Claudio Michaelis

One-shot learning and robustness of deep learning models.

Computer Vision for the Amazon Scout delivery robot.

Developed a new task for few-shot object detection in the lab of Rich Zemel.

Developed the first prototype of a holographic projector for two-photon microscopy that can control light of two wavelengths at once.

Developed the algorithms for Holo4D, a holographic projector for two-photonmicroscopy.

Stand-in lead for the Holo4D project.

Thesis on one-shot learning and robustness of deep learning models in the groups of Matthias Bethge & Alexander Ecker. Figured out that deep learning models tend to take shortcuts (https://www.nature.com/articles/s42256-020-00257-z), such as CNNs classifying images by their texture (https://openreview.net/forum?id=Bygh9j09KX). Built an object detector that can detect almost any object from a single example (https://arxiv.org/abs/2011.04267).

Thesis developing a holographic multi photon microscopy system for biological research in the labs of Alfred Leitenstorfer (Physics) & Elisa May (Bioogy).

Bachelor Thesis developing a tunable diode laser in the group of Dominik Schneble