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

An anti-disturbance generation method and device for an object detection model

An object detection and model technology, applied in the computer field, can solve problems such as poor applicability, low efficiency, and large amount of data processing, and achieve the effect of improving generation efficiency and applicability

Pending Publication Date: 2019-06-18
HUAWEI TECH CO LTD +1
View PDF4 Cites 18 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the adversarial perturbation generated by the existing adversarial perturbation generation method is not universal, that is, multiple adversarial perturbations will be generated for one object detection algorithm, which makes the existing adversarial perturbation generation method a large amount of data processing and low efficiency , poor applicability

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
  • An anti-disturbance generation method and device for an object detection model
  • An anti-disturbance generation method and device for an object detection model
  • An anti-disturbance generation method and device for an object detection model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0026] See figure 1 , figure 1 is a schematic flowchart of a method for generating an adversarial disturbance of an object detection model provided by an embodiment of the present invention. For the convenience of understanding and description, the embodiment of the present invention takes an object detection model in the field of image processing as an example to describe its anti-disturbance generation method in detail. The target object described in the embodiment of the present invention may be an object such as a person, an animal, or a building included in the image to be detected. It should be noted that the first, second, and i-th before the set of adversarial disturbances, adversarial samples, and target object confidences described in this embodiment are only used to distinguish the adversarial disturbances and countermeasures corresponding to different adversarial disturbance correction processes. Confidence sets of samples or target objects, without other restric...

Embodiment 2

[0050] See figure 2 , figure 2 It is a structural schematic diagram of an object detection model anti-disturbance generation device provided by an embodiment of the present invention. The above-mentioned anti-disturbance generation device includes:

[0051] The acquiring unit 10 is configured to acquire a first adversarial perturbation and a first training sample set, where the first training sample set includes N training samples.

[0052] An adversarial disturbance correction unit 20, configured to determine a first adversarial sample based on the first training sample in the first training sample set obtained by the acquisition unit 10 and the first adversarial disturbance, and determine the first adversarial sample based on the object detection model corresponding first target object confidence set, and perform a first counter-perturbation correction on the first counter-perturbation according to the first target-object confidence set, so as to obtain a second counter-...

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 discloses an anti-disturbance generation method and device for an object detection model. The method comprises the following steps: acquiring a first anti-disturbance and a first training sample set; And determining a first confrontation sample according to the first training sample and the first confrontation disturbance in the first training sample set, and performing first confrontation disturbance correction on the first confrontation disturbance based on a first target object confidence set corresponding to the first confrontation sample determined by the object detection model to obtain a second confrontation disturbance. After a new training sample is obtained from the first training sample set each time, new anti-disturbance is obtained through correction again based on the new training sample and the anti-disturbance obtained through last anti-disturbance correction. And when the N confrontation disturbances obtained by N times of correction areconverged, determining the (N + 1) th confrontation disturbance obtained by the Nth confrontation disturbance correction as the target confrontation disturbance corresponding to the object detectionmodel. By adopting the embodiment of the invention, the efficiency and the applicability of the anti-disturbance generation method can be improved.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a method and device for generating an anti-disturbance of an object detection model. Background technique [0002] With the continuous development of computer technology and deep learning technology, as a basic and very important research focus in the field of artificial intelligence and computer vision, the development of object detection technology cannot be ignored. Under the existing environment, object detection technology can already be applied in some fields. For example, in automatic driving or intelligent video surveillance, object detection technology has great practical value. Moreover, it is foreseeable that the application field of object detection technology will continue to expand, so people are paying more and more attention to issues such as the stability and security of object detection algorithms used in object detection technology. [0003] In the prior art...

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): G06K9/62
Inventor 徐名源姚春凤冯柏岚黄凯奇张俊格陈晓棠李德榜
Owner HUAWEI TECH 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