Organisation
Project Deep Visual Learning, summer term 2025
Description
The project begins with a basic introduction to deep learning concepts for visual data, covering fundamental architectures such as convolutional neural networks (CNNs) and vision transformers. Students will work with state-of-the-art frameworks like PyTorch to implement models for tasks such as image classification, object detection, and image generation. After the introduction, each student will focus on a dedicated deep learning project, either individually or in small groups. Throughout the project, students will follow a structured workflow, incorporating software engineering principles such as modular code development, reproducibility, and proper documentation. These projects will allow them to dive deeper into a specific application, such as real-world image computation, medical imaging or generative AI. Throughout the project, students will develop practical coding skills, gain experience with large-scale datasets, and learn how to handle the computational challenges of real-world deep learning.
Modalities
In this project students will unlock the power of AI for images, by mastering visual deep learning techniques and their applications. Through hands-on experience, they will learn how to design, train, and evaluate deep learning models for image-based tasks. Key learning objectives include mastering deep neural network architectures, optimizing model performance, handling large-scale image datasets, and exploring advanced topics such as generative models and self-supervised learning. Students will also gain practical experience in real-world setups by working on the VisCom GPU cluster as well as the BwUniCluster, where they will learn how to manage deep learning workloads efficiently. By the end of the project, they will have developed both theoretical knowledge and practical skills essential for applying deep learning in research and industry.
