverage propagation make long-range dependency capturable even high-resolution features. To end, employ an efficient pyramid network obtain propose Relational Semantics Extractor (RSE) and Propagator (RSP) for extraction respectively. Then stack several RSP into head achieve progressive top-down distribution of context. Experiment results on two challenging datasets Cityscapes COCO demonstrate that performs competitively both panoptic high efficiency. It outperforms DeeplabV3 [1] 0.7% 75% fewer FLOPs (multiply-adds) task.