Clothing key point detection method and device, electronic equipment, storage medium and program

Abstract

The embodiment of the present invention discloses a method and device for detecting key points of clothing, electronic equipment, a storage medium, and a program, wherein the method includes: extracting features from images through multiple sets of convolutional neural networks to obtain multiple sets of features; , the size of the convolution kernel of each group of convolutional neural networks in the multiple groups of convolutional neural networks is different; based on the multiple groups of features, the clothing key point detection is performed on the image. The embodiment of the present invention can be effectively applied to the detection of key points of clothing with different scales, solve the problem of scale changes in clothing images caused by complex changes in shooting angles, human body postures, clothing styles, and placement methods, and reduce the difficulty of key point detection. Improve the accuracy of keypoint detection.

Description

technical field [0001] The invention relates to computer vision technology, especially a method and device for detecting key points of clothing, electronic equipment, storage media and programs. Background technique [0002] Clothing recognition is one of the most important and challenging problems in image retrieval. On the Internet today, most user searches are related to online shopping content and clothing. Therefore, clothing recognition is the key problem in solving the same style retrieval, style recognition and clothing recommendation requirements. However, clothing recognition is very difficult. First, clothing itself is a very flexible object due to the large deformation of clothing. The clothing will show different states when the clothing style, body posture, and placement are different; Clothing pictures taken under different conditions and shooting angles are also very different. For example, the difference between a model's posing for a photo and a consumer'...

AI Technical Summary

Problems solved by technology

However, clothing recognition is very difficult. First, clothing itself is a very flexible object due to the large deformation of clothing. The clothing will show different states when the clothing style, body posture, and placement are different; There is also a huge gap between clothing pictures taken under different conditions and shooting angles. For example, the difference between a model’s posing for a photo and a consumer’s selfie is obvious.

Images