We apply CNN, ResNet18, and ResNet50 methods to classify tasks on BreastMNIST, and because we are doing binary classification tasks, the loss function uses the cross-entropy loss function. Through comparative experiment, we found that ResNet18 with SGD performed the best among the three methods, achieving the highest classification accuracy. This indicates that ResNet18 has an advantage in handling the classification problem on the BreastMNIST dataset.

We put our experiment results in the form of accuracy-epoch images in the directory of the corresponding model.

You can download our code to apply it to your dataset, and feel free to communicate with us if you encounter problems.

Author:Group 6

Pre-algorithm research：Jinyuan Fu,Qiuyu Liu

Algorithm implementation：Shaofei Liu,Aiqi Wang

Data and Graphical Analysis：Ziyi Lu

PPT production and copywriting：Qiuyu Liu,Shaofei Liu