CNN-image-classification 👓 🎯

Designing and tuning a convolutional neural network for image classification

1. Experimentation with the topology of the neural network and its hyper-parameters to evaluate the effects on the model performance
2. Building a neural network for the classification of images (fashion-MNIST dataset is used --> made available by Zalando Research)
3. Design a network using convolutional layers, possibly combined with pooling layers for the SCT dataset
4. The end objective is to design a network using convolutional layers, possibly combined with pooling layers and to tune the parameters, #layers, layer-sizes et cetera to achieve a relatively good performance for the Cifar-10 dataset containing 60,000 32x32 colored images (10 classes)

Techniques/approaches/update-functions etc. that are used in this project:

• Mean-subtraction normalization
• Gradient Descent with Momentum
• Gradient Descent Nesterov Momentum
• L2-weight decay
• Ada-Delta
• DropOut regularization

For more information, please visit: CS231n: Convolutional Neural Networks for Visual Recognition

Datasets/resources that were used:

• Zalando Research: Fashion-MNIST
• Canadian Institute For Advanced Research: CIFAR-10 dataset