Patents by Inventor Yu-Chiang Wang

Yu-Chiang Wang 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: 20250252301
    Abstract: Embodiments of the present disclosure relate to fine-tuning a neural network model using a weight-decomposed low-rank adaptation (DoRA). DoRA reduces the number of parameters that are fine-tuned, thereby reducing memory and the time needed to fine-tune the parameters. the Pre-trained weights are decomposed into two components, magnitude and direction, which are separately fine-tuned. The magnitude components are fine-tuned while the direction components remain unchanged (frozen). Then low-rank adaptation (LoRA) is used to fine-tune the direction components, efficiently minimizing the number of trainable parameters. Compared with using LoRA to fine-tune the weights directly, using DoRA exhibits a closer resemblance to full fine-tuning's learning behavior and improves upon LoRA in commonsense reasoning and visual instruction tuning tasks. By employing DoRA, both the learning capacity and training stability of LoRA is enhanced.
    Type: Application
    Filed: August 16, 2024
    Publication date: August 7, 2025
    Inventors: Min-Hung Chen, Shih-Yang Liu, Chien-Yi Wang, Hongxu Yin, Pavlo Molchanov, Yu-Chiang Wang
  • Publication number: 20250239093
    Abstract: Semantic segmentation generally refers to a machine learning process that associates a label or category with every pixel in an image. This can be used to recognize a collection of pixels that form distinct categories of objects, which may have applications in autonomous driving for example where the vehicle needs to identify other vehicles, pedestrians, traffic signs, pavement, and other road features from captured images of a surrounding environment. While using pixel-level annotations may be ideal for fully-supervised training of the semantic segmentation model, collecting such annotations is time-consuming and expensive, and therefore limits the scalability and practicality of fully-supervised training methods. The present disclosure enables weakly supervised training of a semantic segmentation model aided by learned prompts embedded with semantic knowledge.
    Type: Application
    Filed: August 13, 2024
    Publication date: July 24, 2025
    Inventors: Min-Hung Chen, Ci-Siang Lin, Chien-Yi Wang, Yu-Chiang Wang
  • Patent number: 11748629
    Abstract: A computing device for handling anomaly detection, comprises an encoder, for receiving an input image, to generate a first latent vector comprising a semantic latent vector and a visual appearance latent vector according to the input image and at least one first parameter of the encoder; and a training module, coupled to the encoder, for receiving the input image and the first latent vector, to update the at least one first parameter according to the input image and the first latent vector and a loss function.
    Type: Grant
    Filed: May 6, 2020
    Date of Patent: September 5, 2023
    Assignee: Moxa Inc.
    Inventors: Wei-Yu Lee, Yu-Chiang Wang
  • Publication number: 20220405634
    Abstract: A learning module for handling classification tasks, configured to perform the following instructions: receiving a first plurality of parameters from a training module; and generating a first loss of a first task in a first domain and a second loss of a second task in a second domain according to the first plurality of parameters.
    Type: Application
    Filed: December 29, 2021
    Publication date: December 22, 2022
    Applicant: Moxa Inc.
    Inventors: Wei-Yu Lee, Jheng-Yu Wang, Yu-Chiang Wang
  • Publication number: 20210224606
    Abstract: A computing device for handling anomaly detection, comprises an encoder, for receiving an input image, to generate a first latent vector comprising a semantic latent vector and a visual appearance latent vector according to the input image and at least one first parameter of the encoder; and a training module, coupled to the encoder, for receiving the input image and the first latent vector, to update the at least one first parameter according to the input image and the first latent vector and a loss function.
    Type: Application
    Filed: May 6, 2020
    Publication date: July 22, 2021
    Inventors: Wei-Yu Lee, Yu-Chiang Wang
  • Patent number: 11010871
    Abstract: A computing device for handling image super-resolution (ISR), comprises a generator module, for receiving at least one input image, to generate an output image according to at least one first parameter and a first plurality of feature maps generated by at least one first channel attention (CA); a discriminator module, for receiving the output image and a high resolution (HR) image, to generate a second plurality of feature maps and a third plurality of feature maps by at least one second CA, and to generate at least one score according to the second plurality of feature maps, the third plurality of feature maps and at least one second parameter; and a feedback module, for receiving the at least one score, to update the at least one first parameter and the at least one second parameter according to the at least one score and an objective function.
    Type: Grant
    Filed: February 20, 2020
    Date of Patent: May 18, 2021
    Assignee: Moxa Inc.
    Inventors: Wei-Yu Lee, Po-Yu Chuang, Yu-Chiang Wang
  • Publication number: 20210133925
    Abstract: A computing device for handling image super-resolution (ISR), comprises a generator module, for receiving at least one input image, to generate an output image according to at least one first parameter and a first plurality of feature maps generated by at least one first channel attention (CA); a discriminator module, for receiving the output image and a high resolution (HR) image, to generate a second plurality of feature maps and a third plurality of feature maps by at least one second CA, and to generate at least one score according to the second plurality of feature maps, the third plurality of feature maps and at least one second parameter; and a feedback module, for receiving the at least one score, to update the at least one first parameter and the at least one second parameter according to the at least one score and an objective function.
    Type: Application
    Filed: February 20, 2020
    Publication date: May 6, 2021
    Inventors: Wei-Yu Lee, Po-Yu Chuang, Yu-Chiang Wang
  • Publication number: 20140072177
    Abstract: A method for identifying vehicle license plate, comprising: mounting a server apparatus, and establishing a vehicle database and a vehicle license plate identification library in the server apparatus so as to generate a vehicle data and a vehicle license plate identification program; establishing a communication link, and receiving the vehicle data and the vehicle license plate identification program by the mobile device, and installing the vehicle license plate identification program and storing the vehicle data; capturing plate number of a vehicle by a camera module mounted in the mobile device for generating an image and then analyzing the plate number in the image by an analyzing program to generate an analysis result; comparing the analysis result with the vehicle data by a comparison module of the mobile device for generating a comparison result; and displaying the comparison result.
    Type: Application
    Filed: September 12, 2012
    Publication date: March 13, 2014
    Inventors: PEI-YUAN CHOU, Shang-En Wu, Yu-Chiang Wang