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

Weak supervised object detection method based on pixel gradient map of specific object

A technology for specific object and target detection, applied in the field of computer vision, which can solve problems such as labeling deviation and affecting training results.

Inactive Publication Date: 2018-12-18
XIAMEN UNIV
View PDF3 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In the process of manually labeling the bounding box, labeling bias will also be introduced, which will affect the training results

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
  • Weak supervised object detection method based on pixel gradient map of specific object
  • Weak supervised object detection method based on pixel gradient map of specific object
  • Weak supervised object detection method based on pixel gradient map of specific object

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0058] The following embodiments will further illustrate the present invention in conjunction with the accompanying drawings.

[0059] First, the symbols mainly used in the present invention are defined. use here Represents an input image in RGB format, y∈{0, 1} K Indicates the category label of the corresponding image, where, Represents the range of real numbers, H and W represent the height and width of the image, respectively, and K represents the number of categories in the dataset. Each image can obtain K specific object pixel gradient maps, such as figure 1 shown. s ∈ [0, 1] K Indicates the category prediction result of the corresponding image, where K indicates the number of categories in the dataset. At the same time, use and Denote the average image of the training dataset and the corresponding classification prediction results, respectively.

[0060] The present invention uses AlexNet (Krizhevsky, Alex, Ilya Sutskever, and GeoffreyE.Hinton."Imagenet clas...

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 a weak supervised object detection method based on a pixel gradient map of a specific object, which relates to a computer vision technology. A pixel gradient map of a specificobject is proposed. During the training process, a pixel gradient map of a particular object is extracted from the image. Based on the pixel gradient map of a particular object, the model can roughlyestimate the shape and position of the target object. The accumulated pixel gradient map of a particular object to conceal the corresponding image is used, and then the concealed data set to fine-tune the model is used, so that the model can find more parts of the object; an average-Maximum value pooling neural network layer, which can greatly help weak monitoring target detection, is proposed. The algorithm does not increase the complexity of the network model, nor does it use additional monitoring information. A large number of experimental results show that the performance of weak surveillance target detection and location is excellent.

Description

technical field [0001] The invention relates to computer vision technology, in particular to a weakly supervised target detection method based on a specific object pixel gradient map. Background technique [0002] Object detection is a basic research topic in the field of computer vision, which mainly includes two different types of detection tasks: object instance detection and object category detection. The goal of the first type of detection task is to identify and localize an object or objects known to be specific in an input image, such as a specific car. This type of task can be considered as a matching problem between the sample set of a specific object and the specific object in the input image to be detected. The difference between the target in the sample set and the input image to be detected is mainly due to the change of imaging conditions. The second category of detection tasks focuses on classifying and locating all possible individuals covered by predefined ...

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
CPCG06N3/084G06V10/473G06V2201/07G06N3/045G06F18/24G06F18/214
Inventor 纪荣嵘沈云航
Owner XIAMEN UNIV
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

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com