20230146676. PORTRAIT STYLIZATION FRAMEWORK TO CONTROL THE SIMILARITY BETWEEN STYLIZED PORTRAITS AND ORIGINAL PHOTO simplified abstract (Lemon Inc.)
PORTRAIT STYLIZATION FRAMEWORK TO CONTROL THE SIMILARITY BETWEEN STYLIZED PORTRAITS AND ORIGINAL PHOTO
Organization Name
Inventor(s)
Jing Liu of Los Angeles CA (US)
Chunpong Lai of Los Angeles CA (US)
Guoxian Song of Singapore (SG)
Linjie Luo of Los Angeles CA (US)
PORTRAIT STYLIZATION FRAMEWORK TO CONTROL THE SIMILARITY BETWEEN STYLIZED PORTRAITS AND ORIGINAL PHOTO - A simplified explanation of the abstract
This abstract first appeared for US patent application 20230146676 titled 'PORTRAIT STYLIZATION FRAMEWORK TO CONTROL THE SIMILARITY BETWEEN STYLIZED PORTRAITS AND ORIGINAL PHOTO
Simplified Explanation
The patent application describes systems and methods for controlling the similarity between stylized portraits and an original photo. Here are the key points:
- The system receives an input image and encodes it using a variational autoencoder to generate a latent vector.
- The latent vector is blended with latent vectors representing a face in the original user portrait image.
- The resulting blended latent vector is provided to a generative adversarial network (GAN) generator to generate a controlled stylized image.
- The system allows the user to interactively determine the level of stylization vs. personalization in the resulting stylized portrait.
- In some examples, layers of the stylized GAN generator can be swapped with layers of the original GAN generator.
Potential applications of this technology:
- Personalized artwork: Users can create stylized portraits that closely resemble their original photos, allowing for a unique and personalized artistic representation.
- Social media filters: The technology can be used to develop filters that apply stylization to user portraits while maintaining a recognizable likeness.
- Digital avatars: The system can generate stylized portraits that can be used as digital avatars in various applications, such as gaming or virtual reality.
Problems solved by this technology:
- Balancing stylization and personalization: The system allows users to control the level of stylization in their portraits, ensuring that the resulting images maintain a desired resemblance to the original photo.
- Generating controlled stylized images: By blending latent vectors and swapping layers in the GAN generator, the system provides a way to generate stylized portraits while maintaining control over the output.
Benefits of this technology:
- Customization: Users have the ability to determine the level of stylization in their portraits, allowing for personalized and unique artwork.
- User-friendly: The interactive nature of the system enables users to easily adjust the stylization vs. personalization balance according to their preferences.
- Versatility: The technology can be applied to various applications, such as personalized artwork, social media filters, and digital avatars.
Original Abstract Submitted
systems and methods directed to controlling the similarity between stylized portraits and an original photo are described. in examples, an input image is received and encoded using a variational autoencoder to generate a latent vector. the latent vector may be blended with latent vectors that best represent a face in the original user portrait image. the resulting blended latent vector may be provided to a generative adversarial network (gan) generator to generate a controlled stylized image. in examples, one or more layers of the stylized gan generator may be swapped with one or more layers of the original gan generator. accordingly, a user can interactively determine how much stylization vs. personalization should be included in a resulting stylized portrait.
