Data set expansion and training method and device of an image classification model

Abstract

The embodiment of the invention discloses a data set expansion and training method and device of an image classification model, the image classification model is realized based on a convolutional neural network, and the data set expansion method comprises the following steps: for each picture sample in at least part of picture samples in a training data set of the image classification model, respectively executing the following operations: obtaining a class activation graph CAM of a preset class corresponding to the picture sample; obtaining an area corresponding to a preset target from the CAM graph by adopting a preset algorithm, and determining a position coordinate of the area in the picture sample; obtaining a cut picture from the picture sample by utilizing the position coordinates; and labeling the cut picture as the same category as the picture sample, and storing the picture sample into the training data set. According to the scheme of the invention, in the training process of the image classification model, effective training data set samples can be added.

Description

technical field [0001] The embodiments of the present application relate to but are not limited to the field of image classification, and in particular, relate to a data set expansion method, training method and device for an image classification model. Background technique [0002] In recent years, deep neural networks have become the most important tools in computer vision, such as image classification, object detection, and face recognition, both at the level of theoretical research and practical application. However, the training of deep neural networks generally requires a large amount of training data to obtain ideal results, and the acquisition of labeled data often requires a large cost. The sample enhancement or data augmentation (Data Augmentation) method is a method of automatically generating training samples without manual labeling, which can effectively improve the diversity of training samples, improve model robustness, and avoid overfitting. Typical data enh...

AI Technical Summary

Problems solved by technology

However, random cropping can easily lead to incomplete or even missing target information, and a large amount of non-target information is retained. The samples generated at this time often have a negative effect. This negative effect is not obvious on large data sets, but it is not obvious in small data sets or Small target data sets (targets account for a small proportion of the image) will have a negative impact on normal target recognition

Images