US 12,488,569 B2
Unpaired image-to-image translation using a generative adversarial network (GAN)
Nick Shadbeh Evans, Lynnwood, WA (US); Jasper Patrick Corleis, Renton, WA (US); and Amir Afrasiabi, University Place, WA (US)
Assigned to The Boeing Company, Arlington, VA (US)
Filed by The Boeing Company, Chicago, IL (US)
Filed on Dec. 1, 2021, as Appl. No. 17/540,198.
Claims priority of provisional application 63/125,350, filed on Dec. 14, 2020.
Prior Publication US 2022/0189145 A1, Jun. 16, 2022
Int. Cl. G06V 10/774 (2022.01); G06F 18/21 (2023.01); G06F 18/214 (2023.01); G06N 3/045 (2023.01); G06N 3/0475 (2023.01); G06N 3/094 (2023.01); G06T 5/60 (2024.01); G06V 10/776 (2022.01); G06V 10/82 (2022.01); G06V 20/17 (2022.01); G06V 20/70 (2022.01)
CPC G06V 10/774 (2022.01) [G06V 10/776 (2022.01); G06V 10/82 (2022.01); G06V 20/17 (2022.01); G06V 20/70 (2022.01); G06F 18/214 (2023.01); G06F 18/217 (2023.01); G06N 3/045 (2023.01); G06N 3/0475 (2023.01); G06N 3/094 (2023.01); G06T 5/60 (2024.01); G06T 2207/20084 (2013.01)] 20 Claims
OG exemplary drawing
 
1. An apparatus for identifying objects, the apparatus comprising:
a processor configured to execute instructions stored on a memory; and
the memory storing computer-readable instructions that, when executed by the processor, cause the processor to:
receive a dataset comprising an image having first real image features;
train a neural network to recognize the first real image features in the dataset;
perform a synthetic image augmentation to generate synthetic image features corresponding to the first real image features in the image using the neural network,
wherein the synthetic image augmentation allows for training of a computer vision function for recognizing second real image features, corresponding to the synthetic image features, in a real-world environment;
associate, in real-time, the image to a synthetic image corresponding to the image by correlating features in the image with the synthetic image features,
wherein the synthetic image features are of the synthetic image;
measure a realism gap between the image and the synthetic image corresponding to the image,
wherein to measure the realism gap, the instructions further cause the processor to:
classify each of the synthetic image features of the synthetic image to obtain a synthetic classification, and
compare the synthetic classification to a classification of the image,
wherein the classification of the image relates to classification of each of the first real image features;
determine that the realism gap does not satisfy a threshold; and
update the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.