公开(公告)号：US20210334543A1

公开(公告)日：2021-10-28

申请号：US17222055

申请日：2021-04-05

Applicant: AJOU UNIVERSITY INDUSTRY-ACADEMIC COOPERATION FOUNDATION

Inventor： Yong Seok HEO ,  Sangyong PARK

IPC: G06K9/00 ,  G06N3/04 ,  G06K9/42 ,  G06T7/11

Abstract: The present disclosure provides a knowledge distillation based semantic image segmentation method including inputting an input image to a teacher network and a student network; normalizing a first feature vector corresponding to each pixel in a feature map of a last layer of the teacher network and normalizing a second feature vector corresponding to each pixel in a feature map of a last layer of the student network; generating the first channel and space association matrix and the second channel and space association matrix based on the normalized first feature vector and the normalized second feature vector, and defining a first loss function based on an Euclidean norm value of the difference between the first channel and space association matrix and the second channel and space association matrix.