Garment color changing method

Abstract

The invention discloses a garment color changing method. The garment color changing method comprises the following steps: 1, pretreatment work; and 2, generating an adversarial network, wherein in the step 1, the preprocessing work comprises two parts, one part is to preprocess the portrait image, and the other part is to extract a color guide condition from a clothing part in the other portrait image. In the step 1, the portrait image is preprocessed, that is, the garment part is grayed; and in the second step, the aim of changing the color of the clothes is achieved by using the generative adversarial network. The method is based on a clothing color changing framework of a generative adversarial network, firstly, a portrait image is preprocessed, then, a color guiding condition is extracted, the preprocessed portrait image and the extracted color guiding condition are used as input by using the generative adversarial network, and a portrait image after clothing color changing is generated. And finally a relatively real portrait image after color change of the clothes is obtained in continuous adversarial learning of the generative adversarial network.

Description

technical field [0001] The invention relates to the technical field of clothing color changing, in particular to a clothing color changing method. Background technique [0002] In recent years, the fashion field is also a relatively popular scientific research segment, and most of the work focuses on virtual try-on, clothing retrieval and other directions. The clothing color changing task is similar to the virtual try-on task, but the difference is that the virtual try-on needs to completely change the clothing, including patterns, styles, colors, etc., while the virtual try-on only changes the color of the clothing. Clothing color changing task can also be regarded as an image colorization task in the specific domain of fashion, and existing general image colorization methods usually cannot be directly applied to this task. [0003] Han et al. (Han X, Wu Z, Wu Z, et al.Viton: An image-based virtual try-onnetwork[C] / / Proceedings of the IEEE conference on computer vision and...

AI Technical Summary

Problems solved by technology

User-provided guidance proposed by Zhang et al. (Zhang R Y, Zhu J Y, Isola P, et al. Real-time user-guided imagecolorization with learned deep priors [J]. ACM Transactions on Graphics, 2017, 36(4): 119) The image colorization method of is a general image colorization method, which can be applied to this scene, but due to the limitation of the amount of data in the clothing dataset, the final color change effect is poor

Images