Improving Fully Convolution Network for Semantic Segmentation

TL;DR

This paper proposes the Improved Fully Convolution Network (IFCN), which enhances semantic segmentation by expanding receptive fields and fusing multi-scale context through dense skip connections, leading to state-of-the-art results.

Contribution

The paper introduces IFCN, a novel architecture with a context network and dense skip connections that improve segmentation accuracy without post-processing.

Findings

Significant performance improvements on ADE20K, Pascal Context, Pascal VOC 2012, and SUN-RGBD datasets.

Effective fusion of rich-scale context enhances segmentation reliability.

Architecture modifications outperform previous FCN models.

Abstract

Fully Convolution Networks (FCN) have achieved great success in dense prediction tasks including semantic segmentation. In this paper, we start from discussing FCN by understanding its architecture limitations in building a strong segmentation network. Next, we present our Improved Fully Convolution Network (IFCN). In contrast to FCN, IFCN introduces a context network that progressively expands the receptive fields of feature maps. In addition, dense skip connections are added so that the context network can be effectively optimized. More importantly, these dense skip connections enable IFCN to fuse rich-scale context to make reliable predictions. Empirically, those architecture modifications are proven to be significant to enhance the segmentation performance. Without engaging any contextual post-processing, IFCN significantly advances the state-of-the-arts on ADE20K (ImageNet scene parsing), Pascal Context, Pascal VOC 2012 and SUN-RGBD segmentation datasets.