US 12,437,511 B2
Image classification method, electronic device and storage medium
Chien-Wu Yen, New Taipei (TW)
Assigned to HON HAI PRECISION INDUSTRY CO., LTD., New Taipei (TW)
Filed by HON HAI PRECISION INDUSTRY CO., LTD., New Taipei (TW)
Filed on Feb. 13, 2023, as Appl. No. 18/108,780.
Claims priority of application No. 202211321457.3 (CN), filed on Oct. 26, 2022.
Prior Publication US 2024/0144649 A1, May 2, 2024
Int. Cl. G06V 10/00 (2022.01); G06V 10/75 (2022.01); G06V 10/762 (2022.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01)
CPC G06V 10/764 (2022.01) [G06V 10/751 (2022.01); G06V 10/762 (2022.01); G06V 10/82 (2022.01)] 17 Claims
OG exemplary drawing
 
1. An image classification method applied to an electronic device, the method comprising:
acquiring a classification network, an image to be classified, a plurality of training images, and a plurality of test images;
obtaining a first classification model by training the classification network based on the plurality of training images, comprising: calculating a loss value of the classification network based on the plurality of training images; and obtaining the first classification model by adjusting the classification network based on the loss value, until the loss value decreases to a minimum value;
obtaining a prediction result of each test image of the plurality of test images by inputting each test image into the first classification model, determining a plurality of target images from the plurality of test images according to the prediction result of each test image, calculating a prediction accuracy rate of the first classification model based on the plurality of target images;
in response that the prediction accuracy rate is less than a preset value, obtaining a second classification model by adjusting the first classification model according to the plurality of target images, the second classification model comprising a flattening layer, a fully connected layer, and a classification layer;
acquiring an initial feature matrix output from the flattening layer by inputting the image to be classified into the second classification model;
in response that a dimension of the initial feature matrix is less than a dimension of an initial weight matrix in the fully connected layer, obtaining an input feature matrix by performing a dimension raising process on the initial feature matrix;
obtaining a target weight matrix by rearranging elements in the initial weight matrix;
generating a target vector according to the target weight matrix and the input feature matrix; and
obtaining a classification result of the image to be classified by inputting the target vector into the classification layer.