Image processing method and device and storage medium

Abstract

The embodiment of the invention discloses an image processing method and device and a storage medium. The method comprises the steps of obtaining loss values of pixel points in a multi-label trainingimage; obtaining the number of pixel points corresponding to the labels of the multi-label training image; the number of the pixel points is adjusted according to the number adjusting parameters, andthe adjusted number of the pixel points of the label is obtained; determining a target pixel point from the pixel points according to the adjusted number of the pixel points and the loss value of thepixel points of the label, and obtaining a target image for neural network model back propagation training according to the target pixel point; according to the scheme, the number of pixel points in each label can be adjusted; the target pixel points are determined from the pixel points of the tags according to the number of the adjusted pixel points and the loss values of the pixel points in thetags, and in the subsequent back propagation process, only the target pixel points in the target image need to participate in back propagation, so that the training efficiency of the model can be improved.

Description

technical field [0001] The invention relates to the field of data identification, in particular to an image processing method, device and storage medium. Background technique [0002] The imbalance of sample labels is one of the common problems in machine learning. During the training process of machine learning, the distribution of the number of samples (such as the number of pixels) between various labels in the training image may be very different. Imbalance, that is, some labels have a large number of samples, while some labels have a small number of samples. If all samples participate in backpropagation, the training efficiency of the model will be very low. In order to improve training efficiency, it can be adjusted to participate in backpropagation of samples. Contents of the invention [0003] Embodiments of the present invention provide an image processing method, device, and storage medium, which can improve model training efficiency. [0004] An embodiment of ...

AI Technical Summary

Problems solved by technology

[0002] The imbalance of sample labels is one of the common problems in machine learning. During the training process of machine learning, the distribution of the number of samples (such as the number of pixels) between various labels in the training image may be very different. Imbalance, that is, some labels have a large number of samples, while some labels have a small number of samples. If all samples participate in backpropagation, the training efficiency of the model will be very low. In order to improve training efficiency, it can be adjusted to participate in backpropagation sample of

Images