Comparative analysis on different convolutional neural network (CNN) for classification
Rui Zeng, Yunfan Zhang · 2022 IEEE 5th International Conference on Information Systems and Computer Aided Education (ICISCAE) · 2022
CNN is one of the representative algorithms of deep learning. With the development of theory and the improvement of numerical computing equipment, CNN has developed rapidly. A variety of models have been derived and applied to different places. Taking the classification of vegetables and fruits as the data reference, this paper studies three CNN models: AlexNet, VGG and NIN. Our purpose is to analyze which model is most suitable for this image classification task and compare the advantages and disadvantages of the three models. By adjusting different hyper parameters; relative learning rate (Lr), epochs(#epochs) and batch size(batch _size) on the training loss, train accuracy and test accuracy among three models. The VGGNet with Lr = 0.00001, #epochs = 80 and batch_size = 32 has the best qualified for this graphic classification task.