GSANet: Semantic Segmentation with Global and Selective Attention

TL;DR

GSANet introduces a novel attention-based architecture for semantic segmentation that enhances accuracy on edge devices and benchmarks well on standard datasets.

Contribution

The paper presents GSANet with a new sparsemax global attention and selective attention mechanism, and benchmarks it with low-complexity networks like MobileNetEdge.

Findings

GSANet achieves state-of-the-art accuracy on ADE20k and Cityscapes.

GSANet improves segmentation accuracy with MobileNetEdge.

The proposed attention mechanisms enhance multi-scale contextual information aggregation.

Abstract

This paper proposes a novel deep learning architecture for semantic segmentation. The proposed Global and Selective Attention Network (GSANet) features Atrous Spatial Pyramid Pooling (ASPP) with a novel sparsemax global attention and a novel selective attention that deploys a condensation and diffusion mechanism to aggregate the multi-scale contextual information from the extracted deep features. A selective attention decoder is also proposed to process the GSA-ASPP outputs for optimizing the softmax volume. We are the first to benchmark the performance of semantic segmentation networks with the low-complexity feature extraction network (FXN) MobileNetEdge, that is optimized for low latency on edge devices. We show that GSANet can result in more accurate segmentation with MobileNetEdge, as well as with strong FXNs, such as Xception. GSANet improves the state-of-art semantic segmentation accuracy on both the ADE20k and the Cityscapes datasets.