ResNet50 Based Classification on Mnist and Cifar10
Boshan Chen · Applied and Computational Engineering · 2023
Convolutional neural network (CNN) is gradually trending in the past few years due to its learning capacity and great generalization power. CNNs use a different version of regularization, which makes it on lower extreme. In this task, we studied the performance of Residual neural networks (ResNet50) on Mnist and Cifar10 datasets. We built the model mainly based on ResNet50 with cross-entropy loss. Then we trained Mnist and Cifar10 datasets on it separately and evaluated its prediction accuracy. The prediction accuracy of this model on the test set of Mnist and Cifar10 is 35.87% and 95.35% respectively.