Image recognition method based on improved Focal loss function

Abstract

The invention discloses an image recognition method based on an improved Focal loss function, and the method comprises the steps: improving a modulation factor of the Focal loss function on the basis of an existing Focal loss function, enabling the function to have higher attention to a difficult sample, and enabling the function to have relatively lower attention to a simple sample; and then, on the basis of the convolutional neural network model based on the Focal loss function, predicting the residual negative sample set, screening all difficult samples, dividing the samples into N equal parts, respectively adding the N equal parts into the original training set to form N new training sets, then training a plurality of models, and determining a final prediction picture label result according to voting selection of the N models. According to the method, on the basis of an original Focal loss function, the attention degree on a difficult sample is higher, and the model generalization ability is improved; and under the condition of excessive negative samples, difficult samples are screened out by utilizing the original model and added into the training set, so that the accuracy of the model is improved, available samples are fully utilized, and the calculation amount during model training is also reduced.

Description

technical field [0001] The invention belongs to the technical field of multi-label image recognition based on deep learning, and in particular relates to an image recognition method based on an improved Focal loss function. Background technique [0002] Convolutional Neural Network (CNN) is a neural network specially designed for image recognition problems. CNN can effectively obtain the features of the original image, and rarely go through the preprocessing process. Compared with traditional machine learning, convolutional neural network can automatically extract image features, which is derived from two important characteristics of convolutional neural network: sparse connection and value sharing. Sparse connection means that the nodes of the convolutional layer are only connected to some nodes of the previous layer, and are only used to learn local features. Weight sharing means that in the process of scanning a picture with a convolution kernel, each position in the pi...

AI Technical Summary

Problems solved by technology

For some methods to solve the sample imbalance, some do not make full use of all samples, resulting in the lack of part of the sample data, and some use all the samples, which makes the calculation amount very large and increases the cost

Images