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

Abnormal image detection method combining attention mechanism and information entropy minimization

A technology of abnormal images and detection methods, applied in neural learning methods, computer components, instruments, etc., can solve the problem that abnormal samples do not have good discrimination ability, and achieve good generalization ability, less information redundancy, and strong generalization The effect of the ability

Pending Publication Date: 2020-04-28
ZHEJIANG UNIV OF TECH
View PDF7 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] However, these deep neural network models still face a problem: the models trained only by normal samples not only have good generalization ability to normal samples, but also have a certain generalization ability to abnormal samples, which leads to the inference of the applied model. Abnormal samples do not have good discrimination ability

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
  • Abnormal image detection method combining attention mechanism and information entropy minimization
  • Abnormal image detection method combining attention mechanism and information entropy minimization
  • Abnormal image detection method combining attention mechanism and information entropy minimization

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0031] In order to make the present invention easier to understand and its advantages clearer, the technical solutions in the embodiments of the present invention will be described in detail below in conjunction with the drawings and specific embodiments.

[0032] (1) In order to verify the accuracy of the present invention on abnormal image detection tasks, it is now verified by three commonly used data sets: COIL100, MNIST, and CIFAR10. The following explains the content and image size of each dataset:

[0033] COIL-100: This data set is a collection of natural pictures, including shooting 100 objects from different angles, and taking an image every 5 degrees. Each object contains 72 images, and the image size is 128*128.

[0034] MNIST: This data set is a classic data set, which is often used as an anomaly detection data set. It is a database composed of 10 categories of handwritten digits, with 6000 pictures in each category, and the size of each image is 28*28.

[0035] ...

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 attention mechanism and information entropy minimization combined abnormal image detection method, which comprises the following steps of: (1) selecting and integrating a current mainstream data set as a training data set; (2) selecting a deep neural network for novelty detection, and improving the structure of a mainstream detection model; (3) preprocessing the sampleimage to be suitable for network input; (4) training a detection model; and (5) detecting by using the trained detection model to obtain a detection result. The invention provides an abnormal image detection method combining an attention mechanism and information entropy minimization. By making a certain improvement on a mainstream method, the model has better representation on normal sample characteristics after being trained, the sensitivity to abnormal samples is relatively high, and the discrimination capability of novelty detection on normal samples and abnormal samples is improved.

Description

technical field [0001] The present invention proposes an abnormal image detection method combining attention mechanism and information entropy minimization, which has a stronger ability to identify new or unknown samples. Background technique [0002] Abnormal image detection has a wide range of applications. The abnormal image detection task requires the model to have the ability to identify new or unknown samples. The key to abnormal image detection is to make the model have good expressive ability for normal samples and maintain a certain sensitivity for abnormal samples. In general, both capabilities of the model should be enhanced during training. A common assumption in abnormal image detection is that abnormal samples not only have characteristic differences from normal samples in high-dimensional data space, but also in low-dimensional data space. Therefore, the key to the abnormal image detection task becomes how to better reconstruct the high-dimensional data spa...

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): G06K9/62G06N3/04G06N3/08
CPCG06N3/082G06N3/045G06F18/2411G06F18/214
Inventor 郭东岩吴宇鹏田苗邵燕燕张剑华陈胜勇
Owner ZHEJIANG UNIV OF TECH
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