Assignment 5 focuses on image classification with convolutional neural networks.
The Jupyter notebook in this repo is the notebook I submitted to be graded. This notebook was a guided application of ML techniques involving normalization and one-hot encoding of the Cifar-10 dataset, building a baseline nueral network model, training the baseline model, plotting and evaluating the models accuracy. Part 2 was a guided application of ML techniques involving pretrained CNN models as the starting point for a multi-class image classifier, building and attaching the top layers of the model to the transfer layers, and training and evaluating the model. Part 3 was a guided application of ML techniques involving using all the techniques covered in this and previous assignments to improve the models constructed in part 1 & 2. These techniques included Regularization, Initialization methds, Batch Normalization, Data augmentation, Learning Rate Scheduling, etc.
The end results was that I built two models, both convolutional neural networks. The first was built as a Baseline CNN for image classification and the second imported the lower layers of a pre-trained model and built on those lower layers. I did not finish Part 3 of this assignment.