The award is given for fundamental work on learning powerful representations for image and video recognition from reduced supervision.
Anna Kukleva (right) receives the Otto Hahn Medal from Prof. Dr. Claudia Felser, Director at the MPI for Chemical Physics of Solids (left). © David Ausserhofer / Max-Planck-Gesellschaft
Dr. Anna Kukleva, postdoc in the “Computer Vision and Machine Learning” department led by Director Bernt Schiele, has been awarded the Otto Hahn Medal of the Max Planck Society. Every year since 1978, the Max Planck Society has awarded the Otto Hahn Medal to up to 30 young researchers for outstanding scientific achievements, mostly in connection with their doctorate. Anna Kukleva received the award during the Annual Meeting of the Max Planck Society in Magdeburg.
Anna Kukleva's thesis with the title "Advancing Image and Video Recognition with Less Supervision" addresses challenges caused by the demand for large and supervised datasets. The award citation states: “Through this groundbreaking research, she explores the frontiers of learning with minimal supervision, revolutionizing how we understand and train models for image and video recognition. The work delves into self-supervised and unsupervised methods, demonstrating how to exploit the data without relying on labels. Moreover, this thesis makes substantial strides in reducing the costs and efforts associated with data annotation by proposing methods that omit precise annotations in multimodal learning scenarios. Last but not least, this research extends into open-world scenarios, where models are required to generalize beyond pre-defined classes through novel methods for vision-language model adaptation.”
In January 2020, Anna Kukleva began her doctoral studies at Saarland University and the Max Planck Institute for Informatics, completing them in August 2024 with “summa cum laude.” Since September 2024, she has been a postdoctoral researcher in Bernt Schiele’s department. The core of her research lies in algorithmic processing of visual data to identify elements like objects, actions, or scenes, and understand learning dynamics of the algorithms. Rather than relying entirely on large, human-annotated datasets, she explores how computers can largely learn patterns independently (self-supervised learning), with minimal assistance (semi-supervised learning), and, in rare cases, with limited amounts of fully annotated data (fully-supervised learning). A central aspect of her work is to examine how well these learning methods can be transferred when only a few examples are available, or when dealing with unknown scenarios not included during training.
Further Information:
Anna Kukleva’s personal website: https://annusha.github.io/
Dissertation: https://publikationen.sulb.uni-saarland.de/handle/20.500.11880/38297
Website of the department “Computer Vision and Machine Learning”:
https://www.mpi-inf.mpg.de/de/departments/computer-vision-and-machine-learning
Editor:
Philipp Zapf-Schramm
Max Planck Institute for Informatics
Phone: +49 681 9325 5409
Email: pzs@mpi-inf.mpg.de