In this project, we implemented a classic convolutional neural network, comparing the performance using different regularization techniques, such as data augmentation, droput and batch-normalization on CIFAR10. In the end, we also showed how transfer learning can improve performance and reduce training time.
Framework : PyTorch