Unlock instant, AI-driven research and patent intelligence for your innovation.

Image entity extraction method based on mask R-CNN and GAN

A technology of entity extraction and image, applied in neural learning methods, instruments, biological neural network models, etc., can solve the problems of not clustering together, empty detection results, etc., to achieve accurate object detection results and optimize network parameters.

Active Publication Date: 2022-04-19
BEIJING INSTITUTE OF TECHNOLOGYGY +1
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

And only considering whether the classification of each pixel is correct may cause the correctly classified pixels to not get together, making the final detection result appear empty.

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 entity extraction method based on mask R-CNN and GAN
  • Image entity extraction method based on mask R-CNN and GAN
  • Image entity extraction method based on mask R-CNN and GAN

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0031] The specific implementation scenarios and procedures are as follows: figure 1 shown. First collect training data for the object to be detected, train the MaskR-CNN network, and obtain the network model (corresponding to step 1 in the content of the invention); use the mask branch in Mask R-CNN as a generator, and add a discriminator to replace the original Mask R -The cross-entropy loss in the CNN network forms a generative confrontation network (corresponding to step 2 in the summary of the invention), wherein the generator such as image 3 As shown, removing the classification branch and bounding box regression branch in Mask R-CNN gives a generator based on Mask R-CNN, and the discriminator is as follows figure 2 As shown, it contains multiple hierarchical structures composed of convolutional layers, activation functions, regularization layers, and a fully connected layer; the network parameters are further optimized according to the training method of the generati...

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 relates to an image entity extraction method based on Mask R-CNN and GAN, and belongs to the technical field of computer vision and object detection. This method replaces the cross-entropy loss calculation part in Mask R-CNN with the generation confrontation network GAN, uses the part of the network that generates the mask as the generator, adds a discriminator, and classifies the mask and the true value generated by the generator. Through adversarial learning, the network's marking of object pixels is more in line with the true value, and more accurate marking results are obtained. The method adopts the Mask R-CNN network for training, obtains the network parameters as initial parameters, increases the discriminator to replace the cross-entropy loss in the Mask R-CNN network, and further optimizes the network parameters, so that the network can more accurately classify objects belonging to Pixels are marked to get more accurate object detection results.

Description

technical field [0001] The invention relates to an image entity extraction method, in particular to an image entity extraction method based on Mask R-CNN and GAN, and belongs to the technical field of computer vision and object detection. Background technique [0002] Image entity extraction refers to the automatic detection of entities of interest from images. Object detection methods in computer vision can automatically extract entities in images. Mask R-CNN is an object detection method based on deep learning, which can automatically obtain the area of ​​each object in the image. Mask-RCNN is a two-stage network, the first stage scans the image and generates candidate regions, and the second stage classifies the candidate regions and generates bounding boxes and masks. [0003] The input of Mask-RCNN is an image, and the output includes three branches, one branch outputs the category label, that is, the category of the object contained in the image; one branch outputs t...

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): G06V10/764G06V10/82G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/241
Inventor 闫斌李吟孙正晨裴明涛江帆张峻玮
Owner BEIJING INSTITUTE OF TECHNOLOGYGY