Readspeaker menu

Dr. rer. nat. Björn Barz

Former PostDoc Researcher
Björn Barz

Contact

Address: Carl Zeiss AG
Corporate Research & Technology
Jena
Germany
Phone: +49 (0) 3641 9 46416
Room: 1212
E-mail: bjoern (dot) barz (at) uni-jena (dot) de
GitHub: GitHub Callidior

Curriculum Vitæ

since 2021 Machine Learning Scientist
Carl Zeiss AG
Corporate Research & Technology
2020-2021 PostDoc Researcher
Friedrich Schiller University Jena, Computer Vision Group
Research focus: Integration of prior knowledge
2016-2020 PhD Student
Friedrich Schiller University Jena, Computer Vision Group
PhD Thesis: "Semantic and Interactive Content-based Image Retrieval"
2014-2016 M.Sc. in Computer Science
Friedrich Schiller University Jena
Master's Thesis: "Detection of Anomalous Intervals in Multivariate Spatio-Temporal Time-Series"
2011-2014 B.Sc. in Computer Science
Friedrich Schiller University Jena
Bachelor's Thesis: "Interactive Learning of Object Detectors"

Research Interests

  • Deep Learning
  • Image Retrieval
  • Metric Learning
  • Active Learning
  • Object Detection
  • Natural Language Processing

Further Activities

Projects

Semantic and Interactive Content-based Image Retrieval


Overview of major publications from 2018-2020 (click to open as PDF).


Hierarchy-based Semantic Image Embeddings

Learn to embed images and classes into a semantic space where the dot product corresponds to the semantic similarity.

Interview with Björn Barz (in German) about this work in the German radio Deutschlandfunk (March 23rd, 2019):
https://ondemand-mp3.dradio.de/file/dradio/2019/03/23/wie_die_bildersuche_mit_bildern_noch_besser_wird_interview_dlf_20190323_1644_c6dd615b.mp3


Detecting Anomalous Regions in Multivariate Spatio-Temporal Time-Series

The "libmaxdiv" library (C++/Python) implements the "Maximally Divergent Intervals (MDI)" algorithm for detection of anomalous regions in multivariate spatio-temporal time-series.

https://cvjena.github.io/libmaxdiv/

Maximally Divergent Intervals


ARTOS - Adaptive Real-Time Object Detection System

Conveniently learn models for visual object detections with a few clicks - in just a few seconds!

https://cvjena.github.io/artos/

ARTOS GUI

Summer term 2021

Winter term 2020/21

Summer term 2020

Supervised Theses

  • Adversarial Domain Adaptation for Fine-Grained Recognition (Bernd Gruner, Master's thesis, November 2020)
  • Identification in Wildlife Monitoring (Matthias Körschens, Master's thesis, May 2018)

Publications