Learning Multi-level Region Consistency with Dense Multi-label Networks for Semantic Segmentation

TL;DR

This paper introduces a dense multi-label network module that enforces multi-level region consistency in semantic segmentation, improving prediction plausibility and overall accuracy.

Contribution

It proposes a novel dense multi-label module that can be integrated into existing systems to enhance region consistency at multiple levels.

Findings

Reduces noisy and implausible predictions

Boosts segmentation accuracy

Removes label confusion

Abstract

Semantic image segmentation is a fundamental task in image understanding. Per-pixel semantic labelling of an image benefits greatly from the ability to consider region consistency both locally and globally. However, many Fully Convolutional Network based methods do not impose such consistency, which may give rise to noisy and implausible predictions. We address this issue by proposing a dense multi-label network module that is able to encourage the region consistency at different levels. This simple but effective module can be easily integrated into any semantic segmentation systems. With comprehensive experiments, we show that the dense multi-label can successfully remove the implausible labels and clear the confusion so as to boost the performance of semantic segmentation systems.