For Project 2 of the Udacity Deep Learning Nanodegree, I built a Convolutional Neural Network (i.e. CNN or ConvNet) for Image Classification using TensorFlow.
The images used are from the CIFAR-10 dataset and consist of 10 different objects. Some of the steps I took in my network include:
PREPROCESSING: normalization, one-hot encoding
NETWORK ARCHITECTURE: multi-layer stack including conv layers, fully connected layers, maxpooling, ReLu activation, flattening, dropouts
HYPERPARAMETERS: epochs, batch size, keep_prob
The network is built on TensorFlow and code is written in Python (v3) and is presented via Jupyter Notebook. The network was trained via the cloud using an AWS EC2 instance.