About

I am a researcher in the computer vision group of the Friedrich Schiller University of Jena (Germany). I finished my PhD in 2011 supervised by Joachim Denzler and I am heading the visual recognition part of the group now. In 2012/13, I worked as a PostDoc in the computer vision group of ICSI/EECS (UC Berkeley, California).

My research interests are mainly machine learning aspects of computer vision applications, such as transfer learning for object recognition and fast inference for semantic segmentation. If you have any questions concerning my research or a specific detail in my publications, do not hesitate to write an e-mail. 

CVPR 2014 papers

posted Mar 31, 2014, 7:09 AM by Erik Rodner

The group in Jena got two papers accepted for CVPR 2014 and I am a co-author on both of them:

Christoph Göring and Erik Rodner and Alexander Freytag and Joachim Denzler. 
Nonparametric Part Transfer for Fine-grained Recognition
Short Summary: Detection of parts with part transfer and feature extraction at part locations. The approach is super simple but works great!

Daniel Haase and Erik Rodner and Joachim Denzler. 
Instance-weighted Transfer Learning of Active Appearance Models
Short Summary: Domain adaptation techniques for Active Appearance Models, simple approach motivated from importance sampling

TASK-CV Workshop

posted Mar 31, 2014, 7:00 AM by Erik Rodner

Kate Saenko and several other people are organizing a workshop at ECCV 2014 this year about transfer learning and adaptation: http://www.cvc.uab.es/adas/task-cv2014/. I will help with reviewing and the deadline is the 14th of July, so keep this in mind when thinking about a venue for submitting a domain adaptation paper.

NVIDIA hardware gift

posted Jan 17, 2014, 2:53 PM by Erik Rodner

NVIDIA is supporting my research with a hardware gift. In particular, I received a Tesla K40 for intensive parallel processing for big data machine learning. Thanks to NVIDIA and their academic program.

ICRA paper accepted

posted Jan 13, 2014, 3:35 PM by Erik Rodner

A paper together with Daniel, Judy, Kate, and Trevor has been accepted at ICRA today. The paper shows how to quickly build adapted detection models and in our live demonstration, we were able to learn detectors on-the-fly in less than 30 seconds. I will upload the camera ready version and the corresponding video soon.

Journal article in Analytica Chimica Acta accepted

posted Jul 23, 2013, 9:54 AM by Erik Rodner   [ updated Jul 23, 2013, 9:55 AM ]

The article "Automatic Identification of Novel Bacteria using Raman Spectroscopy and Gaussian Processes" together with Michael Kemmler, Petra Rösch, Jürgen Popp, and Joachim Denzler has been accepted for Analytica Chimica Acta. The main topic of this article is novelty detection when dealing with classification of Raman spectra of bacteria. I will publish a draft version of this article soon.

GCPR Active learning paper accepted for oral presentation!

posted Jun 24, 2013, 4:45 PM by Erik Rodner

The paper "Labeling examples that matter: Relevance-Based Active Learning with Gaussian Processes" with Alex Freytag and Joachim Denzler has been accepted for GCPR 2013 as an oral presentation. In this work, we show how to perform active learning for binary classification with GP regression.

Journal article about GP-OCC accepted!

posted Jun 3, 2013, 9:11 PM by Erik Rodner   [ updated Jun 24, 2013, 5:01 PM ]

My journal article "Nonparametric One-class Classification with Gaussian Processes" with Michael Kemmler, Esther Wacker, and Joachim Denzler has been accepted for publication in the journal "Pattern Recognition". 

ICLR presentation

posted May 30, 2013, 4:07 PM by Erik Rodner

The ICLR presentation of Judy about our domain adaptation method is online now, it also gives a short introduction to visual domain adaptation in general: http://techtalks.tv/talks/efficient-learning-of-domain-invariant-image-representations/58086/

Final versions for CVPR and SCIA

posted Apr 16, 2013, 11:06 AM by Erik Rodner   [ updated Apr 16, 2013, 11:11 AM ]

CVPR 2013
Jeff Donahue, Judy Hoffman, Erik Rodner, Kate Saenko, Trevor DarrellSemi-Supervised Domain Adaptation with Instance Constraints
Paul Bodesheim, Alexander Freytag, Erik Rodner, Michael Kemmler, Joachim Denzler. Kernel Null Space Methods for Novelty Detection

SCIA 2013
Paul Bodesheim, Alexander Freytag, Erik Rodner and Joachim DenzlerApproximations of Gaussian Process Uncertainties for Visual Recognition Problems. Scandinavian Conference on Image Analysis (SCIA 2013)

ICLR Paper accepted as Oral

posted Apr 4, 2013, 1:01 PM by Erik Rodner

The ICLR 2013 paper "Efficient Learning of Domain-invariant Image Representations" with Judy, Jeff, Kate, and Trevor has been accepted for oral presentation. Judy also already released the source code for the algorithm, so feel free to use it for your research.

1-10 of 26

Contact and Profiles