Senior Researcher
Data61-CSIRO 

Tower A, 7 London Circuit
Canberra ACT 2601
Australia

e-mail
mehrtash (döt) harandi (ät) data61 (döt) csiro (döt) au


About me

I am a member of the Machine Learning Research Group (MLRG) in Data61-CSIRO and a senior adjunct lecturer at ANU, working with prof. Richard Nock, prof. Richard Hartley, prof. Fatih Porikli, Dr. Mathieu Salzmann, Dr. Basura Fernando, Dr. Anoop Cherian and Dr. Piotr Koniusz.
I am interested in various aspects of learning, especially with a flavor of visual data. I regularly review papers for top international conferences (e.g. CVPR, ICCV, NIPS, ECCV, ACCV) and journals (e.g. IEEE TPAMI, IJCV, IEEE TIP, IEEE TNNLS, Pattern Recognition).


Team

  1. Masoud Faraki, PhD Student (2014 - Present)
  2. Samitha Herath, PhD Student (2016 - Present)
  3. Soumava Kumar Roy, PhD Student (2016 - Present)
  4. Yao Lu, PhD student (2017 - )
  5. Tong Zhang, PhD Student  (2017 - )

Teaching
  1. Engineering Data Analytics -  Australian National University (ANU) 2016, 2017 
  2. Individual Project -  Australian National University (ANU) 2015
  3. Signal & Image Processing II - University of Queensland (2010-2012)

News

  • Our paper "Joint Dimensionality Reduction and Metric Learning: A Geometric Take" is accepted to ICML'17. 
  • Our paper "Learning Discriminative Alpha-Beta Divergences for Positive Definite Matrices" is accepted to ICCV'17.  
  • We now have results showing that CNNs (e.g., AlexNet and VGG) with orthogonal layers (orthogonality is imposed on the weights) perform better than their non-orthogonal counterparts even on very large-scale datasets such as imagenet data (check out our paper Generalized BackPropagation Ètude De Cas: Orthogonality) 
  • 2 papers accepted to CVPR'17.
  • Our paper "Dimensionality Reduction on SPD Manifolds: The Emergence of Geometry-Aware Methods" is accepted by TPAMI.
  • Our survey on action recognition (https://arxiv.org/abs/1605.04988) is accepted to Image and Vision Computing.
  • Serving as an Area Chair for WACV 2017.
  • Our paper "Distribution-Matching Embedding for Visual Domain Adaptation" is accepted by JMLR.