Next Article in Journal
Use of Discrete Wavelet Transform to Assess Impedance Fluctuations Obtained from Cellular Micromotion
Next Article in Special Issue
EDSSA: An Encoder-Decoder Semantic Segmentation Networks Accelerator on OpenCL-Based FPGA Platform
Previous Article in Journal
A Robust Nonrigid Point Set Registration Method Based on Collaborative Correspondences
Previous Article in Special Issue
SGC-VSLAM: A Semantic and Geometric Constraints VSLAM for Dynamic Indoor Environments

Fast CNN Stereo Depth Estimation through Embedded GPU Devices

Universidad Tecnológica de Chile INACAP, Av. Vitacura 10.151, Vitacura 7650033, Santiago, Chile
Departamento de Ingeniería Eléctrica y Electrócnica, University of Bío-Bío, Concepción 4051381, Chile
Institute of Informatics, Universidad Austral de Chile, Valdivia 5111187, Chile
Escuela Superior Politécnica del Litoral, ESPOL, Campus Gustavo Galindo, Guayaquil EC090101, Ecuador
Computer Vision Center, Edifici O, Campus UAB, Bellaterra, 08193 Barcelona, Spain
Author to whom correspondence should be addressed.
Sensors 2020, 20(11), 3249;
Received: 5 April 2020 / Revised: 27 May 2020 / Accepted: 3 June 2020 / Published: 7 June 2020
(This article belongs to the Special Issue Intelligent Systems and Sensors for Robotics)
Current CNN-based stereo depth estimation models can barely run under real-time constraints on embedded graphic processing unit (GPU) devices. Moreover, state-of-the-art evaluations usually do not consider model optimization techniques, being that it is unknown what is the current potential on embedded GPU devices. In this work, we evaluate two state-of-the-art models on three different embedded GPU devices, with and without optimization methods, presenting performance results that illustrate the actual capabilities of embedded GPU devices for stereo depth estimation. More importantly, based on our evaluation, we propose the use of a U-Net like architecture for postprocessing the cost-volume, instead of a typical sequence of 3D convolutions, drastically augmenting the runtime speed of current models. In our experiments, we achieve real-time inference speed, in the range of 5–32 ms, for 1216 × 368 input stereo images on the Jetson TX2, Jetson Xavier, and Jetson Nano embedded devices. View Full-Text
Keywords: stereo matching; deep learning; embedded GPU stereo matching; deep learning; embedded GPU
Show Figures

Figure 1

MDPI and ACS Style

Aguilera, C.A.; Aguilera, C.; Navarro, C.A.; Sappa, A.D. Fast CNN Stereo Depth Estimation through Embedded GPU Devices. Sensors 2020, 20, 3249.

AMA Style

Aguilera CA, Aguilera C, Navarro CA, Sappa AD. Fast CNN Stereo Depth Estimation through Embedded GPU Devices. Sensors. 2020; 20(11):3249.

Chicago/Turabian Style

Aguilera, Cristhian A., Cristhian Aguilera, Cristóbal A. Navarro, and Angel D. Sappa 2020. "Fast CNN Stereo Depth Estimation through Embedded GPU Devices" Sensors 20, no. 11: 3249.

Find Other Styles
Note that from the first issue of 2016, MDPI journals use article numbers instead of page numbers. See further details here.

Article Access Map by Country/Region

Back to TopTop