Competition vs. Concatenation in Skip Connections of Fully Convolutional Networks

TL;DR

This paper introduces a novel fully convolutional network architecture, CDFNet, which uses competitive maxout activations instead of concatenation in skip connections to improve semantic segmentation performance.

Contribution

The paper proposes competitive dense and unpooling blocks with maxout activations, enhancing layer competition and specialization in segmentation networks.

Findings

Improved segmentation accuracy on VISCERAL benchmark

Enhanced learning of specialized sub-networks

Outperforms state-of-the-art methods

Abstract

Increased information sharing through short and long-range skip connections between layers in fully convolutional networks have demonstrated significant improvement in performance for semantic segmentation. In this paper, we propose Competitive Dense Fully Convolutional Networks (CDFNet) by introducing competitive maxout activations in place of naive feature concatenation for inducing competition amongst layers. Within CDFNet, we propose two architectural contributions, namely competitive dense block (CDB) and competitive unpooling block (CUB) to induce competition at local and global scales for short and long-range skip connections respectively. This extension is demonstrated to boost learning of specialized sub-networks targeted at segmenting specific anatomies, which in turn eases the training of complex tasks. We present the proof-of-concept on the challenging task of whole body segmentation in the publicly available VISCERAL benchmark and demonstrate improved performance over multiple learning and registration based state-of-the-art methods.