COMP0197: Applied Deep Learning

UCL Module | CS | UCL Moodle Page

Term 2, Academic Year 2024-25

1. Development environment

The module tutorials (see bellow) and coursework use Python, NumPy and an option between TensorFlow and PyTorch. The Development environment document contains details of the supported development environment, though it is not mandatory.

2. Tutorials

Quick start

To run the tutorial examples, follow the instruction below.

First, set up the conda environments:

conda create --name comp0197_pt python=3.11 -y
conda activate comp0197_pt
conda install pytorch torchvision cpuonly -c pytorch -y

conda create --name comp0197_tf python=3.11 -y
conda activate comp0197_tf
pip install tensorflow-cpu pillow

Additional libraries and/or data required for individual tutorials are specified in the readme file in each tutorial directory.

Scripts with "_tf" and "_pt" postfix are using TensorFlow 2 and PyTorch, respectively.

All visual examples will be saved in files, without requiring graphics.

Then, change directory cd to each individual tutorial folders and run individual training scripts, e.g.:

conda activate comp0197_pt
python train_pt.py   

or

conda activate comp0197_tf
python train_tf.py  

Convolutional neural networks

Image classification
Image segmentation

Recurrent neural networks

Text classification
Character generation

Variational autoencoder

MNIST generation

Generative adversarial networks

Face image simulation

3. Reading list

A collection of books and research papers is provided in the Reading List.