| CPC G06V 10/7715 (2022.01) [G06V 10/26 (2022.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01)] | 20 Claims |
|
1. An image recognition method applied to an electronic device, the method comprising:
acquiring an image to be recognized, and acquiring a plurality of initial labeled images and an initial labeled result of each of the plurality of initial labeled images from databases, wherein the initial labeled result of the initial labeled image comprises a correspondence between an initial category and an initial label of each initial object in the initial labeled image, the initial label of each initial object refers to a correspondence between an initial color and an initial serial number of the each initial object;
constructing a first semantic segmentation network;
for the initial labeled result of each initial labeled image, obtaining a comparison result by comparing the initial labeled result with a preset labeled result, comprising: determining that the initial labeled result does not match the preset labeled result, in response that each initial category has a corresponding preset category in the preset labeled result but at least one initial label in the initial labeled result does not have a corresponding target label in the preset labeled result, wherein the preset labeled result indicates a relationship between each preset category of each target object and each target label, and each target label indicates a relationship between a preset color label and a preset serial number;
in response that the comparison result indicates that the initial labeled result of one of the plurality of initial labeled images does not match the preset labeled result, obtaining a target image corresponding to the one of the plurality of initial labeled images and obtaining a target labeled result of the target image, by inputting the one of the plurality of initial labeled images into the first semantic segmentation network, which refers to a network for re-labelling the one of the plurality of initial labeled images;
obtaining a plurality of target images and obtaining the target labeled result of each target image;
obtaining a second semantic segmentation network by training the first semantic segmentation network based on the plurality of target images and the target labeled result of each target image; and
obtaining a labeled result of the image to be recognized by inputting the image to be recognized into the second semantic segmentation network.
|