This is the repository for ECE6254 final project. It is also my first humble attempt to realize something in Python. The goal of this project is to implement the convolutional network developed by [VGG group] (http://www.robots.ox.ac.uk/~vgg/research/very_deep/). We also change the layer structures and compare the performance among all variations. The dataset we choose to apply the CNN to is [The CIFAR-10 dataset] (https://www.cs.toronto.edu/~kriz/learning-features-2009-TR.pdf) due to the computational limit of CPUs. The code skeleton is based on this [Jupyter Notebook] (http://nbviewer.jupyter.org/github/julienr/ipynb_playground/blob/master/keras/convmnist/keras_cnn_mnist.ipynb). Plus, you can see the [original paper] (http://arxiv.org/pdf/1409.1556.pdf) for more details.
What's included:
- small VGGNet implementation
- visualization of weights and layer outputs
- test accuracy and prediction probability
- classification performance evaluation (confusion matrices and ROC/AUC)
Team members: [Chia-Sheng Hsu] (https://www.linkedin.com/in/chiasheng), Mythili Vadali, Jingting Yao, Yi-Chi Shao
Instructor: [Prof. Justin Romberg] (http://users.ece.gatech.edu/justin/Justin_Romberg.html)