US 11,775,845 B2
Character recognition method and apparatus, electronic device and computer readable storage medium
Xiaoqiang Zhang, Beijing (CN); Chengquan Zhang, Beijing (CN); and Shanshan Liu, Beijing (CN)
Assigned to Beijing Baidu Netcom Science and Technology Co., Ltd., Beijing (CN)
Filed by Beijing Baidu Netcom Science and Technology Co., LTD, Beijing (CN)
Filed on Mar. 23, 2021, as Appl. No. 17/209,380.
Claims priority of application No. 202010616043.8 (CN), filed on Jun. 30, 2020.
Prior Publication US 2021/0209401 A1, Jul. 8, 2021
Int. Cl. G06N 5/00 (2023.01); G06N 5/022 (2023.01); G06V 30/148 (2022.01); G06V 30/262 (2022.01); G06V 30/18 (2022.01); G06V 30/196 (2022.01); G06V 10/764 (2022.01); G06V 20/62 (2022.01); G06V 30/10 (2022.01)
CPC G06N 5/022 (2013.01) [G06V 10/764 (2022.01); G06V 20/62 (2022.01); G06V 30/153 (2022.01); G06V 30/18181 (2022.01); G06V 30/1988 (2022.01); G06V 30/274 (2022.01); G06V 30/10 (2022.01)] 9 Claims
OG exemplary drawing
 
1. A character recognition method, comprising:
determining semantic information and first position information of each individual character recognized from an image, wherein the semantic information comprises at least character category information and convolutional features of the image;
constructing a graph network according to the semantic information and the first position information of each individual character; and
determining a character recognition result of the image according to a feature of each individual character calculated by the graph network,
wherein, the determining the semantic information and the first position information of each individual character recognized from the image, comprises:
determining the semantic information of each individual character according to convolutional features of the image and character category information of each individual character; and
determining the first position information of each individual character according to the convolutional features of the image and second position information of each individual character; and
wherein the method further comprises:
before the determining the semantic information of each individual character according to the convolutional features of the image and the character category information of each individual character, determining the character category information of each individual character through a first preset neural network and the convolutional features of the image, wherein the character category information comprises a probability that each feature point corresponding to each individual character belongs to any character category.