Abstract

To improve the accuracy of stereo matching, the multi-scale dense attention network (MDA-Net) is proposed. The network introduces two novel modules in the feature extraction stage to achieve better exploit of context information: dual-path upsampling (DU) block and attention-guided context-aware pyramid feature extraction (ACPFE) block. The DU block is introduced to fuse different scale feature maps. It introduces sub-pixel convolution to compensate for the loss of information caused by the traditional interpolation upsampling method. The ACPFE block is proposed to extract multi-scale context information. Pyramid atrous convolution is adopted to exploit multi-scale features and the channel-attention is used to fuse the multi-scale features. The proposed network has been evaluated on several benchmark datasets. The three-pixel-error evaluated over all ground truth pixels is 2.10% on KITTI 2015 dataset. The experiment results prove that MDA-Net achieves state-of-the-art accuracy on KITTI 2012 and 2015 datasets.

Highlights

  • The depth information of objects is quite important for many computer vision tasks such as three-dimensional reconstruction, robot navigation, and autonomous driving

  • REVIEW of the Siamese feature extraction module is shown in Figure 2.3 of 12 extraction blocks

  • The result shows that using dual-path upsampling (DU) blocks to fuse multi-scale features can improve the accuracy of stereo matching effectively

Read more

Summary

Introduction

The depth information of objects is quite important for many computer vision tasks such as three-dimensional reconstruction, robot navigation, and autonomous driving. Obtain context information features for better disparity estimation in the stereo image pairs.trying. A multi-scale dense attention network (MDA-Net) is proposed to exploit to context information for better depth estimation. The block is proposed for high-level features to extract richer context information. The contributions of this contributions of this paper are summarized as follows: paper are summarized as follows: 1. A novel network without any post-processing for stereo matching is proposed; The. DUnetwork block iswithout introduced as a more effective upsampling of fusing multi-scale. A novel any post-processing for stereo matchingmethod is proposed; features; DU block is introduced as a more effective upsampling method of fusing multi-scale features; ACPFE block block is is adopted adopted to to extract extract richer richer context context information.

Multi-Scale
Siamese Feature Extraction
Stacked Dense Blocks
Dual-Path Upsampling Block
Attention-Guided Context-Aware Pyramid Feature Extraction Block
The outputs of Dense
Disparity Regression and Loss Function
Experiments
Datasets
Experimental Details
Ablation Study
Performance Comparison
Visualizations of depth the depth results of KITTI stereo column shows
Findings
Conclusions
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call