Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Image segmentation method and device, equipment and storage medium

An image segmentation and image technology, applied in the field of artificial intelligence, can solve the problems of image difference, excessive labor and cost requirements, and low efficiency, and achieve the effect of reducing labor and cost, and improving the efficiency and accuracy of image segmentation.

Pending Publication Date: 2021-11-26
TENCENT TECH (SHENZHEN) CO LTD
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the real scene is complex and diverse, such as weather, lighting, etc. will cause differences in the images of the same object
Collecting images in all scenes and giving corresponding fine annotations requires too much manpower and cost, and the efficiency is low

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 segmentation method and device, equipment and storage medium
  • Image segmentation method and device, equipment and storage medium
  • Image segmentation method and device, equipment and storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment approach 1

[0139] Embodiment 1: A discriminator is used to determine a first probability that a source-domain predicted sample image is a source-domain image, and a second probability that a target-domain predicted sample image is a target-domain image. Then, based on the first probability and the second probability, determine the adversarial learning loss value between the source domain image segmentation network to be trained, the target domain image segmentation network to be trained, and the discriminator. Based on the source label and the source predicted label, the segmentation loss value of the source domain image segmentation network to be trained is determined. Based on the adversarial learning loss value and the segmentation loss value, the target loss value is obtained.

[0140] Specifically, the neural network architecture involved in the training process is as Figure 7 As shown, it includes a source domain image segmentation network 501 to be trained, a target domain image...

Embodiment approach 2

[0145] Embodiment 2: A discriminator is used to determine the first probability that the predicted sample image in the source domain is an image in the source domain, and the second probability that the predicted sample image in the target domain is an image in the target domain. Then, based on the first probability and the second probability, determine the adversarial learning loss value between the source domain image segmentation network to be trained, the target domain image segmentation network to be trained, and the discriminator. Based on the source label and the source predicted label, the segmentation loss value of the source domain image segmentation network to be trained is determined. Based on the network parameters of the source domain classifier to be trained and the network parameters of the target domain classifier to be trained, the distance loss value between the source domain classifier to be trained and the target domain classifier to be trained is determine...

Embodiment approach 3

[0149] Embodiment 3: Determine the maximum average difference between the predicted sample image in the source domain and the predicted sample image in the target domain. Then, based on the source labels and source prediction labels, the segmentation loss value of the source domain image segmentation network to be trained is determined. Based on the adversarial learning loss value and the segmentation loss value, the target loss value is obtained.

[0150]Specifically, through the Maximum Mean Discrepancy (MMD) algorithm, the maximum mean difference between the predicted sample image and the target domain predicted sample image is determined, and the maximum mean difference is used to measure the distance between two different but related distributions . Based on the difference between the source label and the source prediction label, the segmentation loss value of the source domain image segmentation network to be trained is determined, and then the weighted sum between the ...

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 embodiment of the invention provides an image segmentation method and device, equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: extracting target image features of a target domain to-be-processed image through a trained target domain image segmentation network, obtaining a target domain prediction image based on the target image features, wherein the target domain prediction image comprises a target object with a target label, and the trained target domain image segmentation network is obtained by performing joint iterative training on a to-be-trained target domain image segmentation network and a to-be-trained source domain image segmentation network based on a domain transfer learning mode, and the source domain sample image comprises a source domain sample object with a source label. Through the domain transfer learning mode, the target domain image segmentation network used for segmenting the target domain image is obtained under the condition that each pixel in the target domain sample image does not need to be subjected to refined labeling, so that the manpower and the cost are reduced, and the image segmentation efficiency and the image segmentation accuracy are improved.

Description

technical field [0001] The embodiments of the present invention relate to the technical field of artificial intelligence, and in particular to an image segmentation method, device, equipment and storage medium. Background technique [0002] Image segmentation is a fundamental problem in video and image understanding. The image segmentation process is as follows: after the image x is input into the image segmentation model, the image segmentation model predicts the category of each pixel in the image x, and then according to the category of each pixel, the image x is segmented to obtain the predicted image y. [0003] Since the image segmentation model achieves image segmentation by predicting the category of each pixel in the image, it is also necessary to manually fine-tune each pixel in the sample image when training the image segmentation model. However, complex and diverse real scenes, such as weather and lighting, will cause differences in images of the same object. C...

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): G06T7/11G06K9/62G06N3/04G06N3/08
CPCG06T7/11G06N3/04G06N3/08G06T2207/20076G06T2207/20081G06T2207/20084G06F18/24G06F18/214
Inventor 张启明李志锋刘威
Owner TENCENT TECH (SHENZHEN) CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products