Title:
FIRST PERSON POINT-OF-VIEW IMAGE RECOGNITION METHOD AND APPARATUS, AND COMPUTER-READABLE STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2022/117096
Kind Code:
A1
Abstract:
A first person point-of-view image recognition method, comprising: S1: acquiring multiple images containing a first person point-of-view and multiple images not containing the first person point-of-view; S2: classifying and labeling the images according to whether the images contain hand features and hand feature information; S3: enhancing the classified and labeled images so as to obtain diversified image training samples; S4: inputting the image training samples into a pre-constructed neural network for training to obtain a first point-of-view image recognition model; S5: inputting an image to be recognized into the first point-of-view image recognition model; S6: determining whether said image is a first person point-of-view image according to the output of the first point-of-view image recognition model. The described method can automatically identify whether the image is a first person point-of-view image. Users can automatically distinguish whether a video frame is a first person point-of-view video frame by providing an inputted video frame as needed. The described method is conducive to post-processing videos, and has advantages of rapid processing and high accuracy.
Inventors:
GAO RUIDONG (CN)
CHEN BOLIN (CN)
CAI JINLIN (CN)
CHEN BOLIN (CN)
CAI JINLIN (CN)
Application Number:
PCT/CN2021/135527
Publication Date:
June 09, 2022
Filing Date:
December 03, 2021
Export Citation:
Assignee:
ARASHI VISION INC (CN)
International Classes:
G06K9/00
Foreign References:
CN112381055A | 2021-02-19 | |||
CN110348572A | 2019-10-18 | |||
CN111814810A | 2020-10-23 | |||
CN111062871A | 2020-04-24 | |||
CN105718879A | 2016-06-29 | |||
US7171042B2 | 2007-01-30 |
Download PDF: