CPC G06V 20/62 (2022.01) [G06V 10/26 (2022.01); G06V 10/44 (2022.01); G06V 10/454 (2022.01); G06V 10/82 (2022.01); G06V 30/10 (2022.01); G06V 30/15 (2022.01); G06V 30/19027 (2022.01)] | 18 Claims |
1. A method of recognizing a text, comprising:
according to a preset feature extraction network and a to-be-recognized image, extracting a feature map of the to-be-recognized image;
providing the feature map to a preset segmentation network, determining segmentation information of a text region of the to-be-recognized image;
according to the segmentation information, determining boundary key points in the text region;
according to the boundary key points, converting a text in the text region into a text with a target arrangement sequence; and
inputting the text obtained by conversion into a preset recognition model for recognition processing;
wherein according to the segmentation information, determining the boundary key points in the text region comprising:
according to offsets between each pixel point in a first boundary region and two boundary key points in the first boundary region in the segmentation information, determining position information of the two boundary key points in the first boundary region; and according to offsets between each pixel point in a second boundary region and two boundary key points in the second boundary region in the segmentation information, determining position information of the two boundary key points in the second boundary region, wherein the first boundary region is located at a head portion of the text region, and the second boundary region is located at a tail portion of the text region; and
according to the position information of the two boundary key points in the first boundary region and the position information of the two boundary key points in the second boundary region, determining other boundary key points other than boundary key points in the first boundary region and boundary key points in the second boundary region in the text region.
|