US 12,260,531 B2
Using imager with on-purpose controlled distortion for inference or training of an artificial intelligence neural network
Patrice Roulet, Montreal (CA); Pierre Konen, Saint-Bruno (CA); Pascale Nini, Orford (CA); Simon Thibault, Quebec City (CA); Jocelyn Parent, Montreal (CA); Viacheslav Natashyn, Chateauguay (CA); and Julie Buquet, Montreal (CA)
Assigned to IMMERVISION, INC., Montreal (CA)
Filed by ImmerVision, Inc., Montreal (CA)
Filed on Nov. 17, 2020, as Appl. No. 16/950,218.
Claims priority of provisional application 62/936,647, filed on Nov. 18, 2019.
Prior Publication US 2021/0150679 A1, May 20, 2021
Int. Cl. G06T 5/80 (2024.01); G06N 3/084 (2023.01); G06N 3/086 (2023.01); G06T 3/00 (2024.01)
CPC G06T 5/80 (2024.01) [G06N 3/084 (2013.01); G06N 3/086 (2013.01); G06T 3/00 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A method for inference processing of at least one input digital image file with controlled distortion using an artificial intelligence neural network to improve the output of the neural network, the method comprising:
a. receiving, by a neural network, an input digital image file with controlled distortion created by an imager;
b. inference processing, by the neural network, the input digital image file, wherein the neural network is formed by algorithms or software codes running on a computing device and has been specifically trained for processing images with controlled distortion, wherein the inference processing is performed without removing the controlled distortion from the input digital image file;
c. outputting, by the neural network, interpreted data derived from the input digital image file by the inference processing, the interpreted data output being an output digital image file with the controlled distortion; and
d. dewarping the output digital image file to remove at least in part the controlled distortion.