Patents by Inventor Sudhakar SAH

Sudhakar SAH has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).

  • Publication number: 20240256842
    Abstract: A system, method, and computer readable medium for deploying neural networks in low bit environments. The system comprises a runtime platform, a first set of configuration parameters identifying limitations of the runtime platform, and a quantization platform for quantizing neural networks. The quantization platform receives a neural network associated with a framework and quantizing the neural network into a smaller neural network and generates a dataset comprising a second set of configuration parameters for compiling the smaller neural network into instructions for the runtime platform. The second set of configuration parameters are responsive to the limitations of the first set of configuration parameters. The runtime environment implements the smaller neural network in accordance with the second set of configuration parameters.
    Type: Application
    Filed: January 26, 2023
    Publication date: August 1, 2024
    Applicant: Deeplite Inc.
    Inventors: Muhammad Saad ASHFAQ, MohammadHossein ASKARI HEMMAT, Sudhakar SAH, Ehsan SABOORI, Ahmed HASSANIEN, Olivier MASTROPIETRO, Alexander HOFFMAN
  • Publication number: 20240249121
    Abstract: A system, method and computer readable medium for implementing neural networks. The method can include providing a neural network, providing a lookup table based on the neural network, packing weights and activations of the neural network associated with the at least one convolution with a first set of bitwise operations, unpacking the packed weights and activations, with a second set of bitwise operations, to determine one or more inputs for the look up table. The method includes accessing, within the lookup table, an output corresponding to the one or more inputs, and implement the at least one convolution based on the output.
    Type: Application
    Filed: January 19, 2024
    Publication date: July 25, 2024
    Applicant: Deeplite Inc.
    Inventors: Muhammad Saad ASHFAQ, Saptarshi MITRA, Alexander HOFFMAN, Darshan Chandrashekhar GANJI, Ahmed HASSANIEN, Sudhakar SAH, Ehsan SABOORI, MohammadHossein ASKARI HEMMAT
  • Publication number: 20240037702
    Abstract: A system, device and method are provided for generating image processing models for selected hardware. The method, illustratively, includes obtaining a reference model, a desired image resolution based on target hardware, and a training set of images comprising images with the desired image resolution and images with a higher resolution. The method includes generating an updated model by: iteratively training the reference model with a combined set of features, the combined set of features comprising features determined from the images with the higher resolution with at least one stem and features determined from the images with the desired resolution. The method includes outputting the trained updated model to the target hardware to process images with the desired image resolution.
    Type: Application
    Filed: July 25, 2023
    Publication date: February 1, 2024
    Applicant: Deeplite Inc.
    Inventors: Sudhakar SAH, Ivan LAZAREVICH, Ahmadreza JEDDI, Honnesh ROHMETRA, Ehsan SABOORI
  • Patent number: 9430713
    Abstract: According to an aspect of the present invention, there is provided a method of detecting a repetitive pattern. The method includes: clustering a plurality of pixels that form an input image according to color and obtaining one or more color layers composed of pixels included in each cluster; selecting one or more effective layers from the color layers, wherein each of the effective layers includes a predetermined number or more of pixel components, each composed of a plurality of pixels and having a predetermined shape or a predetermined size of area; selecting a unit pattern repeatedly disposed at different locations in each effective layer from the pixel components included in each effective layer; calculating distances between the unit patterns in each effective layer; and calculating a repetition cycle of the unit pattern of the input image based on the calculated distances in each effective layer.
    Type: Grant
    Filed: March 23, 2016
    Date of Patent: August 30, 2016
    Assignee: SAMSUNG SDS CO., LTD.
    Inventors: Sudhakar Sah, Sang Hak Lee, Jong Hang Kim, Seong Jong Ha, Yu Ri Ahn, Yeon Hee Kwon
  • Patent number: 9411659
    Abstract: Provided is a data processing method which can increase data processing speed without adding a new node to a distributed system. The data processing method may include: calculating a conversion number of cores corresponding to a number of processing blocks included in a graphics processing unit (GPU) of a node of a distributed system; calculating a adding up number of cores by adding up a number of cores included in a central processing unit (CPU) of the node of the distributed system and the conversion number of cores; splitting job data allocated to the node of the distributed system into a number of job units data equal to the adding up number of cores; and allocating a number of job units data equal to the number of cores included in the CPU to the CPU of the node of the distributed system and a number of job units data equal to the conversion number of cores to the GPU of the node of the distributed system.
    Type: Grant
    Filed: August 13, 2014
    Date of Patent: August 9, 2016
    Assignee: SAMSUNG SDS CO., LTD.
    Inventor: Sudhakar Sah
  • Publication number: 20160210527
    Abstract: According to an aspect of the present invention, there is provided a method of detecting a repetitive pattern. The method includes: clustering a plurality of pixels that form an input image according to color and obtaining one or more color layers composed of pixels included in each cluster; selecting one or more effective layers from the color layers, wherein each of the effective layers includes a predetermined number or more of pixel components, each composed of a plurality of pixels and having a predetermined shape or a predetermined size of area; selecting a unit pattern repeatedly disposed at different locations in each effective layer from the pixel components included in each effective layer; calculating distances between the unit patterns in each effective layer; and calculating a repetition cycle of the unit pattern of the input image based on the calculated distances in each effective layer.
    Type: Application
    Filed: March 23, 2016
    Publication date: July 21, 2016
    Applicant: SAMSUNG SDS CO., LTD.
    Inventors: Sudhakar SAH, Sang Hak LEE, Jong Hang KIM, Seong Jong HA, Yu Ri AHN, Yeon Hee KWON
  • Patent number: 9330450
    Abstract: According to an aspect of the present invention, there is provided a method of detecting a repetitive pattern. The method includes: clustering a plurality of pixels that form an input image according to color and obtaining one or more color layers composed of pixels included in each cluster; selecting one or more effective layers from the color layers, wherein each of the effective layers includes a predetermined number or more of pixel components, each composed of a plurality of pixels and having a predetermined shape or a predetermined size of area; selecting a unit pattern repeatedly disposed at different locations in each effective layer from the pixel components included in each effective layer; calculating distances between the unit patterns in each effective layer; and calculating a repetition cycle of the unit pattern of the input image based on the calculated distances in each effective layer.
    Type: Grant
    Filed: May 29, 2014
    Date of Patent: May 3, 2016
    Assignee: SAMSUNG SDS CO., LTD.
    Inventors: Sudhakar Sah, Sang Hak Lee, Jong Hang Kim, Seong Jong Ha, Yu Ri Ahn, Yeon Hee Kwon
  • Publication number: 20150052536
    Abstract: Provided is a data processing method which can increase data processing speed without adding a new node to a distributed system. The data processing method may include: calculating a conversion number of cores corresponding to a number of processing blocks included in a graphics processing unit (GPU) of a node of a distributed system; calculating a adding up number of cores by adding up a number of cores included in a central processing unit (CPU) of the node of the distributed system and the conversion number of cores; splitting job data allocated to the node of the distributed system into a number of job units data equal to the adding up number of cores; and allocating a number of job units data equal to the number of cores included in the CPU to the CPU of the node of the distributed system and a number of job units data equal to the conversion number of cores to the GPU of the node of the distributed system.
    Type: Application
    Filed: August 13, 2014
    Publication date: February 19, 2015
    Applicant: SAMSUNG SDS CO., LTD.
    Inventor: Sudhakar SAH
  • Publication number: 20140355873
    Abstract: According to an aspect of the present invention, there is provided a method of detecting a repetitive pattern. The method includes: clustering a plurality of pixels that form an input image according to color and obtaining one or more color layers composed of pixels included in each cluster; selecting one or more effective layers from the color layers, wherein each of the effective layers includes a predetermined number or more of pixel components, each composed of a plurality of pixels and having a predetermined shape or a predetermined size of area; selecting a unit pattern repeatedly disposed at different locations in each effective layer from the pixel components included in each effective layer; calculating distances between the unit patterns in each effective layer; and calculating a repetition cycle of the unit pattern of the input image based on the calculated distances in each effective layer.
    Type: Application
    Filed: May 29, 2014
    Publication date: December 4, 2014
    Applicant: SAMSUNG SDS CO., LTD.
    Inventors: Sudhakar SAH, Sang Hak LEE, Jong Hang KIM, Seong Jong HA, Yu Ri AHN, Yeon Hee KWON