17949517. METHODS AND SYSTEMS OF GENERATING IMAGES UTILIZING MACHINE LEARNING AND EXISTING IMAGES WITH DISENTANGLED CONTENT AND STYLE ENCODING simplified abstract (Robert Bosch GmbH)

METHODS AND SYSTEMS OF GENERATING IMAGES UTILIZING MACHINE LEARNING AND EXISTING IMAGES WITH DISENTANGLED CONTENT AND STYLE ENCODING

Organization Name

Robert Bosch GmbH

Inventor(s)

Mansur Arief of Verona PA (US)

Ji Eun Kim of Pittsburgh PA (US)

Shashank Shekhar of Ranchi (IN)

Mohammad Sadegh Norouzzadeh of Pittsburgh PA (US)

Ding Zhao of Pittsburgh PA (US)

METHODS AND SYSTEMS OF GENERATING IMAGES UTILIZING MACHINE LEARNING AND EXISTING IMAGES WITH DISENTANGLED CONTENT AND STYLE ENCODING - A simplified explanation of the abstract

This abstract first appeared for US patent application 17949517 titled 'METHODS AND SYSTEMS OF GENERATING IMAGES UTILIZING MACHINE LEARNING AND EXISTING IMAGES WITH DISENTANGLED CONTENT AND STYLE ENCODING

Simplified Explanation

The abstract describes systems and methods for generating new images for training a machine-learning model by altering the style of captured images and utilizing latent spaces to generate new images.

• Image data is produced from an image captured by an image sensor.
• The style of the image is altered (e.g., color, shading, orientation).
• Altered image data is encoded into a first latent space.
• An image from a database is selected based on similarity to the altered image and decoding of the first latent space.
• Style encodings of the first latent space are extracted to classify the style of the altered image data in a second latent space.
• New images are generated using a reconstructor model that combines the two latent spaces.

Potential Applications

This technology can be applied in image recognition, computer vision, and artificial intelligence research.

Problems Solved

This technology solves the problem of generating diverse training data for machine-learning models by altering image styles.

Benefits

The benefits of this technology include improved model training, increased accuracy in image recognition tasks, and enhanced creativity in image generation.

Potential Commercial Applications

One potential commercial application of this technology is in developing advanced image recognition systems for industries such as healthcare, security, and autonomous vehicles.

Possible Prior Art

Prior art may include similar methods for image style alteration and generation using latent spaces in machine learning and computer vision research.

How does this technology compare to existing image style transfer methods in terms of efficiency and accuracy?

This article does not provide a direct comparison with existing image style transfer methods, leaving the question of efficiency and accuracy differences unanswered.

What are the potential limitations or challenges in implementing this technology in real-world applications?

The article does not address potential limitations or challenges in implementing this technology, leaving this question unanswered.

Original Abstract Submitted

Systems and methods for generating new images for training a machine-learning model are disclosed. Image data is produced regarding an image captured by an image sensor. The image data is altered such that the style of the image (e.g., color, shading, orientation, etc.) is altered. The altered image data is encoded into a first latent space. An image from a database is selected based on its similarity to the altered image and a decoding of the first latent space. Style encodings of the first latent space are extracted to classify a style of the altered image data in a second latent space. New images are then generated utilizing a reconstructor model that combines the two latent spaces. These new images can be used to train an image-recognition model.