Feedforward semantic segmentation with zoom-out features

TL;DR

This paper presents a feedforward neural network architecture for semantic segmentation that leverages multi-scale region features, achieving state-of-the-art accuracy without complex inference procedures.

Contribution

The authors introduce a novel zoom-out feature extraction method combined with a simple feedforward network for semantic segmentation, avoiding explicit structured prediction.

Findings

Achieved 64.4% average accuracy on PASCAL VOC 2012

Outperformed previous state-of-the-art methods

Simplified the segmentation process by eliminating complex inference

Abstract

We introduce a purely feed-forward architecture for semantic segmentation. We map small image elements (superpixels) to rich feature representations extracted from a sequence of nested regions of increasing extent. These regions are obtained by "zooming out" from the superpixel all the way to scene-level resolution. This approach exploits statistical structure in the image and in the label space without setting up explicit structured prediction mechanisms, and thus avoids complex and expensive inference. Instead superpixels are classified by a feedforward multilayer network. Our architecture achieves new state of the art performance in semantic segmentation, obtaining 64.4% average accuracy on the PASCAL VOC 2012 test set.