Automatic garment image labeling method

Abstract

The invention discloses an automatic garment image labeling method. The method comprises the steps of 1, extracting a salient region of an image; 2, extracting an SIFT feature of the image, obtaininga visual vocabulary by utilizing K-mean clustering, and performing different weighted operations according to whether the SIFT feature of the training image is located in the salient region or not toobtain a bag-of-word representation of the visual vocabulary; and 3, training a classification model by using a support vector machine to realize image classification and labeling. Compared with the prior art, the method has the following advantages and effects that a large amount of labor is liberated to a great extent; the color of a garment can be directly positioned; the data labeling accuracycan be improved; and wrong labeling caused by subjective factors can be reduced.

Description

technical field [0001] The invention relates to an automatic tagging method for clothes images. Background technique [0002] At present, it mainly relies on a large amount of manual labor to label clothes. This labeling process has the following problems: [0003] 1. Due to the common understanding gap between the underlying visual features and high-level semantic concepts of clothing images, manual labeling of clothing images has been difficult to meet user needs; [0004] 2. The entire labeling process is very complicated. When the amount of data is particularly large, it will require a lot of manual labor. Not only is it time-consuming and laborious, but it also has a lot of subjectivity; [0005] 3. During the labeling process of clothing data, due to factors such as the limited energy of the labeler or the subjectivity of the labeler, the 100% accuracy of the labeling data cannot be guaranteed, that is, the quality of the labeling cannot be judged, and manual supervis...

AI Technical Summary

Problems solved by technology

[0003] 1. Due to the common understanding gap between the underlying visual features and high-level semantic concepts of clothing images, manual labeling of clothing images has been difficult to meet user needs;

[0004] 2. The entire labeling process is very complicated. When the amount of data is particularly large, it will require a lot of manual labor. Not only is it time-consuming and laborious, but it also has a lot of subjectivity;

[0005] 3. In the process of labeling clothing data, due to factors such as the limited energy of the labeler or the subjectivity of the labeler, the 100% accuracy of the labeling data cannot be guaranteed, that is, the quality of the labeling cannot be judged, and manual supervision is required

[0006] Therefore, the labeling process of a large amount of data is a very time-consuming and difficult to guarantee the accuracy of the task