Main Content

Advanced Topics in Computer Vision and Pattern Recognition (Seminar)

This page is currently under construction. Detailed course information will be added in the coming weeks. In the meantime, please refer to the university course catalog. 

Course Information

Instructor: Alexander Krawczyk

Introduction session: October 16, 3 p.m.; Lecture Hall V (HS V, MZG) 04A23 (HS V A4) (Seminarraum)

Block seminar: 2 days in February/March 2025 (presence mandatory)

Target Audience

Bachelor/Master students in Computer Science, and Data Science (maximum 14 participants)

Prerequisites: Students are strongly encouraged to have completed or be enrolled in an introductory course in machine learning and/or computer vision.

Course Overview

Visual perception is a fundamental human ability, enabling us to navigate safely through our environment and make informed decisions. Computer Vision (CV) aims to replicate this ability in machines, enabling them to understand and gain insights from digital images and videos. By identifying meaningful patterns in visual data, CV enhances automated planning and decision-making. Today, a wide range of deep-learning architectures enables complex tasks like object detection and image generation, as well as fusing visual data with other modalities such as text prompts.

The Master's-level seminar, "Advanced Topics in Computer Vision and Pattern Recognition," allows students to explore and discuss the latest trends and state-of-the-art methods for solving challenging computer vision tasks. This seminar provides students with an opportunity to delve deeply into the latest advances in the field.

Key topics include:

  • Important Machine Learning architectures such as: Convolutional Neural Networks, Diffusion Models, Vision-Language Models (VLMs), Vision-Language-Action (VLA) models
  • Application areas:
    • Visual Question Answering (VQA),
    • Human-Centric Vision (pose estimation and activity recognition),
    • Object recognition &  tracking, and
    • Semantic segmentation,
    • Image and video synthesis,
    • Embodied AI (robotics and intelligent agents).

Qualification:
The seminar is designed to sharpen essential academic skills. Students will learn to synthesize relevant information in order to understand key trends, potential impacts, and the strengths and weaknesses of modern approaches to computer vision, as well as identify open research questions. Through active discussions with their peers and presentations of their own research on these topics, students will improve their ability to present and communicate advanced scientific content to a broad audience. These activities prepare students for independent research and advanced work in computer vision.

Format:
The course is a research-focused seminar based on student participation. Each student is responsible for presenting a recent research paper and leading the subsequent discussion. Students will also be required to write a final term paper, which will provide an opportunity to conduct in-depth research on a topic of interest.