18079386. GENERATING LARGE DATASETS OF STYLE-SPECIFIC AND CONTENT-SPECIFIC IMAGES USING GENERATIVE MACHINE-LEARNING MODELS TO MATCH A SMALL SET OF SAMPLE IMAGES simplified abstract (Microsoft Technology Licensing, LLC)

GENERATING LARGE DATASETS OF STYLE-SPECIFIC AND CONTENT-SPECIFIC IMAGES USING GENERATIVE MACHINE-LEARNING MODELS TO MATCH A SMALL SET OF SAMPLE IMAGES

Organization Name

Microsoft Technology Licensing, LLC

Inventor(s)

Maurice Diesendruck of Bellevue WA (US)

Harsh Shrivastava of Redmond WA (US)

GENERATING LARGE DATASETS OF STYLE-SPECIFIC AND CONTENT-SPECIFIC IMAGES USING GENERATIVE MACHINE-LEARNING MODELS TO MATCH A SMALL SET OF SAMPLE IMAGES - A simplified explanation of the abstract

This abstract first appeared for US patent application 18079386 titled 'GENERATING LARGE DATASETS OF STYLE-SPECIFIC AND CONTENT-SPECIFIC IMAGES USING GENERATIVE MACHINE-LEARNING MODELS TO MATCH A SMALL SET OF SAMPLE IMAGES

The present disclosure discusses a system for generating large datasets of style-matching images that match the styles and content of an initial small sample set of input images. This system uses a generative machine-learning model with a selection of style-mixed stored images to synthesize new images that accurately match the style, content, characteristics, and patterns of the input images while also providing added variety and diversity to the dataset.

• Utilizes a generative machine-learning model to produce large datasets of synthesized images
• Conditionally samples synthesized images that match the style and content of the initial sample set
• Provides added variety and diversity to the large image dataset

Potential Applications: - Content creation for marketing and advertising purposes - Style transfer in image editing software - Fashion design and trend forecasting

Problems Solved: - Generating large datasets of style-matching images efficiently - Ensuring synthesized images accurately match the style and content of the input images

Benefits: - Increased efficiency in generating large image datasets - Improved accuracy in matching styles and content - Enhanced variety and diversity in image datasets

Commercial Applications: Title: Style-Matching Image Generation System for Content Creation and Fashion Design This technology can be used in industries such as marketing, advertising, image editing software development, and fashion design to streamline content creation processes and enhance the quality of synthesized images.

Prior Art: Readers can explore prior research on generative machine-learning models, style transfer techniques, and image dataset generation methods to gain a deeper understanding of the background of this technology.

Frequently Updated Research: Stay informed about advancements in generative machine-learning models, style transfer algorithms, and image synthesis techniques to keep up with the latest developments in this field.

Questions about Style-Matching Image Generation System: 1. How does the system ensure that the synthesized images accurately match the style and content of the input images? 2. What are the potential limitations of using a generative machine-learning model for image dataset generation?

Original Abstract Submitted

The present disclosure relates to utilizing a style-matching image generation system to generate large datasets of style-matching images having matching styles and content to an initial small sample set of input images. For example, the style-matching image generation system utilizes a selection of style-mixed stored images with a generative machine-learning model to produce large datasets of synthesized images. Further, the style-matching image generation system utilizes the generative machine-learning model to conditionally sample synthesized images that accurately match the style, content, characteristics, and patterns of the initial small sample set and that also provide added variety and diversity to the large image dataset.