Topic: Image classification is one of the most important topics in machine learning: given an image, you have to classify it to a class and assign to it the appropriate class label (e.g., ‘dog’, ‘car’. The tools of choice are deep neural networks, notably Convolutional Neural Networks (CNN).
Exercise: The tasks are: a) to train a CNN neural network with the training part of CIFAR10 dataset (https://www.cs.toronto.edu/~kriz/cifar.html) and b) to classify the test part/split of CIFAR10 dataset (https://www.cs.toronto.edu/~kriz/cifar.html).
Instructions: You are given a toy neural network on which you can rely on. You can build a more optimized one. You must also experiment with all the hyperparameters, such as activation functions, loss function etc., as they are not the optimal ones.
You can download the exercise’s resources from here.
Material for better understanding: A lecture on Convolutional Neural Networks (CNN), e.g., from: https://icarus.csd.auth.gr/convolutional-neural-networks-lecture/
Knowledge Assessment questionnaire: https://aiia.csd.auth.gr/gr/cvml-knowledge-self-assessment/
For the solutions to the exercises, please contact koroniioanna@csd.auth.gr