Invention Title:

GENERATING VIRTUAL HAIRSTYLE USING LATENT SPACE PROJECTORS

Publication number:

US20240221259

Publication date:

2024-07-04

Section:

Physics

Class:

G06T13/40

Inventors:

Menglei Chai Los Angeles, CA, United States

Pavel Savchenkov London, Great Britain (UK)

Aleksandr Belskikh London, Great Britain (UK)

Antoine Chassang London, Great Britain (UK)

Anna Kovalenko London, Great Britain (UK)

Applicant:

Snap Inc. Santa Monica, CA, United States

Drawings (4 of 12)

Drawing 01 for GENERATING VIRTUAL HAIRSTYLE USING LATENT SPACE PROJECTORS

Drawing 02 for GENERATING VIRTUAL HAIRSTYLE USING LATENT SPACE PROJECTORS

Drawing 03 for GENERATING VIRTUAL HAIRSTYLE USING LATENT SPACE PROJECTORS

Drawing 04 for GENERATING VIRTUAL HAIRSTYLE USING LATENT SPACE PROJECTORS

Smart overview of the Invention

The technology focuses on generating realistic virtual hairstyles by utilizing a Generative Adversarial Network (GAN) model. Initially, a first image of a face is created using the GAN model. Subsequently, 3D virtual hair is applied to this image, resulting in a second image featuring the virtual hair. The second image is then projected into a GAN latent space to produce a third image that showcases realistic virtual hair.

Blending and Output Generation

The final step involves blending the realistic virtual hair from the third image with the original face image. This process results in a new image that presents a realistic hairstyle corresponding to the previously applied 3D virtual hair. A trained neural network facilitates this transformation, taking the second image with 3D virtual hair as input and generating an output image with realistic hair.

Background Context

The rise of digital images and advancements in portable computing devices have transformed how users capture and share images. Augmented reality (AR) technology plays a significant role in enhancing user experiences by integrating electronic information with real-world environments. However, processing images under varying conditions remains challenging and resource-intensive.

Applications and User Interaction

The technology can be applied across various domains, including entertainment, gaming, and virtual conversations. It allows for content manipulation in real-time scenarios such as video calls or interactive media sharing. Users can interact with augmented reality content seamlessly, enhancing their engagement with digital images.

System Architecture

The underlying infrastructure supports an interaction system that includes multiple client systems connected over a network. This system enables data exchange between clients and servers, allowing for real-time communication and media sharing. The architecture includes features such as APIs for accessing server functionalities, facilitating user interactions, and managing multimedia content efficiently.