DOE Patents title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Uncertainty-refined image segmentation under domain shift

Abstract

A method for digital image segmentation is provided. The method comprises training a neural network for image segmentation with a labeled training dataset from a first domain, wherein a subset of nodes in the neural net are dropped out during training. The neural network receives image data from a second, different domain. A vector of N values that sum to 1 is calculated for each image element, wherein each value represents an image segmentation class. A label is assigned to each image element according to the class with the highest value in the vector. Multiple inferences are performed with active dropout layers for each image element, and an uncertainty value is generated for each image element. The label of any image element with an uncertainty value above a predefined threshold is replaced with a new label corresponding to the class with the next highest value.

Inventors:
; ; ; ; ; ; ;
Issue Date:
Research Org.:
Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
Sponsoring Org.:
USDOE
OSTI Identifier:
1924873
Patent Number(s):
11379991
Application Number:
16/887,311
Assignee:
National Technology & Engineering Solutions of Sandia, LLC (Albuquerque, NM)
Patent Classifications (CPCs):
G - PHYSICS G06 - COMPUTING G06F - ELECTRIC DIGITAL DATA PROCESSING
G - PHYSICS G06 - COMPUTING G06N - COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
DOE Contract Number:  
NA0003525
Resource Type:
Patent
Resource Relation:
Patent File Date: 05/29/2020
Country of Publication:
United States
Language:
English

Citation Formats

Martinez, Carianne, Potter, Kevin Matthew, Donahue, Emily, Smith, Matthew David, Snider, Charles J., Korbin, John P., Roberts, Scott Alan, and Collins, Lincoln. Uncertainty-refined image segmentation under domain shift. United States: N. p., 2022. Web.
Martinez, Carianne, Potter, Kevin Matthew, Donahue, Emily, Smith, Matthew David, Snider, Charles J., Korbin, John P., Roberts, Scott Alan, & Collins, Lincoln. Uncertainty-refined image segmentation under domain shift. United States.
Martinez, Carianne, Potter, Kevin Matthew, Donahue, Emily, Smith, Matthew David, Snider, Charles J., Korbin, John P., Roberts, Scott Alan, and Collins, Lincoln. Tue . "Uncertainty-refined image segmentation under domain shift". United States. https://www.osti.gov/servlets/purl/1924873.
@article{osti_1924873,
title = {Uncertainty-refined image segmentation under domain shift},
author = {Martinez, Carianne and Potter, Kevin Matthew and Donahue, Emily and Smith, Matthew David and Snider, Charles J. and Korbin, John P. and Roberts, Scott Alan and Collins, Lincoln},
abstractNote = {A method for digital image segmentation is provided. The method comprises training a neural network for image segmentation with a labeled training dataset from a first domain, wherein a subset of nodes in the neural net are dropped out during training. The neural network receives image data from a second, different domain. A vector of N values that sum to 1 is calculated for each image element, wherein each value represents an image segmentation class. A label is assigned to each image element according to the class with the highest value in the vector. Multiple inferences are performed with active dropout layers for each image element, and an uncertainty value is generated for each image element. The label of any image element with an uncertainty value above a predefined threshold is replaced with a new label corresponding to the class with the next highest value.},
doi = {},
journal = {},
number = ,
volume = ,
place = {United States},
year = {2022},
month = {7}
}

Works referenced in this record:

Transaction periodicity forecast using machine learning-trained classifier
patent-application, December 2021


Uncertainty guided semi-supervised neural network training for image classification
patent-application, July 2021


Learning of Detection Model Using Loss Function
patent-application, October 2020


Adapting a generative adversarial network to new data sources for image classification
patent, January 2020


Training artificial neural networks with constraints
patent-application, September 2021


Methods to estimate effectiveness of a medical treatment
patent-application, September 2020


Unification of models having respective target classes with distillation
patent-application, February 2021


Training Artificial Neural Networks Using Context-Dependent Gating With Weight Stabilization
patent-application, August 2020


Robust anti-adversarial machine learning
patent-application, May 2020


User classification from data via deep segmentation for semi-supervised learning
patent-application, May 2021


Annealed dropout training of neural networks
patent-application, October 2016


Information processing apparatus
patent-application, January 2019


Face detection
patent-application, March 2017


Encoded data along tape based on colorspace schemes
patent, November 2020


Learning coach for machine learning system
patent-application, June 2020


Medical image classification based on a generative adversarial network trained discriminator
patent, March 2021


Training and application method of neural network model, apparatus, system and storage medium
patent-application, May 2020


Computer Program For Performing Drawing-Based Security Authentication
patent-application, October 2020


Training a neural network model
patent-application, November 2020


Method and apparatus for multi-level stepwise quantization for neural network
patent-application, November 2021


Information processing apparatus and method
patent-application, August 2020


Generative adversarial network medical image generation for training of a classifier
patent, March 2020


V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation
conference, October 2016


Device and method for multiclass object detection
patent-application, April 2012