Mateusz Malinowski

Mateusz Malinowski is a researcher at the Max Planck Institute for Informatics, part of the Saarland Informatics Campus, working in the Computer Vision and Machine Learning department. His work focuses on the intersection of visual understanding and language processing, with particular emphasis on developing systems that can answer questions about visual content.

He received his Master's Degree with Honors in Computer Science from Saarland University in Germany, and completed his undergraduate studies in Computer Science at the University of Wrocław in Poland. His doctoral research at Universität des Saarlandes (completed in 2017) laid foundational work for visual question answering systems.

Malinowski's research interests center around the synergy between machine vision and natural language understanding, with specific focus on visual question answering, text-to-image retrieval, and spatial reasoning. His work has pioneered approaches to understanding real-world images through questioning, developing the DAQUAR dataset (the first question-answering dataset about real-world images) and neural architectures that can interpret both visual and linguistic inputs. His research bridges computer vision, natural language processing, and cognitive science to create more holistic AI systems.

His publication record shows a clear progression in visual question answering research, evolving from foundational work on spatial relations and pooling methods (2013-2014) to sophisticated neural architectures for visual question answering (2015-2017), and further to more complex tasks like visual what-if questioning and boundary prediction (2018-2019). The research consistently demonstrates innovation in multimodal learning, with increasing sophistication in handling the interplay between visual and linguistic information.

Malinowski has served as a reviewer for top-tier conferences including NIPS, CVPR, ECCV, and journals such as TPAMI and IJCV, demonstrating his standing in the computer vision and machine learning communities.

He has advised Master's students including Ashkan Mokarian (2016) on deep learning for image captioning and Sreyasi Nag Chowdhury (2015) on contextual media retrieval. He has also contributed to teaching as a teaching assistant for Deep Learning Seminar (2015) and Probabilistic Graphical Models (2013) courses.

His work is associated with the Scalable Learning and Perception group at the Max Planck Institute, where he collaborates with researchers including Mario Fritz, Marcus Rohrbach, and Bernt Schiele on advancing the state of the art in computer vision and multimodal learning.