Revisiting Multi-Scale Feature Fusion for Semantic Segmentation. (arXiv:2203.12683v2 [cs.CV] UPDATED)

It is commonly believed that high internal resolution combined with expensive
operations (e.g. atrous convolutions) are necessary for accurate semantic
segmentation, resulting in slow speed and large memory usage. In this paper, we
question this belief and demonstrate that neither high internal resolution nor
atrous convolutions are necessary. Our intuition is that although segmentation
is a dense per-pixel prediction task, the semantics of each pixel often depend
on both nearby neighbors and far-away context; therefore, a more powerful
multi-scale feature …

arxiv cv feature fusion scale segmentation semantic