High-quality depth estimation method based on depth prediction and an enhanced sub-network

A deep prediction and sub-network technology, applied in image enhancement, neural learning methods, biological neural network models, etc., can solve time-consuming problems and achieve the effects of simple procedures, improved accuracy, and easy construction

Inactive Publication Date: 2018-09-07
DALIAN UNIV OF TECH
View PDF6 Cites 22 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although it improves the performance of

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
  • High-quality depth estimation method based on depth prediction and an enhanced sub-network
  • High-quality depth estimation method based on depth prediction and an enhanced sub-network
  • High-quality depth estimation method based on depth prediction and an enhanced sub-network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0032] The high-quality depth estimation method based on depth prediction and enhanced sub-network of the present invention will be described in detail below in conjunction with the embodiments and accompanying drawings:

[0033] A method for high-quality depth estimation based on depth prediction and enhanced subnetworks, such as figure 1 As shown, the method comprises the following steps;

[0034] 1) Prepare initial data;

[0035] 1-1) Use two public datasets, the indoor dataset NYUV2 dataset, and the outdoor dataset Make3D dataset to train and evaluate the invention;

[0036] 1-2) For the indoor dataset NYUV2 dataset, 464 scenes, 1449 color images and corresponding depth images are selected as training data. According to the official division method, the training data is divided into 795 color images and corresponding depth images as the training set, and 654 color images and corresponding depth images as the test set.

[0037] 1-3) For the outdoor data set Make3D data s...

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 high-quality depth estimation method based on depth prediction and an enhanced sub-network, and belongs to the field of image processing and computer vision. The method comprises performing depth prediction on a color map by constructing a depth prediction sub-network, and using a depth enhanced sub-network to recover the resolution of a low-resolution image obtained by the depth prediction sub-network, thereby obtaining a high-resolution depth prediction map. The method is simple and easy to implement. The system is easy to construct, and can obtain a corresponding high-quality depth map from a single color map end to end by using a convolutional neural network. The information lost by the downsampling of the depth prediction sub-network is restored by the high-frequency information of the color image, and a high-quality and high-resolution depth map is finally obtained. The spatial pooling pyramid structure effectively solves the problem of predictive accuracy caused by different object sizes.

Description

technical field [0001] The invention belongs to the field of image processing computer vision, and relates to the use of a depth prediction sub-network to predict the depth of a color image, and the use of a depth enhancement sub-network to restore the resolution of a low-resolution depth map obtained by the depth prediction sub-network, thereby obtaining a high resolution The depth prediction map of , specifically relates to a high-quality depth estimation method based on depth prediction and enhanced subnetworks. Background technique [0002] The depth of field of a real scene, that is, depth information, is a method to measure the third dimension of a scene, and is widely used in various computer vision tasks, such as pose estimation, 3D modeling, etc. Although high-quality texture information can be easily obtained using color cameras, the acquisition of depth information is still a very challenging topic. In the traditional method, the acquisition of depth information ...

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
IPC IPC(8): G06T7/50G06N3/04G06N3/08
CPCG06N3/08G06T7/50G06T2207/10024G06T2207/20081G06T2207/20084G06N3/045G06T7/55G06T2207/10028
Inventor 叶昕辰李豪杰李阳段祥越
Owner DALIAN UNIV OF TECH
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