MASTER-PRAKTIKUM: DEEP LEARNING FOR COMPUTER VISION AND NLP (211428-SoSe23)

This praktikum provides the opportunity to practice the development of neural network-based methods for machine learning tasks. In this praktikum, the focus will be on some general-domain vision and/or NLP tasks of varying complexity. Initial basic tasks will ensure a basic level of proficiency while a final mini-project will test your skills on a more complicated and possibly cutting-edge problem. The course requires knowledge of Python programming, and at least theoretical understanding of basic deep learning.
The course will start with an introductory session that presents the praktikum in more detail, includes a short introduction to Pytorch and some of its ecosystem, which we will be using for the projects. The students will form groups of up to three members to work on their projects. We aim to develop multiple mini-projects of varying complexity over the course of the semester, with a few deadlines throughout the semester. We will have regular in-person meetings to discuss progress and clear problems.
The grades are given mainly based on reports delivered by the deadlines throughout the semester, as well as accompanying intermediate presentations or discussions, a final presentation at the end of the semester, trackable code contributions and performance of the developed methods. We will host the code publicly on Github or similar.

Requirements:
Minimum requirements: practical programming experience in Python. Knowledge of deep learning (neural networks), as well as knowledge of basic machine learning concepts.
Recommended: experience with Latex, Git, Pycharm or other IDEs