18666613. METHOD FOR FEW-SHOT UNSUPERVISED IMAGE-TO-IMAGE TRANSLATION simplified abstract (NVIDIA CORPORATION)

METHOD FOR FEW-SHOT UNSUPERVISED IMAGE-TO-IMAGE TRANSLATION

Organization Name

NVIDIA CORPORATION

Inventor(s)

Ming-Yu Liu of Redwood City CA (US)

Xun Huang of Pittsburgh PA (US)

Tero Tapani Karras of Helsinki (FI)

Timo Aila of Tuusula (FI)

Jaakko Lehtinen of Helsinki (FI)

METHOD FOR FEW-SHOT UNSUPERVISED IMAGE-TO-IMAGE TRANSLATION - A simplified explanation of the abstract

This abstract first appeared for US patent application 18666613 titled 'METHOD FOR FEW-SHOT UNSUPERVISED IMAGE-TO-IMAGE TRANSLATION

Simplified Explanation

The patent application describes a few-shot, unsupervised image-to-image translation algorithm called FUNIT. This algorithm can translate images of previously unseen target classes using only a few input images of the target type.

Key Features and Innovation

• FUNIT algorithm enables image-to-image translation with minimal input images of the target class.
• The network learns to extract appearance patterns from few input images for translation tasks.
• Generalizable appearance pattern extractor is created for translating images of unseen classes.

Potential Applications

The technology can be used in various applications such as:

• Artistic image editing
• Fashion design
• Virtual try-on applications
• Augmented reality experiences

Problems Solved

The technology addresses the following problems:

• Limited availability of training data for new target classes
• Efficient image translation with minimal input images
• Generalization to unseen classes for image translation tasks

Benefits

• Enables quick and accurate image translation with minimal input
• Increases efficiency in image editing and design tasks
• Enhances creativity in various visual applications

Commercial Applications

• "FUNIT: Few-shot Unsupervised Image-to-Image Translation" can be utilized in industries such as fashion, entertainment, and e-commerce for efficient image editing and design tasks, leading to improved user experiences and increased customer engagement.

Questions about FUNIT

How does the FUNIT algorithm improve image-to-image translation efficiency?

The FUNIT algorithm improves efficiency by learning to extract appearance patterns from a few input images for translation tasks, enabling accurate image translation with minimal input.

What are the potential applications of the FUNIT technology beyond image editing?

The FUNIT technology can be applied in various fields such as fashion design, virtual try-on applications, and augmented reality experiences, expanding its utility beyond traditional image editing tasks.

Original Abstract Submitted

A few-shot, unsupervised image-to-image translation (“FUNIT”) algorithm is disclosed that accepts as input images of previously-unseen target classes. These target classes are specified at inference time by only a few images, such as a single image or a pair of images, of an object of the target type. A FUNIT network can be trained using a data set containing images of many different object classes, in order to translate images from one class to another class by leveraging few input images of the target class. By learning to extract appearance patterns from the few input images for the translation task, the network learns a generalizable appearance pattern extractor that can be applied to images of unseen classes at translation time for a few-shot image-to-image translation task.