How It Works
This demo offers reconstruction of an existing face into a new AI-generated face via deep learning technology.In Detail:
- U-Net architecture based deep learning network has been developed and trained using a labeled dataset comprised of hundreds of thousands of face images of varying image quality and demographics (age, gender, ethnicity, etc.). Further data collection and training is in progress to enhance the realistic facial construction.
- As a face is detected and categorized by size and shape, one of the many 'Masking Methods' developed on our own will be chosen and applied to optimize the facial replacement.
- An existing face is replaced seamlessly with a generated face, matching the facial features and shades, via Inpainting technique.
✔ When a press should use photos that contain people who did not give consent to release
✔ When you wish to share a picture to social media which contains faces of other people taken unexpectedly
✔ For online sellers without a professional model who had to mask faces or cut them out from the pictures to describe the goods
✔ To ensure the privacy of customers and/or workers from on- and off-line advertisement pictures for small businesses
✔ To protect privacy while posting your photos freely to on-line boards and/or community sites
This demo can be utilized in a wide range of applications that are conventionally limited by Portrait Rights. Our service will be officially released in the first half of 2020, optimized for the aforementioned applications.