Abstract

The state-of-the-art in semantic segmentation is cur-rently represented by fully convolutional networks (FCNs).However, FCNs use large receptive fields and many pool-ing layers, both of which cause blurring and low spatialresolution in the deep layers. As a result FCNs tend to pro-duce segmentations that are poorly localized around objectboundaries. Prior work has attempted to address this issuein post-processing steps, for example using a color-basedCRF on top of the FCN predictions. However, these approaches require additional parameters and low-level features that are difficult to tune and integrate into the originanetwork architecture. Additionally, most CRFs use colorbased pixel affinities, which are not well suited for semantic segmentation and lead to spatially disjoint predictions. To overcome these problems, we introduce a Boundary Neural Field (BNF), which is a global energy model integrating FCN predictions with boundary cues. The boundary information is used to enhance semantic segment coherence and to improve object localization. Specifically, we first show that the convolutional filters of semantic FCNs provide good features for boundary detection. We then employ the predicted boundaries to define pairwise potentialsin our energy. Finally, we show that our energy decomposes semantic segmentation into multiple binary problems, which can be relaxed for efficient global optimization. We report extensive experiments demonstrating that minimization of our global boundary-based energy yields results superior to prior globalization methods, both quantitatively as well as qualitatively.