Image semantic segmentation method based on context and shallow space coding and decoding network

A semantic segmentation and context technology, applied in the field of computer vision and deep learning, can solve the problems of not obtaining high-quality semantic context features, ignoring the shallow spatial details of the network, etc.

Pending Publication Date: 2020-05-08
JIANGXI UNIV OF SCI & TECH
View PDF10 Cites 18 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] At present, some semantic segmentation models based on deep learning use fully convolutional network structures to obtain effective semantic context information, while

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
  • Image semantic segmentation method based on context and shallow space coding and decoding network
  • Image semantic segmentation method based on context and shallow space coding and decoding network
  • Image semantic segmentation method based on context and shallow space coding and decoding network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0053] In order to verify the effectiveness of the proposed module, the following ablation experiments were performed on the CamVid dataset. The present invention uses four schemes to evaluate the performance of the hybrid expansion module and the residual pyramid feature extraction module: (1) The context path at the encoding end uses only the hybrid expansion convolution module; (2) the context path at the encoding end uses only the residual pyramid feature Extraction module; (3) There is no hybrid expansion convolution and residual pyramid feature extraction module on the encoding side context path; (4) Hybrid expansion convolution module and residual pyramid feature extraction module are used on the encoding side context path. The experimental results are shown in Table 1. It can be seen from the table that the best segmentation performance is obtained when the hybrid expansion convolution module and the residual pyramid feature extraction module are used at the same time, i...

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

As for current image semantic segmentation study, a more effective algorithm based on two factors used for how to extract effective semantic context information and restore spatial detail informationis designed. In some existing semantic segmentation models, spatial detail information of a network shallow layer is ignored and some existing semantic segmentation models do not acquire high-qualitysemantic context features. In order to solve the problem, the invention provides an image semantic segmentation method based on context and shallow space coding and decoding networks. A two-branch strategy is adopted at a coding end, a new semantic context module is designed for a context branch to obtain high-quality semantic context information, a space branch is designed to be of an inverted U-shaped structure, a chain type inverted residual module is combined, and the semantic information is improved while space detail information is reserved. At the decoding end, an optimization module isdesigned to further optimize the fused context information and space information. Through a large number of experiments, competitive results are obtained on the three reference data sets CamVid, SUNRGB-D and Cityscapes.

Description

Technical field [0001] The present invention relates to the fields of computer vision and deep learning, and is specifically an image semantic segmentation method based on context and shallow space coding and decoding networks. Background technique [0002] Semantic segmentation is one of the most popular research problems in the field of computer vision, due to its wide application prospects, such as the current autopilot technology, VR technology, and automatic medical analysis, which have attracted much attention. The semantic segmentation task is to assign a corresponding semantic label to each pixel in the image, such as car, road, sky, etc. It is also considered a pixel-level classification task. Different from image classification and target detection, image semantic segmentation needs to identify the object category in the image and find the location of the object in the image. Therefore, semantic segmentation is also one of the most challenging tasks in computer vision. ...

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): G06T7/10
CPCG06T7/10G06T2207/20081G06T2207/20084Y02T10/40
Inventor 罗会兰黎宵童康
Owner JIANGXI UNIV OF SCI & 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