I wrote the project to help me understand the CNN deep learning better and to practice writing deep learning model with pytorch.
It's a simple image classifier with 3 convolutional layers and 2 fully connected layers. The training dataset is classic CIFAR-10. After 50-epoch-training, the test error is 0.38 and training error is 0.36. So, to enhance the performance of the classifier, I believe I have to increate the complexity of the model or train it with longer time.