Sunil Khatri Receives the DAGM Best Master's Thesis Award at the GCPR

Sunil Khatri received the Best Master’s Thesis Award from the German Association for Pattern Recognition (DAGM). The award ceremony took place during the German Conference on Pattern Recognition (GCPR) 2026. The award recognizes his outstanding master's thesis, which was supervised by Simon Reiß, Steven Landgraf, and Markus Ulrich.

In his thesis, Sunil Khatri investigated how various image processing tasks can be solved using only a few examples, without adjusting the model parameters. This approach, known as “in-context learning,” has become particularly well-known through large language models. To date, large, computationally intensive models have often been used to apply this concept to image data.

With “Generalizer,” he investigated and adapted a compact approach—based entirely on convolution operations—for various image processing tasks. These include semantic segmentation, filling in missing image regions, and removing rain from images. With just 1.27 million parameters, the model achieved an inference time of approximately 22 milliseconds per run in the measurements conducted. The results demonstrate the potential of small models to handle various visual tasks with significantly lower computational effort.

We are delighted by this recognition of his scientific achievement. Congratulations, Sunil, on this well-deserved award!