17493661. SYSTEM AND METHOD OF CONVOLUTIONAL NEURAL NETWORK simplified abstract (TAIWAN SEMICONDUCTOR MANUFACTURING COMPANY, LTD.)

From WikiPatents
Jump to navigation Jump to search

SYSTEM AND METHOD OF CONVOLUTIONAL NEURAL NETWORK

Organization Name

TAIWAN SEMICONDUCTOR MANUFACTURING COMPANY, LTD.

Inventor(s)

Chao-Tsung Huang of Hsinchu City (TW)

Hsiu-Pin Hsu of New Taipei City (TW)

SYSTEM AND METHOD OF CONVOLUTIONAL NEURAL NETWORK - A simplified explanation of the abstract

This abstract first appeared for US patent application 17493661 titled 'SYSTEM AND METHOD OF CONVOLUTIONAL NEURAL NETWORK

Simplified Explanation

The patent application describes a method and system for generating an output image from an input image using convolutional neural networks (CNN) and non-local operations.

  • The method involves downscaling the input image to create a scaled image.
  • A first CNN modeling process with non-local operations is performed on the scaled image to generate global parameters.
  • A second CNN modeling process with second non-local operations is performed on the input image using the global parameters to generate the output image.

Potential applications of this technology:

  • Image processing and enhancement: The method can be used to improve the quality and resolution of images, making it useful in applications such as image editing, medical imaging, and surveillance systems.
  • Computer vision: The method can be applied to tasks like object recognition, image classification, and scene understanding, improving the accuracy and performance of computer vision systems.

Problems solved by this technology:

  • Limited image quality: The method addresses the problem of low-resolution or poor-quality images by using CNN and non-local operations to enhance the details and overall appearance of the output image.
  • Computational efficiency: By using global parameters and non-local operations, the method reduces the computational complexity compared to traditional CNN approaches, making it more efficient for real-time applications.

Benefits of this technology:

  • Improved image quality: The method enhances the details, sharpness, and overall visual appeal of images, resulting in better image quality for various applications.
  • Faster processing: By utilizing global parameters and non-local operations, the method achieves faster processing times, making it suitable for real-time applications where speed is crucial.
  • Versatility: The method can be applied to a wide range of image processing tasks and computer vision applications, providing a flexible and adaptable solution.


Original Abstract Submitted

A method the following operations: downscaling an input image to generate a scaled image; performing, to the scaled image, a first convolutional neural networks (CNN) modeling process with first non-local operations, to generate global parameters; and performing, to the input image, a second CNN modeling process with second non-local operations that are performed with the global parameters, to generate an output image corresponding to the input image. A system is also disclosed herein.