In this project, I have used a simple 3 layered CNN to perform image classification on CIFAR-10 dataset. This highlights a very naive approach to train a simple CNN. Accuracy observed is 74 % here, but more layers can be added and different loss functions could be tested to improve accuracy. For the purpose, I have used Cross Entropy as my loss functions and SGD as my optimizer
I have also added a jupyter notebook if someone would prefer to use that.