End-to-end binocular stereo matching network with extremely small calculation amount based on full binary convolution

A technology of binocular stereo matching and computational complexity, applied in computing, biological neural network models, instruments, etc., can solve the problem of large computational complexity of stereo matching solutions, and achieve the effect of improving feature quality and reducing overhead.

Pending Publication Date: 2022-07-29
SOUTHEAST UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] Purpose of the invention: In order to solve the problem of excessive calculation of existing stereo matching schemes, a lightweight end-to-end binocular stereo matching network based on full binary convolution is proposed

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • End-to-end binocular stereo matching network with extremely small calculation amount based on full binary convolution
  • End-to-end binocular stereo matching network with extremely small calculation amount based on full binary convolution
  • End-to-end binocular stereo matching network with extremely small calculation amount based on full binary convolution

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0053] The present invention will be further explained below in conjunction with the accompanying drawings.

[0054] A binary 2D convolution module dedicated to stereo matching, such as figure 1 As shown in (a), when the input and output feature maps are 3-dimensional data:

[0055] Step A1: Binarization operation on the input feature map: The input feature map will be processed by the RSign2D symbol function, and then binarized to -1 / +1 after adding the corresponding offset value to the channel dimension;

[0056] Step A2: Binarization operation on the weights: firstly obtain the mean value of the weights in the channel dimension as the channel scaling factor, and then use the Sign function to binarize the weights to -1 / +1;

[0057] Step A3: Binary convolution operation: Convolve the binary input feature map obtained in step A1 with the binary weight obtained in step A2, and then multiply the channel dimension by the channel scaling factor obtained in step A2 to obtain two. ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a full binary convolution end-to-end binocular stereo matching network (PBCStereo) with extremely low calculation amount, which is used for a binocular depth estimation task. Based on the design of a binary convolution module, a binary up-sampling module and an input layer coding method, all convolution processes in PBCStereo are binary convolution. Compared with other binocular stereo matching methods depending on floating point convolution, the PBCSterio saves more than 10 times of calculation times, and for an input image pair with the input resolution of 512 * 256, the calculation amount overhead of the PBCSterio for completing depth estimation is only 0.64 G OPs. Therefore, the PBCStereo is more easily deployed on edge equipment with limited computing resources, meanwhile, the PBCStereo also realizes considerable accuracy on SceneFlow and KITTI data sets, the end point error on the SceneFlow is 1.84, the error percentage of three pixel points on the KITTI 2012 is 4.46%, and the error percentage of three pixel points on the KITTI 2015 is 4.73%.

Description

technical field [0001] The invention proposes an end-to-end binocular stereo matching network based on full binary convolution with very little calculation amount, which is used for the task of depth estimation, and belongs to the field of computer vision stereo matching. Background technique [0002] In complex computer vision tasks such as autonomous driving, augmented reality, and robot navigation, the key to the technology is to obtain the depth information of images correctly, and binocular stereo matching has always been the focus of research due to its high accuracy and low cost. At present, many schemes for binocular stereo matching using deep learning have greatly improved the accuracy of data processing, but at the same time, these schemes have more and more parameters, more and more computing requirements, and deeper network structures. Applications that deploy depth estimation usually run on mobile devices and embedded platforms, including drones, vehicles, smart...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06N3/04G06N3/08G06T3/40
CPCG06N3/08G06T3/4007G06N3/045
Inventor 齐志蔡家璇刘昊
Owner SOUTHEAST UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap