US 12,394,171 B2
Region recognition method, apparatus and device, and readable storage medium
Yuqiang Ren, Shenzhen (CN); Xingjia Pan, Shenzhen (CN); Weiming Dong, Shenzhen (CN); Xudong Zhu, Shenzhen (CN); Haolei Yuan, Shenzhen (CN); Xiaowei Guo, Shenzhen (CN); and Changsheng Xu, Shenzhen (CN)
Assigned to TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Guangdong (CN)
Filed by Tencent Technology (Shenzhen) Company Limited, Guangdong (CN)
Filed on Apr. 27, 2022, as Appl. No. 17/730,717.
Application 17/730,717 is a continuation of application No. PCT/CN2021/090152, filed on Apr. 27, 2021.
Claims priority of application No. 202010446547.X (CN), filed on May 25, 2020.
Prior Publication US 2022/0254134 A1, Aug. 11, 2022
Int. Cl. G06V 10/24 (2022.01); G06V 10/22 (2022.01); G06V 10/25 (2022.01); G06V 10/32 (2022.01)
CPC G06V 10/242 (2022.01) [G06V 10/22 (2022.01); G06V 10/25 (2022.01); G06V 10/32 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A region recognition method, performed by a computer device and comprising:
acquiring an input image, the input image comprising image content of a to-be-recognized region;
acquiring a first rotation angle of the image content in the input image, the first rotation angle being estimated by a recognition model, and the first rotation angle representing a tilt angle of the image content in the input image relative to a reference angle;
rotating a plurality of convolution kernels of a plurality of network branches in the recognition model at the first rotation angle before the plurality of network branches converges, wherein the convolution kernels are of different types and are separately obtained for each of the plurality of network branches;
performing convolution processing on an image feature of the input image using the rotated convolution kernels to obtain a target feature;
recognizing the target feature to obtain region data corresponding to the image content; and
determining, in the input image by using the region data, a target region corresponding to the image content.