Learning a Discriminative Feature Network for Semantic Segmentation

TLDR

Semantic segmentation methods still struggle with intra‑class inconsistency and inter‑class indistinction. The authors propose a Discriminative Feature Network comprising a Smooth Network and a Border Network to address these challenges. The Smooth Network uses channel attention and global average pooling to select discriminative features, while the Border Network employs deep semantic boundary supervision to distinguish bilateral boundary features. The DFN achieves state‑of‑the‑art mean IoU of 86.2 % on PASCAL VOC 2012 and 80.3 % on Cityscapes.

Abstract

Most existing methods of semantic segmentation still suffer from two aspects of challenges: intra-class inconsistency and inter-class indistinction. To tackle these two problems, we propose a Discriminative Feature Network (DFN), which contains two sub-networks: Smooth Network and Border Network. Specifically, to handle the intra-class inconsistency problem, we specially design a Smooth Network with Channel Attention Block and global average pooling to select the more discriminative features. Furthermore, we propose a Border Network to make the bilateral features of boundary distinguishable with deep semantic boundary supervision. Based on our proposed DFN, we achieve state-of-the-art performance 86.2% mean IOU on PASCAL VOC 2012 and 80.3% mean IOU on Cityscapes dataset.