Classification and Retrieval of Images

Piotr Koniusz

pk.jpgI am a senior researcher in MLRG at Data61/CSIRO (former NICTA). I am also a senior honorary lecturer at Australian National University (ANU). Previously, I worked as a post-doctoral researcher in the team LEAR, INRIA, France. I received my B.Sc. degree in Telecommunications and Software Engineering in 2004 from the Warsaw University of Technology, Poland, and completed my PhD degree in Computer Vision in 2013 at CVSSP, University of Surrey, UK.

My interests include visual concept detection, visual category recognition, action recognition, zero-, one- and few shot learning, domain adaptation, image-to-image translation, feature and representation learning, invariance learning and understanding, feature pooling, spectral learning and graphs, as well as tensor, kernel methods, linearisations, sparsity and deep learning methods.

If you are interested in PhD studies at the Australian National University and/or Data61/CSIRO, read here. If you are an ANU student pursuing BSc/MSc (etc.) looking for honours projects, I do not normally accept any students with perhaps an exception outlined here.

Latest News

  1. 26.07.2019. One ICCV'19, one UAI'19, two CVPR'19, two WACV'19, one IJCV and one TIP papers accepted.
    Congrats Lei Wang, Fatemeh Shiri and Hongguang Zhang!
  2. 01.05.2019. I am organising an ICCV 2019 tutorial entitled Second- and Higher-order Representations in Computer Vision with my colleagues
    Mehrtash Harandi, Lei Wang and Ruiping Wang.
  3. 01.05.2019. PK will be serving as an Area Chair for ICLR 2020, this year I have served as an AC for ICLR 2019, WACV 2019 , ICCV 2019 and BMVC 2019.
  4. 01.05.2019. I am the main advisor of four PhD students at CECS/ANU: Lei Wang, Fatima Shiri, Yusuf Tas and Hongguang Zhang.
    I am co-supervising (second supervisor) Christian Simon (ANU) with Dr. Mehrtash Harandi (Monash), Arian Prabowo and
    Xianjing Wang with A/Prof. Flora Salim (RMIT), and Sheila Caceres (UoS) with Prof. Fabio Ramos (UoS).
  5. 23.10.2018. I am organising an ACCV 2018 workshop and a mini-challenge entitled
    Museum Exhibit Identification Challenge (Open MIC) for Domain Adaptation and Few-Shot Learning.
  6. 11.09.2018. The Open MIC dataset is available for non-commercial academic research Open MIC Dataset (ECCV 2018).
  7. 04.07.2018. Two ECCV'18 (oral ~2% acceptance rate/poster) and one BMVC'18 (spotlight ~6% acceptance rate) papers accepted.
    Congrats Yusuf Tas and Tsung-Yu Lin!
  8. 24.02.2018. Two CVPR'18 and one WACV'18 papers accepted. Congrats Fatemeh Shiri and Hongguanh Zhang!
  9. 23.10.2017. PK listed as an Outstanding Reviewer by ICCV 2017 Awards Page (23.10.2017), CVPR 2017 Awards Page (20.06.2017),
    ECCV 2016 Awards Page (21.10.2016), CVPR 2015 Awards Page (04.05.2015), and BMVC 2012 Awards Page (05.09.2012)
  10. 09.12.2016. I am organising a full-day CVPR 2017 workshop entitled Tensor Methods in Computer Vision with my colleagues
    Anoop Cherian and Fatih Porikli.
  11. 14.06.2016. Beyond Covariance: Higher-order Tensor Descriptors and Applications in Computer Vision, (Slides), P. Koniusz,
    Talk at Australian National University
  12. 01.03.2013. Novel Image Representations for Visual Categorisation with Bag-of-Words, (Full Paper), P. Koniusz,
    PhD Dissertation (supervised by Dr. K. Mikolajczyk, reviewed by Prof. M. Bober and Prof. Theo Gevers)
  13. 11.09.2010. Our classification system SURREY_MK_KDA has scored the highest Mean Average Precision (MAP) of 62.15%
    (average over APs of all 9 concepts) amongst all competing approaches in the PASCAL VOC2010 Action Classification
    Teaser Challenge, Overview and results of the action classification taster challenge

Conference Publications

  1. thumb_uai2019.jpg Fisher-Bures Adversary Graph Convolutional Networks,
    Ke Sun, Piotr Koniusz, Jeff Wang,
    Conference on Uncertainty in Artificial Intelligence (UAI), 2019,
    (Full Paper, Full Paper /ArXiv/, ~26% acceptance rate)
  2. thumb_iccv2019.jpg Hallucinating Bag-of-Words and Fisher Vector IDT terms for CNN-based Action Recognition,
    Lei Wang, Piotr Koniusz, Du Q. Huynh,
    International Conference on Computer Vision (ICCV), 2019,
    (Full Paper, Full Paper /ArXiv/)
  3. thumb_cvpr2019a.jpg Few-Shot Learning via Saliency-guided Hallucination of Samples,
    Hongguang Zhang, Jing Zhang, Piotr Koniusz,
    Computer Vision and Pattern Recognition (CVPR), 2019,
    (Full Paper, Full Paper /ArXiv/)
  4. thumb_cvpr2019b.jpg Deep Image Deblurring with Multi-Patch Hierarchical Network at 30fps,
    Hongguang Zhang, Yuchao Dai, Hongdong Li, Piotr Koniusz,
    Computer Vision and Pattern Recognition (CVPR), 2019,
    (Full Paper, Full Paper /ArXiv/)
  5. thumb_wavc2019a.jpg Power Normalizing Second-order Similarity Network for Few-shot Learning,
    Hongguang Zhang, Piotr Koniusz, Winter Conference on Applications of Computer Vision (WACV), 2019,
    (Full Paper, Full Paper /ArXiv/)
  6. thumb_wavc2019b.jpg Recovering Faces from Portraits with Auxiliary Facial Attributes,
    Fatemeh Shiri, Xin Yu, Fatih Porikli, Richard Hartley, Piotr Koniusz,
    Winter Conference on Applications of Computer Vision (WACV), 2019, (Full Paper, Full Paper /ArXiv/)
  7. thumb_eccv2018c.jpg Model Selection for Generalized Zero-shot Learning, Hongguang Zhang, Piotr Koniusz,
    European Conference on Computer Vision (ECCV) Workshop: TASK-CV, 2018, (Full Paper, Full Paper /ArXiv/)
  8. thumb_bmvc2018b.jpg CNN-based Action Recognition and Supervised Domain Adaptation on 3D Body Skeletons via Kernel Feature Maps,
    Yusuf Tas, Piotr Koniusz, British Machine Vision Conference (BMVC), 2018,
    (Full Paper, Full Paper /ArXiv/, Full Paper /BMVA/, spotlight ~6% acceptance rate)
  9. thumb_eccv2018b.jpg Museum Exhibit Identification Challenge for Domain Adaptation and Beyond,
    Piotr Koniusz, Yusuf Tas, Hongguang Zhang, Mehrtash Harandi, Fatih Porikli, Rui Zhang,
    European Conference on Computer Vision (ECCV), 2018,
    (Full Paper, Full Paper /ArXiv/, Full Paper /CVF Open Access/, ECCV'18 Talk /YouTube/, oral ~2% acceptance rate)
    The Open MIC dataset is available for non-commercial academic research Open MIC Dataset (ECCV 2018).
  10. thumb_eccv2018a.jpg Second-order Democratic Aggregation, Tsung-Yu Lin, Subhransu Maji, Piotr Koniusz,
    European Conference on Computer Vision (ECCV), 2018, (Full Paper /ArXiv/, Full Paper /CVF Open Access/)
  11. thumb_cvpr2018b.jpg Zero-Shot Kernel Learning, Hongguang Zhang, Piotr Koniusz,
    Computer Vision and Pattern Recognition (CVPR), 2018, (Full Paper, Full Paper /ArXiv/)
  12. thumb_cvpr2018c.jpg A Deeper Look at Power Normalizations, Piotr Koniusz, Hongguang Zhang, Fatih Porikli,
    Computer Vision and Pattern Recognition (CVPR), 2018, (Full Paper, Full Paper /ArXiv/)
  13. thumb_wacv2018.jpg Identity-preserving Face Recovery from Portraits,
    Fatemeh Shiri, Xin Yu, Fatih Porikli, Richard Hartley, Piotr Koniusz
    Winter Conference on Applications of Computer Vision (WACV), 2018, (Full Paper /ArXiv/)
  14. thumb_dicta2017.jpg Face Destylization, Fatemeh Shiri, Xin Yu, Piotr Koniusz, Fatih Porikli,
    The International Conference on Digital Image Computing: Techniques and Applications (DICTA), 2017, (Full Paper /ArXiv/)
  15. thumb_cvpr2017.jpg Domain Adaptation by Mixture of Alignments of Second- or Higher-Order Scatter Tensors, P. Koniusz, Y. Tas, F. Porikli
    Computer Vision and Pattern Recognition (CVPR), 2017, (Full Paper, Full Paper /ArXiv/)
  16. thumb_wacv2017.jpg Higher-order Pooling of CNN Features via Kernel Linearization for Action Recognition, A. Cherian, P. Koniusz, S. Gould,
    Winter Conference on Applications of Computer Vision (WACV), 2017, (Full Paper /ArXiv/)
  17. thumb_mw2017.jpg Artwork Identification from Wearable Camera Images for Enhancing Experience of Museum Audiences,
    R. Zhang, Y. Tas, P. Koniusz, Museums and the Web (MW), 2017 (acceptance rate 25-33%), (Full Paper, Full Paper /ArXiv/)
  18. thumb_eccv2016.jpg Tensor Representations via Kernel Linearization for Action Recognition from 3D Skeletons, P. Koniusz, A. Cherian, F. Porikli
    European Conference on Computer Vision 2016, (Full Paper, Full Paper /ArXiv/, Poster)
  19. thumb_cvpr2016.jpg Sparse Coding for Third-order Super-symmetric Tensor Descriptors with Application to Texture Recognition,
    P. Koniusz, A. Cherian, Computer Vision and Pattern Recognition 2016 (spotlight>poster),
    (Full Paper, Spotlight, Poster, ANU Talk)
  20. thumb_nips2014.jpg Convolutional Kernel Networks, Julien Mairal, Piotr Koniusz, Zaid Harchaoui, Cordelia Schmid,
    In Proc. of the Advances in Neural Information Processing Systems 2014 (spotlight>poster)
    (Full Paper /HAL-INRIA/)
  21. thumb_icip2011a.jpg Spatial Coordinate Coding To Reduce Histogram Representations, Dominant Angle And Colour Pyramid Match,
    Piotr Koniusz, Krystian Mikolajczyk, In Proc. of the International Conference on Image Processing 2011
    (Full Paper, Talk)
  22. thumb_icip2011b.jpg Soft Assignment Of Visual Words As Linear Coordinate Coding And Optimisation Of Its Reconstruction Error,
    Piotr Koniusz, Krystian Mikolajczyk, In Proc. of the International Conference on Image Processing 2011
    (Full Paper, Poster)
  23. thumb_icpr2010.jpg On a Quest for Image Descriptors Based on Unsupervised Segmentation Maps,
    Piotr Koniusz, Krystian Mikolajczyk, In Proc. of the International Conference on Pattern Recognition 2010
    (Full Paper, Talk)
  24. thumb_bmvc2009.jpg Segmentation Based Interest Points and Evaluation of Unsupervised Image Segmentation Methods,
    Piotr Koniusz, Krystian Mikolajczyk, In Proc. of the British Machine Vision Conference 2009
    (Full Paper, Poster, Teaser)

Journal Publications

  1. thumb_ijcv2019.jpg Identity-preserving Face Recovery from Stylized Portraits,
    Fatemeh Shiri, Xin Yu, Fatih Porikli, Richard Hartley, Piotr Koniusz
    International Journal of Computer Vision (IJCV), 2019, (Full Paper, Full Paper /ArXiv/)
  2. thumb_tip2019.jpg A Comparative Review of Recent Kinect-basedAction Recognition Algorithms,
    Lei Wang, Du Q. Huynh, Piotr Koniusz
    IEEE Transactions on Image Processing (TIP), 2019, (Full Paper, Full Paper /ArXiv/, accepted)
  3. thumb_pami2016.jpg Higher-order Occurrence Pooling for Bags-of-Words: Visual Concept Detection, P. Koniusz, F. Yan, P. H. Gosselin, K. Mikolajczyk,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 2016 (accepted), (Full Paper)
  4. thumb_pami2013.jpg Higher-order Occurrence Pooling on Mid- and Low-level Features: Visual Concept Detection, P. Koniusz,
    F. Yan, P. H. Gosselin, K. Mikolajczyk, Technical Report 2013, (Full Paper, Full Paper /HAL-INRIA/)
  5. thumb_cviu2012.jpg Comparison of Mid-Level Feature Coding Approaches And Pooling Strategies in Visual Concept Detection,
    Piotr Koniusz, Fei Yan, Krystian Mikolajczyk, Computer Vision and Image Understanding 2012
    (Full Paper, Original Draft, More Results /Pascal VOC07/, Full Paper /CVIU Version/, BibTex)
  6. thumb_tmm2013a.jpg Robust Multi-Speaker Tracking via Dictionary Learning and Identity Modelling, M. Barnard, P. Koniusz, W. Wang,
    J. Kittler, S. M. Naqvi, J. Chambers, IEEE Transactions on Multimedia 2013 (Full Paper /TMM Version/)
  7. thumb_tmm2013b.jpg A Robust and Scalable Visual Category and Action Recognition System using Kernel Discriminant Analysis
    with Spectral Regression, M. A. Tahir, F. Yan, P. Koniusz, M. Awais, M. Barnard, K. Mikolajczyk, A. Bouridane,
    J. Kittler, IEEE Transactions on Multimedia 2013 (Full Paper /TMM Version/)

ArXiv/Preprints

  1. Hallucinating Bag-of-Words and Fisher Vector IDT terms for CNN-based Action Recognition,
    Lei Wang, Piotr Koniusz, Du Q. Huynh,
    ArXiv Preprint 2019, (Full Paper /ArXiv/)
  2. Multi-Label Propagation Networks for Multi-Label Few-Shot Learning,
    Christian Simon, Piotr Koniusz, Mehrtash Harandi,
    ArXiv Preprint 2019, (Full Paper /ArXiv/)
  3. Rethinking Class Relations in Few-shot Learning,
    Hongguang Zhang, Songlei Jian, Piotr Koniusz,
    ArXiv Preprint 2019, (Full Paper /ArXiv/)
  4. Multi-stage Second-order Similarity Network for Few-shot Image and Action Recognition,
    Hongguang Zhang, Hongdong Li, Piotr Koniusz,
    ArXiv Preprint 2019, (Full Paper /ArXiv/)
  5. Flight Delay Prediction using Airport Situational Awareness Map,
    Wei Shao, Sichen Zhao, Siyu Tan, Arian Prabowo, Flora Salim, Piotr Koniusz, Jeffrey Chan, Xinhong Hei,
    ArXiv Preprint 2019, (Full Paper /ArXiv/)
  6. Few-Shot Learning via Saliency-guided Hallucination of Samples,
    Hongguang Zhang, Jing Zhang, Piotr Koniusz,
    ArXiv Preprint 2018, (Full Paper /ArXiv/)
  7. Deep Image Deblurring with Multi-Patch Hierarchical Network at 30fps,
    Hongguang Zhang, Yuchao Dai, Hongdong Li, Piotr Koniusz,
    ArXiv Preprint 2018, (Full Paper /ArXiv/)
  8. Power Normalizing Second-order Similarity Network for Few-shot Learning,
    Hongguang Zhang, Piotr Koniusz,
    ArXiv Preprint 2018, (Full Paper /ArXiv/)
  9. TrajCNN: Deep Map Inference from GPS Trajectories,
    Arian Prabowo, Piotr Koniusz, Wei Shao, Flora Salim,
    ArXiv Preprint 2018, (Full Paper /ArXiv/)
  10. Projective Subspace Networks For Few-Shot Learning,
    Christian Simon, Piotr Koniusz, Mehrtash Harandi,
    ArXiv Preprint 2018, (Full Paper /ArXiv/)
  11. Model Selection for Generalized Zero-shot Learning,
    Hongguang Zhang, Piotr Koniusz,
    ArXiv Preprint 2018, (Full Paper /ArXiv/)
  12. Recovering Faces from Portraits with Auxiliary Facial Attributes,
    Fatemeh Shiri, Xin Yu, Fatih Porikli, Richard Hartley, Piotr Koniusz
    ArXiv Preprint 2018, (Full Paper /ArXiv/)
  13. CNN-based Action Recognition and Supervised Domain Adaptation on 3D Body Skeletons via Kernel Feature Maps,
    Yusuf Tas, Piotr Koniusz,
    ArXiv Preprint 2018, (Full Paper /ArXiv/)
  14. Deeper Look at Power Normalizations, Piotr Koniusz, Hongguang Zhang, Fatih Porikli,
    ArXiv Preprint 2017, (Full Paper /ArXiv/)
  15. Zero-Shot Kernel Learning, Hongguang Zhang, Piotr Koniusz,
    ArXiv Preprint 2017, (Full Paper /ArXiv/)
  16. Museum Exhibit Identification Challenge for Domain Adaptation and Beyond,
    Piotr Koniusz, Yusuf Tas, Hongguang Zhang, Mehrtash Harandi, Fatih Porikli, Rui Zhang,
    ArXiv Preprint 2017, (Full Paper /ArXiv/)
  17. End-to-end Statistical Matching for Domain Adaptation, P. Koniusz et al., 2017
  18. Domain Adaptation by Mixture of Alignments of Second- or Higher-Order Scatter Tensors, P. Koniusz, Y. Tas, F. Porikli
    ArXiv Preprint 2016, (Full Paper /ArXiv/)
  19. Tensor Representations via Kernel Linearization for Action Recognition from 3D Skeletons, P. Koniusz, A. Cherian, F. Porikli
    ArXiv Preprint 2016, (Full Paper /ArXiv/)
  20. Dictionary Learning and Sparse Coding for Third-order Super-symmetric Tensors, P. Koniusz, A. Cherian,
    ArXiv Preprint 2015, (Full Paper /ArXiv/)

Studying for a PhD at the Australian National University

I always look for motivated and hard-working students who are interested in doing research with me. However, you must have good mathematical and computational skills, preferably with prior experience in machine learning and/or computer vision. Strong programming skills in Python, Matlab and/or TensorFlow are a must, hands-on deep learning know-how is important too. Integrity, honesty, ability to work under pressure and explore while taking strict directions is a must.

Moreover, the scholarships at the ANU are decided by a panel on the competitive basis. Normally, 1-3 candidates are selected among as many as 15-20 applicants aspiring to join our group. The candidates are expected to come from high-quality universities highly ranked in the QS Word University Rankings, e.g. see QS Global World Ranking for the ANU. The candidates are expected to have graduated with distinction (top 1% of university cohort, often university medalists, HD grades, minimum 5% of cohort and 1st class), have a very high GPA, have outstanding references from ideally professor-level referees and often have already a paper or two in top computer vision and machine learning conferences/journals such as CVPR, ICCV, ECCV, NIPS, ICML, ICLR, TPAMI, IJCV, TIP, TNNLS, BMVC, WACV, etc., and/or even hold patents.

There are two rounds of CECS/ANU scholarships, e.g. one around the 15th of May, and second around the 15th of September (including the domestic round). See details at ANU PhD Scholarships though be sure to e-mail HDR/research office straight away for key dates as they tend to move around (and I am bad at tracking these changes). An option may be also a scholarship from the Chinese Scholarship Council (I believe candidates must be already in touch with CSC by December and shortly after with the ANU). Finally, there is a number of Data61 scholarships (and/or top-ups for existing PhD students) as detailed by Data61 Scholarship Program which follow a similar rigorous process. English-wise, an IELTS (or equivalent) with an overall score of 6.5 with a minimum of 6.0 in each component is required (taken no later than one year ago but the uni. can always proceed firstly with a conditional offer).

I do not normally accept BSc/MSc (etc.) students for honours projects etc. simply due to lack of time. However, I may make an exception if an ANU student is on their way towards obtaining a high distinction (top 1% of university cohort, running for an university medalists, having HD grades and very high GPA, or at least being in minimum 5% of cohort and going for a clear cut 1st class degree), has interest in my research work and is seriously planning on pursuing a PhD under my guidance at the ANU (and has an understanding of general requirements and dedication required during PhD studies already explained above).

Useful Links

  1. CLARET II, Legacy page for the project CLARET II
  2. PK's home page at CECS ANU
  3. PK's home page at INRIA THOTH
  4. Fatih Porikli, The Group Leader at Computer Vision Research Group Data61/CSIRO (former NICTA)
  5. Image Indexing and Retrieval, Centre For Vision, Speech, and Signal Processing, University of Surrey
  6. Affine Covariant Features, Robotics Research Group, University of Oxford
  7. Feature Detectors and Descriptors, The State Of The Art and Beyond
  8. Image and Video Retrieval, The State Of The Art in Retrieval, INRIA TexMex
  9. Image and Video Classification, More References in Computer Vision, INRIA THOTH
Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License