Title:
FEATURE MAP PROCESSING METHOD AND RELATED DEVICE
Document Type and Number:
WIPO Patent Application WO/2023/274191
Kind Code:
A1
Abstract:
The present application relates to the technical field of artificial intelligence. Provided are a feature map processing method and a related device. The method is implemented by calling a neural network model, which comprises a plurality of input adaptation branches and a post-processing portion, wherein outputs of the plurality of input adaptation branches serve as an input of the post-processing portion, and each of the plurality of input adaptation branches can perform down-sampling processing on a feature map at different rates. Specifically, the method comprises: inputting a first feature map into one input adaptation branch of a plurality of input adaptation branches, so as to obtain a second feature map, wherein the input adaptation branch is determined according to the size of the first feature map and a preset size threshold value, the preset size threshold value is an upper-limit size of an input feature map of a post-processing portion, and the size of the second feature map is not greater than the preset size threshold value; and inputting the second feature map into the post-processing portion to perform post-processing, so as to obtain a processing result of the first feature map. By means of the embodiments of the present application, the requirement for computing power can be lowered.
Inventors:
ZHAO YIN (CN)
MAO JUE (CN)
YANG HAITAO (CN)
MAO JUE (CN)
YANG HAITAO (CN)
Application Number:
PCT/CN2022/101753
Publication Date:
January 05, 2023
Filing Date:
June 28, 2022
Export Citation:
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G06K9/00
Foreign References:
| CN112668584A | 2021-04-16 | |||
| CN109447247A | 2019-03-08 | |||
| CN111401379A | 2020-07-10 | |||
| US20200175313A1 | 2020-06-04 | |||
| CN202110740186A | 2021-06-30 |
Other References:
ANGELOS KATHAROPOULOS; FRANÇOIS FLEURET: "Processing Megapixel Images with Deep Attention-Sampling Models", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 3 May 2019 (2019-05-03), 201 Olin Library Cornell University Ithaca, NY 14853 , XP081437146
Attorney, Agent or Firm:
SCIHEAD IP LAW FIRM (CN)
Download PDF: