Weak supervision semantic segmentation method and application thereof

A technology of semantic segmentation and weak supervision, applied in the field of computer vision, can solve the problem of low precision of semantic segmentation, achieve the effect of simplifying computational complexity, enhancing expansion efficiency, and improving expansion efficiency

Active Publication Date: 2020-07-28
HUAZHONG UNIV OF SCI & TECH
View PDF4 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a high-level semantics-based weakly supervised semantic segmentation method and its application to solve the problem of poor semantic segmentation accuracy caused by erasing position and expansion efficiency in the existing erasure area expansion type weakly supervised semantic segmentation method. high technical issues

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
  • Weak supervision semantic segmentation method and application thereof
  • Weak supervision semantic segmentation method and application thereof
  • Weak supervision semantic segmentation method and application thereof

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0040] A weakly supervised semantic segmentation method 100, such as figure 1 As shown, using the pre-trained semantic erasing region expansion classification network for weakly supervised semantic segmentation, the following steps are performed:

[0041]Step 110, performing the first-stage feature extraction and high-level semantic integration classification on the image to be semantically segmented in order to obtain the first category response map corresponding to the image;

[0042] Step 120, erasing the areas with high responsiveness in the first category response map, and performing the second-stage high-level semantic integration classification on the erased category response map to obtain the second category response map;

[0043] Step 130: Add and fuse the corresponding positions of the first category response map and the second category response map respectively to obtain the fused category response map, and perform background threshold cutting processing on the fuse...

Embodiment 2

[0083] An application of any weakly supervised semantic segmentation method as described in Embodiment 1, which is used to perform semantic segmentation on multiple pictures to be semantically segmented to obtain a category segmentation area map, based on multiple pictures to be semantically segmented and their corresponding Class Segmentation Region Maps, Training Semantic Segmentation Networks.

[0084] Another semantic segmentation method uses the above-mentioned semantic segmentation network to perform semantic segmentation on the image to be semantically segmented to complete the semantic segmentation.

[0085] This embodiment adopts the weakly supervised semantic segmentation method described in Embodiment 1. Due to its erasing-type region expansion algorithm based on high-level semantic information, a fused category response map is obtained through multi-stage expansion, and the fused category response map is processed. Region segmentation, the obtained category segment...

Embodiment 3

[0087] A storage medium, in which instructions are stored, and when the computer reads the instructions, the computer is made to execute any weakly supervised semantic segmentation method as described in the first embodiment above and / or as described in the second embodiment above A semantic segmentation method described above.

[0088] The relevant technical solutions are the same as those in Embodiment 1 and Embodiment 2, and will not be repeated here.

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 belongs to the technical field of computer vision, and particularly discloses a weak supervision semantic segmentation method and an application. The method comprises that: a pre-trainedsemantic erasure type region expansion classification network used for weak supervision semantic segmentation is adopted, first-stage feature extraction and high-level semantic integration classification are sequentially carried out on a picture to be semantically segmented, and a first class response graph corresponding to the picture is obtained; an area with high responsivity in the first category response diagram is erased, and second-stage high-level semantic integration classification is performed on the erased category response diagram to obtain a second category response diagram; andthe corresponding positions of the first category response diagram and the second category response diagram are added and fused to obtain a fused category response diagram, and background threshold segmentation processing is performed on the fused category response diagram to obtain a category segmentation region diagram. The erasure type region expansion classification network structure is greatly simplified, the expansion effect is good, the region expansion exploration efficiency is greatly improved, and the weak supervision semantic segmentation effect is further enhanced.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and more specifically relates to a weakly supervised semantic segmentation method and its application. Background technique [0002] Semantic segmentation is one of the classic problems in computer vision, and it can be widely used in fine segmentation scenarios such as vision-based road scene segmentation and remote sensing image segmentation. For a given picture, different category regions (including several foreground object categories and background) are segmented by a certain algorithm. The fully supervised semantic segmentation algorithm based on deep learning requires pixel-level category labeling information, which is often fine and time-consuming, and limits the diversity of object categories and specific practical applications. Other weakly supervised marks such as object boxes, stick figures, and point marks greatly reduce the cost of marking, and can mark more training pictur...

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): G06T7/194G06T7/11G06T7/136G06N3/04
CPCG06T7/194G06T7/11G06T7/136G06N3/045
Inventor 刘佳惠高常鑫桑农
Owner HUAZHONG 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