Optimizing 3D Convolution Kernels on Stereo Matching for Resource Efficient Computations

Xiao, Jianqiang; Ma, Dianbo; Yamane, Satoshi

doi:10.3390/s21206808

Open AccessArticle

Optimizing 3D Convolution Kernels on Stereo Matching for Resource Efficient Computations

by

Jianqiang Xiao

,

Dianbo Ma

and

Satoshi Yamane

^*

Division of Electrical Engineering and Computer Science, Kanazawa University, Kanazawa 920-1192, Japan

^*

Author to whom correspondence should be addressed.

Sensors 2021, 21(20), 6808; https://doi.org/10.3390/s21206808

Submission received: 16 September 2021 / Revised: 4 October 2021 / Accepted: 8 October 2021 / Published: 13 October 2021

(This article belongs to the Special Issue Stereo Vision-Based Perception, Navigation and Control for Intelligent Autonomous Systems)

Download

Browse Figures

Versions Notes

Abstract

Despite recent stereo matching algorithms achieving significant results on public benchmarks, the problem of requiring heavy computation remains unsolved. Most works focus on designing an architecture to reduce the computational complexity, while we take aim at optimizing 3D convolution kernels on the Pyramid Stereo Matching Network (PSMNet) for solving the problem. In this paper, we design a series of comparative experiments exploring the performance of well-known convolution kernels on PSMNet. Our model saves the computational complexity from

256.66

G MAdd (Multiply-Add operations) to

69.03

G MAdd (

198.47

G MAdd to

10.84

G MAdd for only considering 3D convolutional neural networks) without losing accuracy. On Scene Flow and KITTI 2015 datasets, our model achieves results comparable to the state-of-the-art with a low computational cost.

Keywords: stereo matching; lightweight 3D kernels; 3D channel-wise attention; network design; 3D vision

Share and Cite

MDPI and ACS Style

Xiao, J.; Ma, D.; Yamane, S. Optimizing 3D Convolution Kernels on Stereo Matching for Resource Efficient Computations. Sensors 2021, 21, 6808. https://doi.org/10.3390/s21206808

AMA Style

Xiao J, Ma D, Yamane S. Optimizing 3D Convolution Kernels on Stereo Matching for Resource Efficient Computations. Sensors. 2021; 21(20):6808. https://doi.org/10.3390/s21206808

Chicago/Turabian Style

Xiao, Jianqiang, Dianbo Ma, and Satoshi Yamane. 2021. "Optimizing 3D Convolution Kernels on Stereo Matching for Resource Efficient Computations" Sensors 21, no. 20: 6808. https://doi.org/10.3390/s21206808

APA Style

Xiao, J., Ma, D., & Yamane, S. (2021). Optimizing 3D Convolution Kernels on Stereo Matching for Resource Efficient Computations. Sensors, 21(20), 6808. https://doi.org/10.3390/s21206808

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu

Optimizing 3D Convolution Kernels on Stereo Matching for Resource Efficient Computations

Abstract

Share and Cite

Article Metrics

Article Access Statistics

Further Information

Guidelines

MDPI Initiatives

Follow MDPI