US 12,106,481 B2
Computer vision systems and methods for end-to-end training of convolutional neural networks using differentiable dual-decomposition techniques
Shaofei Wang, Philadelphia, PA (US); Vishnu Sai Rao Suresh Lokhande, Madison, WI (US); Maneesh Kumar Singh, Princeton, NJ (US); Konrad Kording, Philadelphia, PA (US); and Julian Yarkony, Jersey City, NJ (US)
Assigned to Insurance Services Office, Inc., Jersey City, NJ (US)
Filed by Insurance Services Office, Inc., Jersey City, NJ (US)
Filed on Dec. 14, 2020, as Appl. No. 17/121,257.
Claims priority of provisional application 62/947,874, filed on Dec. 13, 2019.
Prior Publication US 2021/0182675 A1, Jun. 17, 2021
Int. Cl. G06T 7/10 (2017.01); G06F 18/20 (2023.01); G06F 18/21 (2023.01); G06N 3/08 (2023.01); G06N 5/046 (2023.01); G06T 7/11 (2017.01); G06V 10/26 (2022.01); G06V 10/84 (2022.01); G06V 30/19 (2022.01); G06V 30/262 (2022.01)
CPC G06T 7/10 (2017.01) [G06F 18/2163 (2023.01); G06F 18/29 (2023.01); G06N 3/08 (2013.01); G06N 5/046 (2013.01); G06T 7/11 (2017.01); G06V 10/26 (2022.01); G06V 10/85 (2022.01); G06V 30/19153 (2022.01); G06V 30/274 (2022.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A computer vision system for end-to-end training of a neural network, comprising:
a memory; and
a processor in communication with the memory, the processor:
implementing a fixed point algorithm for dual-decomposition of a maximum-a-posteriori inference problem,
training a neural network to perform semantic image segmentation by applying the fixed point algorithm to training input data, and
processing one or more images to segment an attribute of the one or more images using the trained neural network.