Fully Attentional Network for Semantic Segmentation

lastmod 2026-11-10created 2023-04-202 edits

Fully Attentional Network for Semantic Segmentation (FLANet) is a paper that proposes a fully attentional block which, unlike earlier approaches that apply spatial attention and channel attention separately in semantic segmentation, computes similarities along the spatial and channel axes together in a single attention map. It aims to reduce the loss of information along either axis while keeping computational cost down. This page keeps the link to the original arXiv paper.

Documentation

[2112.04108] Fully Attentional Network for Semantic Segmentation

https://arxiv.org/abs/2112.04108