A fine-grained classification method based on meta-learning

Abstract

The invention discloses a fine-grained classification method based on meta-learning, which comprises the following steps of: establishing an external data set, dividing the data set into a training set, a verification set and a test set, and enabling sample types of the three sets to be not intersected, and enabling the sample types of the test set to be smaller than that of the training set; Performing data enhancement on samples in the data set; A convolutional neural network is established, the input of the convolutional neural network is a color picture, the output of the convolutional neural network is a category to which the color picture belongs, the length of a classification layer is equal to the number of categories of an external data set, and a loss function adopts softmax loss; Training a fine-grained classification network by adopting the training set; and testing the pre-trained convolutional neural network by using the test set, and carrying out fine tuning on the convolutional neural network according to a test result. According to the method, a good general initialization model can be rapidly generated, so that a good classification effect can be obtained only byusing fewer samples when related but different categories are tested, and the problem that no big data set exists during fine-grained classification is solved.

Description

technical field [0001] The invention belongs to the technical field of calculation and calculation, in particular to the technical field of computer vision for fine-grained classification, and relates to a fine-grained classification method based on meta-learning. Background technique [0002] Fine-grained image recognition is a challenging task in image classification, where the goal is to correctly identify objects among numerous subclasses within a large class. In general, fine-grained image classification is to find some subtle local areas, and use the characteristics of these local areas to classify the original image. However, the current fine-grained classification algorithms basically use general models (such as VGG16) for fine-grained classification, which limits the structure of the classification model and the migration effect is not very good. Therefore, we use the method of meta-learning to quickly generate a good general initialization model, and then perform ...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a fine-grained classification method based on meta-learning, which can quickly generate a good general initialization model, so that when testing related but different categories, only fewer samples can be used to obtain better results. Classification effect, to solve the problem of no large data set in fine-grained classification

Images