The main aim of this project is to understand and implement deep neural network (DNN) and convolutional neural network (CNN) and see how these can be used for classification problems.
- Python
In this we will use a subset of images from Tiny ImageNet, which is a subset of the ImageNet dataset. Our subset of Tiny ImageNet contains 30 different categories, we will refer to it as TinyImageNet30. The training set has 450 resized images (64x64 pixels) for each category (13,500 images in total).
- numpy
- h5py
- matplotlib
- pytorch
- Function Implementation - to get the datasets and the dataloaders
- Function Implementation - model class for a simple MLP and CNN Model
- Model Training
- Generating Confusion Matrices and ROC
- Implementing strategies to tackle overfitting
- Model Testing
- Model Fine-tuning on the CIFAR-10 dataset