A Contour-Based Small-Sample Semantic Segmentation Method

A semantic segmentation and small sample technology, applied in the field of digital image intelligent processing, can solve problems such as misclassification, combination of reference image and segmented image, poor object edge segmentation, etc., to improve speed and solve the effect of poor edge segmentation

Active Publication Date: 2022-05-20
FUDAN UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Some existing small-sample semantic segmentation methods have poor object edge segmentation and misclassification problems, because the reference image and the segmented image are not well combined

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
  • A Contour-Based Small-Sample Semantic Segmentation Method
  • A Contour-Based Small-Sample Semantic Segmentation Method
  • A Contour-Based Small-Sample Semantic Segmentation Method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0037] For a segmented image I q , given a reference image I s , it is necessary to segment objects of the same category in the image to be segmented according to the category in the reference image. The specific steps are:

[0038] (1) Use the deep convolutional neural network to extract the respective features of the reference image and the image to be segmented

[0039] For the reference image I s and the image to be segmented I q , the present invention uses the ResNet-50 with parameter sharing to extract the respective corresponding features, which are respectively denoted as F s and F q .

[0040](2) Use the contour generation module to generate rough object contours in the image to be segmented

[0041] The contour generation module combines the features of all levels extracted by the deep convolutional neural network, so that the high-level features can guide the low-level features. High-level features imply abstract contour information, while low-level feature...

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 digital image intelligent processing, in particular to a contour-based small-sample semantic segmentation method. The method of the present invention includes: using a deep convolutional neural network to extract the respective features of the reference image and the image to be segmented; using a profile generation module to generate a rougher object profile in the image to be segmented; using a profile optimization module to optimize the rougher object profile to obtain a more Fine contours; use the label average pooling operation to obtain the semantic prototype corresponding to the reference image; use the region average pooling operation to obtain the semantic prototype of the object corresponding to the contour in the image to be segmented; compare the semantic prototypes to determine whether they belong to the same semantic category. Experimental results show that the present invention can generate accurate segmentation maps and effectively solve the problem of small-sample semantic segmentation.

Description

technical field [0001] The invention belongs to the technical field of digital image intelligent processing, and in particular relates to a contour-based small-sample semantic segmentation method. Background technique [0002] Semantic segmentation refers to the use of algorithms to assign a semantic category to each pixel in an image. Small-sample semantic segmentation means that the trained semantic segmentation model can perform accurate semantic segmentation on unseen semantic categories. [0003] Semantic segmentation tasks have important application value in the fields of autonomous driving, robots, and unmanned security. In recent years, the rise of convolutional neural networks has greatly promoted the development of semantic segmentation, and the performance of semantic segmentation has been continuously refreshed on various public data sets. . However, the existing semantic segmentation methods have poor generalization performance when dealing with new scenes and...

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 Patents(China)
IPC IPC(8): G06T7/13G06T7/12G06N3/08G06N3/04
CPCG06T7/13G06T7/12G06N3/08G06N3/045
Inventor 颜波谭伟敏茹港徽李吉春
Owner FUDAN 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