ECIR2025 Tutorial: Advanced Methods for Visual Information Retrieval and Exploration in Large Multimedia Collections

This tutorial, presented by Prof. Dr. Kai Barthel, M.Sc. Nico Hezel, and M.Sc. Konstantin Schall from the Visual Computing Group at HTW Berlin, will take place at ECIR 2025.

At the time of the conference, our Jupyter notebooks, along with links to the relevant papers and data, will be available here.

Overview

• Introduction
• Presenters
• Format and Schedule

Introduction

This tutorial presents advanced methods for efficiently searching and exploring large visual datasets, addressing the increasing demand for high-performance visual retrieval systems as multimedia content grows exponentially.

We will begin by covering the principles of large visual encoders, emphasizing techniques for generating and improving compact, high-quality general-purpose visual descriptors. Next, we will explore strategies to enhance cross-modal retrieval. Participants will gain insights into approximate nearest neighbor search methods, with a focus on graph-based approaches that maximize search efficiency in dynamic datasets.

In addition, the tutorial introduces innovative visualization techniques for high-dimensional data, such as grid-based sorting, which enable intuitive navigation and exploration of extensive image collections.

Hands-on exercises provide practical experience with the concepts discussed using Jupyter notebooks and interactive demonstrations. This tutorial will equip researchers, practitioners, and students in information retrieval, computer vision, and data science with both basic and advanced skills to efficiently search and explore large visual datasets.

Presenters

Kai Uwe Barthel

Kai Uwe Barthel is a professor at the Institute for Media and Computing at HTW Berlin, heading the Visual Computing Group. His work focuses on technologies for simplifying media retrieval, including image understanding, retrieval, metric learning, and visual exploration. During his PhD at TU Berlin, he specialized in fractal image compression and later led a 3D-video coding research project. As head of R&D at N-Tec Media and LuraTech Inc., he contributed to the JPEG2000 standard, developing advanced image and video compression solutions. Since 2001, he has taught image analysis, machine learning, and information retrieval at HTW Berlin. In 2009, he founded pixolution, a company specializing in visual image search. Prof. Barthel holds numerous patents, publications, and awards.

Nico Hezel

Nico Hezel is a researcher in the Visual Computing Group at HTW Berlin, focusing on content-based information retrieval and large-scale image analysis. He is pursuing a PhD at Humboldt University, developing a "Dynamic Exploration Graph" for efficient image search in evolving datasets. His innovations earned the Best Demo Award at ICMR 2018 and ACM Multimedia 2019 for projects enhancing search capabilities. In 2024, he won the SISAP Indexing Challenge for advancements in graph-based nearest neighbor search. Hezel also lectures on machine learning, computer vision, and image analysis.

Konstantin Schall

Konstantin Schall researches general-purpose retrieval networks within HTW Berlin’s Visual Computing Group, focusing on adaptable retrieval applications. Since 2024, he has been a Principal Machine Learning Scientist at Mediaopt GmbH, leading product retrieval for online shopping. As a PhD candidate at Humboldt University, his work has received awards such as the Best Demo Award at ACM Multimedia and two Best Video Browser System Awards at the Video Browser Showdown (2022, 2023). He also teaches machine learning and information retrieval.

Format and Schedule

This tutorial will take half a day (a total of 3 hours plus breaks) and will be presented in person. We propose the following program:

1. Introduction

• Welcome, collect participant information
• Overview of tutorial goals & schedule
• Motivation: Challenges in visual information retrieval
• Motivation: Challenges in visualizing high-dimensional data
• Access to Jupyter notebook code examples via GitHub

2. Image Encoders for Effective Retrieval

• Overview of image encoding techniques for retrieval
• Optimizing CLIP image encoders
• Enabling cross-modal and general-purpose retrieval
• Evaluating retrieval embedding models

3. Approximate Nearest Neighbor Search (ANNS) in Dynamic Datasets

• ANNS techniques and challenges in high-dimensional spaces
• Graph-based ANNS methods
• Dynamic graph structures for ANNS

4. Visual Exploration

• Challenges of high-dimensional data visualization
• Dimensionality reduction methods for visualization
• Grid-based visual sorting
• Visual exploration techniques for interactive image search

5. Conclusion

• Summary of techniques for effective visualization, search, and exploration of high-dimensional image data
• Final Q&A and recap of key tutorial points