Outline Objects using Deep Reinforcement Learning

TL;DR

This paper introduces DeepOutline, a novel deep reinforcement learning approach for semantic image segmentation that divides the task into boundary detection and tracing, outperforming existing methods on COCO dataset.

Contribution

It presents the first deep reinforcement learning framework for image segmentation, demonstrating a divide and conquer strategy that improves boundary detection and object outlining.

Findings

Outperforms other algorithms on COCO detection leaderboard for large persons

Effective boundary tracing within limited steps

Provides new insights into reinforcement learning for vision tasks

Abstract

Image segmentation needs both local boundary position information and global object context information. The performance of the recent state-of-the-art method, fully convolutional networks, reaches a bottleneck due to the neural network limit after balancing between the two types of information simultaneously in an end-to-end training style. To overcome this problem, we divide the semantic image segmentation into temporal subtasks. First, we find a possible pixel position of some object boundary; then trace the boundary at steps within a limited length until the whole object is outlined. We present the first deep reinforcement learning approach to semantic image segmentation, called DeepOutline, which outperforms other algorithms in Coco detection leaderboard in the middle and large size person category in Coco val2017 dataset. Meanwhile, it provides an insight into a divide and conquer way by reinforcement learning on computer vision problems.