US 11,748,881 B2
Deep learning based instance segmentation via multiple regression layers
Elad Arbel, Santa Clara, CA (US); Itay Remer, Santa Clara, CA (US); and Amir Ben-Dor, Santa Clara, CA (US)
Assigned to Agilent Technologies, Inc., Santa Clara, CA (US)
Filed by Agilent Technologies, Inc., Santa Clara, CA (US)
Filed on Jun. 23, 2022, as Appl. No. 17/847,326.
Application 17/847,326 is a division of application No. 16/846,180, filed on Apr. 10, 2020, granted, now 11,410,303.
Claims priority of provisional application 62/832,880, filed on Apr. 12, 2019.
Claims priority of provisional application 62/832,877, filed on Apr. 11, 2019.
Prior Publication US 2022/0366564 A1, Nov. 17, 2022
Int. Cl. G06T 7/00 (2017.01); G06T 7/10 (2017.01); G06T 7/187 (2017.01); G06T 7/11 (2017.01); G06T 7/174 (2017.01); G06N 20/00 (2019.01); G06F 3/0482 (2013.01); G06F 3/0486 (2013.01); G06N 3/08 (2023.01); G06F 18/2431 (2023.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01); G06V 10/80 (2022.01); G06V 10/82 (2022.01); G06V 10/26 (2022.01); G06V 10/28 (2022.01); G06V 20/69 (2022.01); G06V 10/94 (2022.01); G06V 10/24 (2022.01)
CPC G06T 7/0012 (2013.01) [G06F 3/0482 (2013.01); G06F 3/0486 (2013.01); G06F 18/2431 (2023.01); G06N 3/08 (2013.01); G06N 20/00 (2019.01); G06T 7/0014 (2013.01); G06T 7/10 (2017.01); G06T 7/11 (2017.01); G06T 7/174 (2017.01); G06T 7/187 (2017.01); G06V 10/267 (2022.01); G06V 10/28 (2022.01); G06V 10/764 (2022.01); G06V 10/7753 (2022.01); G06V 10/809 (2022.01); G06V 10/82 (2022.01); G06V 10/945 (2022.01); G06V 20/695 (2022.01); G06V 20/698 (2022.01); G06T 2207/10056 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/20104 (2013.01); G06T 2207/30024 (2013.01); G06T 2207/30096 (2013.01); G06V 10/247 (2022.01)] 18 Claims
OG exemplary drawing
 
1. A method of inference by an artificial intelligence (“AI”) model trained during a training phase, comprising:
receiving, with a computing system, a first image, the first image comprising a field of view (“FOV”) of a first biological sample;
generating, using the AI model that is trained or updated during the training phase by a trained AI system, two or more predicted images based on the first image comprising at least one third encoded predicted image and at least one fourth encoded predicted image, each of the two or more predicted images being different from each other,
wherein the training phase of the AI system comprises training the AI system to generate or update the AI model to predict instances of objects of interest based on at least in part on a plurality of sets of at least two training images comprising at least one third training encoded image and at least one fourth training encoded image that are generated based on an annotated training image,
wherein the at least one third encoded predicted image comprises highlighting of a centroid for each instance of an object of interest in the annotated training image,
wherein the at least one fourth encoded predicted image comprises highlighting of an edge or border for each instance of the object of interest; and
decoding, with the computing system and using the decoder, the two or inure predicted images comprising the at least one third encoded predicted image and the at least one fourth encoded predicted image to generate a second predicted image comprising predicted instances of objects of interest in the first biological sample.