基于深度残差网络的水果图像分类算法研究

Fruits Image Classification Algorithm Based on ResNet

深度学习是基于人工神经网络的一种机器学习方法,能够进行大规模矩阵运算。将其应用于图像识别领域,能够高效完成对高分辨率图像的处理,有效推动计算机视觉领域技术革新。在此基础上,利用深度残差网络模型进行对图像数据的识别与分类,引入迁移学习思想,将ResNet152模型迁移到三种水果的图像数据集中进行训练,实现对图像特征的获取,再对测试集进行分类识别训练,以此促进模型分类识别的准确率提高。通过增加测试集中数据量、增加训练难度进行重复实验,得出结论:ResNet152模型具有较高的准确率,验证该模型在图像识别分类领域的高效性。

Deep learning is a machine learning method based on artificial neural network,which can carry out large-scale matrix operations.Based on this,the use of the depth of the residual network model for identification and classification of image data,this paper introduces the migration learning thought,the migration of ResNet152 model to the image data of three kinds of fruit concentration training,realize the characteristics of image acquisition,classification recognition on test set training,promote the accuracy of classification model.By increasing the amount of data in the test set and increasing the difficulty of training,repeated experiments are carried out,and the conclusion is that the ResNet152 model has a good accuracy rate,which verifies the high efficiency of the model in the field of image recognition and classification.