US 12,131,478 B2
Method for object segmentation based on deep-learning using a trained neural network and system for performing the same
Seung On Bang, Seoul (KR)
Assigned to GYNETWORKS CO., LTD., Incheon (KR)
Filed by GYNETWORKS CO., LTD., Incheon (KR)
Filed on Jan. 26, 2022, as Appl. No. 17/584,540.
Claims priority of application No. 10-2021-0016258 (KR), filed on Feb. 4, 2021; and application No. 10-2021-0164131 (KR), filed on Nov. 25, 2021.
Prior Publication US 2022/0245818 A1, Aug. 4, 2022
Int. Cl. G06T 7/10 (2017.01)
CPC G06T 7/10 (2017.01) [G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/20221 (2013.01)] 5 Claims
OG exemplary drawing
 
1. A method for object segmentation using a trained neural network, comprising:
receiving a target image;
splitting the target image into unit images having a predetermined size;
outputting a unit activation map for the split unit image as a first segmentation result by using the neural network;
merging the unit activation map; and
outputting a second segmentation result according to the merged unit activation map,
wherein the neural network is trained by defining a loss function as a first loss term defining a difference between the entire activation map of an original image and a ground-truth, a second loss term defining a difference between the merged activation map merged with a unit activation map of a unit image split from the original image and the ground-truth, and a third loss term defining a difference between the entire activation map and the merged activation map.