US 12,314,342 B2
Object recognition method and apparatus
Jun Yue, Shenzhen (CN); Jianzhuang Liu, Shenzhen (CN); Songcen Xu, Shenzhen (CN); Youliang Yan, Shenzhen (CN); and Li Qian, Shenzhen (CN)
Assigned to Huawei Technologies Co., Ltd., Shenzhen (CN)
Filed by Huawei Technologies Co., Ltd., Shenzhen (CN)
Filed on Sep. 24, 2021, as Appl. No. 17/484,545.
Application 17/484,545 is a continuation of application No. PCT/CN2020/081436, filed on Mar. 26, 2020.
Claims priority of application No. 201910236937.1 (CN), filed on Mar. 26, 2019.
Prior Publication US 2022/0012533 A1, Jan. 13, 2022
Int. Cl. G06F 18/214 (2023.01); G06F 18/2413 (2023.01); G06F 18/28 (2023.01); G06N 3/04 (2023.01); G06N 3/08 (2023.01); G06V 10/28 (2022.01); G06V 10/44 (2022.01); G06V 10/46 (2022.01)
CPC G06F 18/214 (2023.01) [G06F 18/2413 (2023.01); G06F 18/28 (2023.01); G06N 3/04 (2013.01); G06N 3/08 (2013.01); G06V 10/28 (2022.01); G06V 10/457 (2022.01); G06V 10/462 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method, comprising:
obtaining, by a device, a to-be-recognized image and one or more body regions of the to-be-recognized image, wherein the one or more body regions comprise a to-be-recognized object;
determining, by the device, a saliency score of each body region of the one or more body regions, wherein the respective saliency score of each respective body region represents a saliency degree of an object in the respective body region;
in response to a saliency score of a body region A being greater than or equal to a categorization threshold, determining, by the device, a feature vector of an object in the body region A based on a feature of the object in the body region A, and determining a category of the object in the body region A based on the feature vector of the object in the body region A and a category feature vector in a feature library, wherein the body region A is comprised in one of the one or more body regions, and the category feature vector represents a common feature of objects of a same category or a feature of one category of objects; wherein the category feature vector in the feature library is a center feature vector, the center feature vector represents the common feature of objects of the same category, and determining the category of the object in the body region A based on the feature vector of the object in the body region A and the category feature vector in the feature library comprises:
extracting the feature of the object in the body region A using a convolutional neural network (CNN), to obtain the feature vector of the object in the body region A, wherein the CNN comprises an input layer configured to receive a portion of the to-be-recognized image corresponding to the body region A, a convolutional layer coupled to the input layer, and a neural network layer coupled to the convolutional layer, the neural network layer comprising at least one hidden layer and an output layer coupled to the at least one hidden layer;
calculating a distance between the feature vector of the object in the body region A and a center feature vector corresponding to each category in the feature library; and
determining a category corresponding to a target center feature vector to be the category of the object in the body region A, wherein the target center feature vector is a center feature vector in the feature library that is closest to the feature vector of the object in the body region A.