Image-based Virtual Try-on Systems with Wavelet Transform

Virtual try-on is a clothing technology that allows users to try on clothes virtually without physically wearing them. This is done by digitally overlaying clothing on the user’s body using various methods such as augmented reality, virtual reality and 3D scanning. Virtual try-on has the potential to revolutionize the fashion industry by reducing returns, improving customer experience and increasing sales. There are several major challenges in the field of image-based clothing virtual try on, including: the variety of human poses and the misalignment between clothes and poses, the existence of suitable datasets, and the existence of images with different resolutions for virtual try-on; It is also important to note that the data must be of high quality and represent the real world. In this research, to overcome the challenge of images with different resolutions, we presented an architecture named Appearance Flow with Wavelet(AFW), which is an extension of the Appearance Flow network; In this architecture, a wavelet transform function and the definition of the loss function in its objective function are used to effectively manage the differences in resolution and make the network more resistant to input images with different resolutions and appropriate warping of clothing, according to the person. The experimental results show that these changes lead to improvement in the way the clothes are warped and determine the shape and size of the clothes that are suitable for the desired person’s body, and also cause the network to be resistant to the difference in resolution in the input images.