Attention-Based Self-Supervised Learning Monocular Depth Estimation With Edge Refinement

Abstract

Learning depth from a single image extracted from unlabeled videos has been attracting significant attention in the past few years. In this work, we propose a new depth estimation neural network with edge refinement to predict depth. First, we introduce a dual attention module into depth prediction module to integrate global information into local features and improve local features' capability of representation. Second, to increase the details between objects in scenes, we propose a subnetwork to predict edges in four directions and combine the predicted depth and edges to increase the details by propagation operation. Besides, we integrate the gradients of the image into the photometric reprojection loss to handle the confusion caused by changing brightness. We conduct experiments on KITTI datasets and show that our network achieves the state-of-art result.

Attention Mechanism Used in Monocular Depth Estimation: An Overview

“…Zhang et al reported a selfsupervised method in which channel and spatial attentions are embedded at the bottleneck of an encoder-decoder architecture [41]. Similar works include [42,43], in which channel and spatial attention blocks are parallel or sequentially inserted at the bottleneck. The channel and spatial attention modules can also be inserted into any stage of the decoder [44].…”