Portrait at TENSS

Welcome, I am Alexander Mathis. My main research interests comprise the neural basis of olfactory and motor behaviors. More specifically, I am interested in odor-guided navigation, pose estimation as well as motor control and adaptation.

A trail tracking mouse, whose pose is estimated with DeepLabCut.

I work as a postdoctoral fellow in the group of Matthias Bethge at the Bernstein Center for Computational Neuroscience in Tuebingen and the University of Tuebingen and the lab of Venkatesh N. Murthy at the Department of Molecular and Cellular Biology at Harvard University. My work in these labs was funded by a DFG postdoctoral fellowship as well as a Marie-Curie fellowship (Department news) .

November 2018 Together with Tanmay Nath, Matthias Behtge and Mackenzie Mathis, I devoloped a Python package for DeepLabCut - now also available on pypi. It has GUIs for labeling and allows active learning. We also showed how to use DLC for analyzing 3D poses of Cheetah's (with Amir Patel and An Chi Chen) (see Preprints).
October 2018 Together with Rick Warren, I substantially improved the inference speed of DeepLabCut and benchmarked it. Furthermore, we showed that the inference of poses is very robust to video compression (see Preprints).
August 2018 DeepLabCut just appeared in Nature Neuroscience (see publications). Kunlin Wei and Konrad Kording wrote a nice news & views! DeepLabCut was also covered by the NVIDIA AI developer blog and the Harvard Gazette
June 2018 We are thrilled that several labs are already using DeepLabCut and have shared really cool results with us - we will share some of the example behaviors featuring race horses, rat reaching, fish, & mouse locomtion on the project website .
April 2018 I am very excited to present DeepLabCut & a poster on odor guided navigation at the Federation of European Neuroscience Conference in Berlin - I hope to see you in July (see Abstracts)
April 2018 I was awarded a NVIDIA hardware grant - thanks, it will be very helpful for my current work!
April 2018 We finally shared DeepLabCut, an open source tool for pose estimation based on very little labeled data. We demonstrate the versatility of this framework by tracking various body parts in a broad collection of experimental settings: mice odor trail-tracking, egg-laying behavior in drosophila, and mouse hand articulation in a skilled forelimb task (see Preprints).
April 2018 All code for DeepLabCut is available on my Github account -- happy tracking!
November 2017 A collaborative study between Dulac, Schnitzer & Murthy labs on coding of social stimuli in the amygdala was just published (see publications).
September 2017 Together with Andreas Herz and Martin Stemmler I wrote a review covering conceptual spaces, hierarchical codes, population vectors, ... (see reviews).