Convolutional Oriented Boundaries

TL;DR

Convolutional Oriented Boundaries (COB) is a fast, CNN-based method that produces multiscale, oriented contours and hierarchical segmentations, significantly improving state-of-the-art performance across multiple datasets.

Contribution

COB introduces a novel approach that estimates contour orientation and uses a sparse boundary representation for efficient hierarchical segmentation from CNNs.

Findings

Achieves state-of-the-art contour detection on multiple datasets

Provides accurate region hierarchies and object proposals

Generalizes well to unseen categories and datasets

Abstract

We present Convolutional Oriented Boundaries (COB), which produces multiscale oriented contours and region hierarchies starting from generic image classification Convolutional Neural Networks (CNNs). COB is computationally efficient, because it requires a single CNN forward pass for contour detection and it uses a novel sparse boundary representation for hierarchical segmentation; it gives a significant leap in performance over the state-of-the-art, and it generalizes very well to unseen categories and datasets. Particularly, we show that learning to estimate not only contour strength but also orientation provides more accurate results. We perform extensive experiments on BSDS, PASCAL Context, PASCAL Segmentation, and MS-COCO, showing that COB provides state-of-the-art contours, region hierarchies, and object proposals in all datasets.