US 11,798,145 B2
Image processing method and apparatus, device, and storage medium
Zequn Jie, Shenzhen (CN)
Assigned to TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Shenzhen (CN)
Filed by Tencent Technology (Shenzhen) Company Limited, Shenzhen (CN)
Filed on Mar. 3, 2021, as Appl. No. 17/191,611.
Application 17/191,611 is a continuation of application No. PCT/CN2019/119087, filed on Nov. 18, 2019.
Claims priority of application No. 201811457745.5 (CN), filed on Nov. 30, 2018.
Prior Publication US 2021/0192701 A1, Jun. 24, 2021
Int. Cl. G06K 9/00 (2022.01); G06T 5/20 (2006.01); G06N 3/08 (2023.01); G06N 3/045 (2023.01); G06N 3/048 (2023.01)
CPC G06T 5/20 (2013.01) [G06N 3/045 (2023.01); G06N 3/048 (2023.01); G06N 3/08 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 15 Claims
OG exemplary drawing
 
6. A computer device, comprising a memory and a processor, the memory storing computer-readable instructions, the computer-readable instructions, when executed by the processor, causing the computer device to perform a plurality of operations including:
inputting an original image to a decoder network according to an image transformation instruction, to obtain a first feature map of the original image, the decoder network being configured to extract features of an image;
inputting the first feature map sequentially to a plurality of transformer networks, each transformer network corresponding to at least one piece of transformation requirement information associated with the original image, to obtain a second feature map, each of the transformer networks being configured to perform image transformation to a respective region of the first feature map, further including:
determining, for each of the transformer networks, a conditional tensor according to the piece of transformation requirement information corresponding to the transformer network; and
transforming, based on the conditional tensor corresponding to the transformer network, a region corresponding to the transformer network in a respective feature map outputted by a preceding transformer network, and outputting the respective feature map of the transformer network; and
inputting the second feature map to a reconstruction network, to obtain a target image, the reconstruction network being configured to reconstruct an inputted feature map into a two-dimensional image.