Multi-classifier combined weak annotation image object detection method

A multi-classifier and image object technology, applied in the field of image processing and computer vision, can solve problems that affect the detection effect

Active Publication Date: 2015-12-30
BEIHANG UNIV
View PDF4 Cites 23 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Object detection technology based on weak annotation has a wide range of applications in large-scale image data processing, but when analyzing weakly labeled samples, sample noise information will greatly affect the final detection effect

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
  • Multi-classifier combined weak annotation image object detection method
  • Multi-classifier combined weak annotation image object detection method
  • Multi-classifier combined weak annotation image object detection method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0031] The present invention proposes a multi-classifier combined weakly labeled image object detection method, which includes the following steps:

[0032] (1) Given a multi-category training picture set containing m categories, each category is given a category label, defined as L={L 1 , L 2 ,...,L i} (i=1,2,...,M). Each category contains N i images (i=1,2,...,M).

[0033] (2) Use the objectness detection method to perform objectness analysis (ObjectnessMeasure) on all images, and generate K candidate regions for each image.

[0034] (3) For each area block, use the trained convolutional neural network (CNN) model on the ImageNet dataset to extract fc7-layer 4096-dimensional features for each area.

[0035] (4) Use a clustering algorithm (such as the Kmeans algorithm) to cluster the target area features generated by each category, and get C for the i-th category i =[N i / 100] cluster centers and clustering results for all regional features.

[0036] (5) According to th...

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 provides a multi-classifier combined weak annotation image object detection method. The method comprises the steps that M weak annotation image sets with different labels are input, objective analysis is conducted on all images in the image sets, and an objective region set is generated; image features are generated for the region set, and clustering is conducted on different label feature sets; a middle region classifier is trained for each clustered region set according to a clustering result; a category attribute is calculated for each classifier; a test image is input, objective analysis is conducted to obtain a region block, and region features are generated. A multi-classifier is used for conducting combined detection, and the region containing the category object is judged. The multi-classifier combined weak annotation image object detection method has good performance on the aspect of multi-category image object combined detection and can be applied to fields of image object automatic annotation, image object identification and the like.

Description

technical field [0001] The invention belongs to the technical fields of image processing and computer vision, and relates to a weakly labeled image object detection method combined with multiple classifiers. Background technique [0002] 1. Object detection technology based on weak labeling mainly considers how to use simple labeling information and a large number of unlabeled samples for training and classification. On the basis of low cost, it can better use a large amount of data to obtain relatively good recognition results . In 2010, Alexe et al. proposed the concept of image objectivity, using salient foreground analysis and other methods to extract areas that may contain objects on images without any annotations. This method takes into account the color contrast (ColorContrast), edge density (EdgeDensity ) and SuperpixelsStraddling. [0003] In 2012, Thomas et al. proposed a weakly supervised learning and positioning method based on attribute knowledge. This method ...

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): G06K9/62
CPCG06F18/23213
Inventor 李甲陈小武张宇赵沁平王晨
Owner BEIHANG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap