Title:
METHOD FOR PROCESSING IMAGE THROUGH NEURAL NETWORK, AND ELECTRONIC DEVICE THEREFOR
Document Type and Number:
WIPO Patent Application WO/2024/043604
Kind Code:
A1
Abstract:
An electronic device according to one embodiment may comprise: a memory for storing a first neural network model; and at least one processor. The electronic device can be configured such that the at least one processor: generates device state information related to image processing speed of the electronic device; determines a quantization level for the first neural network model on the basis of the device state information; and generates a second neural network model on the basis of the quantization level so as to restore a deteriorated image. Other various embodiments identified through the specification are possible.
More Like This:
Inventors:
CHUN HYUNGJU (KR)
KANG YOUNGJUN (KR)
KIM YOUNGJO (KR)
PARK HYUNHEE (KR)
CHOI JONGBUM (KR)
HAN CHANGSU (KR)
PARK JAEHYOUNG (KR)
WON JONGHOON (KR)
KANG YOUNGJUN (KR)
KIM YOUNGJO (KR)
PARK HYUNHEE (KR)
CHOI JONGBUM (KR)
HAN CHANGSU (KR)
PARK JAEHYOUNG (KR)
WON JONGHOON (KR)
Application Number:
PCT/KR2023/012115
Publication Date:
February 29, 2024
Filing Date:
August 16, 2023
Export Citation:
Assignee:
SAMSUNG ELECTRONICS CO LTD (KR)
International Classes:
G06N3/08; G06N3/045; G06T5/00
Foreign References:
KR101821285B1 | 2018-01-23 | |||
KR20220027436A | 2022-03-08 | |||
CN113422876A | 2021-09-21 | |||
CN113112009A | 2021-07-13 |
Other References:
WANG KUAN; LIU ZHIJIAN; LIN YUJUN; LIN JI; HAN SONG: "HAQ: Hardware-Aware Automated Quantization With Mixed Precision", 2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), IEEE, 15 June 2019 (2019-06-15), pages 8604 - 8612, XP033687500, DOI: 10.1109/CVPR.2019.00881
Attorney, Agent or Firm:
KWON, Hyuk-Rok et al. (KR)
Download PDF: