Fast Convolution Algorithms for Deep learning and computer vision

February 2020 : Exact time and date TBD

CVML Live Web Lecture Series Concept

Artificial Intelligence and Information analysis (AIIA) Lab, AUTH is proud to launch the live CVML Web lecture series that will cover very important topics Computer vision/machine learning. Top scientists internationally will deliver these lectures, aiming at providing in-depth knowledge on various CVML topics. The 1-hour lectures will take place on Saturdays, to avoid conflicts with other intended registrant schedules/duties:
a) Saturdays 11:00 EET (17:00 Beijing time) and
b) Saturdays 20:00 EET (13:00 EST, 10:00 PST for NY/LA, respectively) for audience in the Americas.
Each lecture will be announced at least 1 week in advance in various relevant email lists and in this page.

Lectures will consist primarily of live lecture streaming and PPT slides. Attendees (registrants) need no special computer equipment for attending the lecture. They will receive the lecture PDF before each lecture and will have the ability to ask questions real-time. Audience should have basic University-level undergraduate knowledge of any science or engineering department (calculus, probabilities, programming, that are typical e.g., in any ECE, CS, EE undergraduate program). More advanced knowledge (signals and systems, optimization theory, machine learning) is very helpful but nor required.

Part A

Computer vision and machine learning fundamentals

  • Introduction to autonomous systems
  • Signals and systems.
  • 1D/2D convolutions
  • Introduction to computer vision
  • Image acquisition, camera geometry
  • Stereo and Multiview imaging
  • 3D building/monument modeling
  • Motion estimation
  • Introduction to neural networks, Perceptron, backpropagation
  • Deep neural networks, Convolutional NNs

Part B

Advanced computer vision and machine learning.

  • Deep learning for object/target detection
  • Object tracking
  • Localization and mapping
  • Object 3D/6D localization
  • Object shot type and pose recognition
  • 3D building/scene modelling
  • GPU programming
  • Fast convolution algorithms
  • Generative Adversarial Networks
  • Recurrent Neural Networks
  • Reinforcement learning
  • Autonomous systems/swarms communications

Part C

Application oriented topics


  • Drone cinematography
  • Drone mission planning  and control
  • Imaging for drone safety
  • Drone mission simulations
  • Privacy protection, ethics and regulatory issues

Autonomous cars:

  • Introduction to car vision
  • Road condition assessment
  • 3D road modeling

Marine applications:

  • Introduction to autonomous marine vehicles

The CVML web lecture series content, lecture timing and exact lecture topics may vary from the above ones depending on intended audience interest and lecturer availability.

The CVML web lecture series is expected to last till 30th August 2020. It will contain max 39 lectures (max 3×13 weeks), organized in batches of 1-3 lectures per week (lasting 1-3 consecutive hours per batch).

The first web lecture on ‘Fast Convolution Algorithms for deep learning and computer vision’ by Professor Ioannis Pitas, will take place on 14th December 2019, 11:00 EET only (17:00 Beijing time).

The same web lecture will be announced soon to take place on 11th January 2020, 20:00 EET (13:00 EST, 10:00 PST for NY/LA, respectively) for audience in the Americas.

February 2020 : Exact time and date TBD

Fast 1D/2D convolution algorithms for machine learning and computer vision

2D convolutions play an extremely important role in machine learning, as they form the first layers of Convolutional Neural Networks (CNNs). They are also very important for computer vision (template matching through correlation, correlation trackers) and in image processing (image filtering/denoising/restoration). 3D convolutions are very important for machine learning (video analysis through CNNs) and for video filtering/denoising/restoration. 1D convolutions are extensively used in digital signal processing (filtering/denoising)  and analysis (also through CNNs)

Therefore, 1D/2D/3D convolution algorithms are very important both for machine learning and for signal/image/video processing and analysis. As their computational complexity is of the order O(N^2), O(N^4) and O(N^6) respectively their fast execution is a must.

This lecture will overview linear and cyclic convolution. Then it will present their fast execution through FFTs, resulting in algorithms having computational complexity of the order O(Nlog2N), O(N^2log2N) for 1D and 2D convolutions respectively. Finally, optimal Winograd 1D and 2D convolution algorithms will be presented having theoretically minimal number of computations.



Nominal registration fee:-
Maximum number of registrants: 50, on a first-come-first-serve basis.
Registration cutoff date: –
After successful registration, you will receive infos on how to join the web lecture by email.

If you haven’t received the invitation mail to join the lecture, please contact


After the registration cutoff date/time, you will receive a link on how to attend the web lecture,

also containing:

a) lecture PDF

b) lecture questionnaire

c) lecture evaluation form.


Each lecture has a) a main lecturer, b) a tutor and c) administrator ( Lecture language will be English.

The tool ‘BigBlueButton’ will be used for web lecturing (PDF+live lecturer video+ questions using chat/audio). Your browser needs flash plugin installed.

You will receive a link on this by email.  The ‘Presentation meeting’ mode on your PC is suggested.


It is strongly recommended you join the lecture 10 min before its formal start, to have an informal chat with the lecturer and other attendants on general topics and have the feeling of a real live class.


During lecture, you can ask questions any time by chat to be replied by the lecture tutor, or afterwards by email, in case they are many. Live audio questions will be allowed in the middle and the end of the lecture (around min 25 and min 55). Live discussion on general topics will follow for another 10 min after the formal end of the lecture.


If you want to receive a certificate of lecture attendance with mark (in the range 0-10, 10 being the best mark, 5 being the pass mark),

you have to submit to the lecture administrator ( within 48 hours after formal lecture end:

a) your replies to lecture questionnaire: very short replies (1-2 text lines) to each of 10 questions.

You should have no problem replying them within 5-10 min, if you understood the lecture topics

b) your filled lecture evaluation form.

If somebody attends all yearly lectures (till 30th August 2019) she/he can get a certificate for the overall CVML web series attendance

with mark (in the range 0-10, 10 being the best mark, 5 being the pass mark).


Prof. Ioannis Pitas (IEEE fellow, IEEE Distinguished Lecturer, EURASIP fellow) received the Diploma and PhD degree in Electrical Engineering, both from the Aristotle University of Thessaloniki, Greece. Since 1994, he has been a Professor at the Department of Informatics of the same University. He served as a Visiting Professor at several Universities.

His current interests are in the areas of image/video processing, machine learning, computer vision, intelligent digital media, human centered interfaces, affective computing, 3D imaging and biomedical imaging. He has published over 1138 papers, contributed in 50 books in his areas of interest and edited or (co-)authored another 11 books. He has also been member of the program committee of many scientific conferences and workshops. In the past he served as Associate Editor or co-Editor of 9 international journals and General or Technical Chair of 4 international conferences. He participated in 70 R&D projects, primarily funded by the European Union and is/was principal investigator/researcher in 42 such projects. He has 30000+ citations to his work and h-index 81+ (Google Scholar).

Prof. Pitas leads the big European H2020 R&D project MULTIDRONE: He is chair of the Autonomous Systems initiative



[1] Y. LeCun, Y. Bengio, and G. Hinton, \Deep learning,» nature, vol. 521, no. 7553, p. 436, 2015.
[2] A. V. Oppenheim, Discrete-time signal processing. Pearson Education India, 1999.
[3] J. F. Henriques, R. Caseiro, P. Martins, and J. Batista, \High-speed tracking with kernelized correlation _lters,» IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 37, no. 3, pp. 583{596, 2015.
[4] O. Zachariadis, V. Mygdalis, I. Mademlis, N. Nikolaidis, and I. Pitas, 2d visual tracking for Sports uav cinematography applications,» in Proceedings of the IEEE Global Conference on Signal and Information Processing (GlobalSIP), 2017.19
[5] S. Winograd, Arithmetic complexity of computations, vol. 33. Siam, 1980.
[6] J. H. McClellan and C. M. Rader, \Number theory in digital signal processing,» 1979.
[7] H. J. Nussbaumer, Fast Fourier transform and convolution algorithms. Springer Science & Business Media, 1981.
[8] I. Pitas, Computational complexity study of multidimensional digital signal processing algorithms. Aristotle University of Thessaloniki, Greece, 1985.
[9] I. Pitas and M. Strintzis, \Multidimensional cyclic convolution algorithms with minimal multiplicative complexity,» IEEE transactions on acoustics, speech, and signal processing, vol. 35, no. 3, pp. 384-390, 1987.
[10] S. Chetlur, C. Woolley, P. Vandermersch, J. Cohen, J. Tran, B. Catanzaro, and E. Shelhamer, \cudnn: E_cient primitives for deep learning,» arXiv preprint arXiv:1410.0759, 2014.
[11] A. Lavin and S. Gray, \Fast algorithms for convolutional neural networks,» in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4013{4021, 2016.
[12] D. Budden, A. Matveev, S. Santurkar, S. R. Chaudhuri, and N. Shavit, Deep tensor convolution on multicores,» arXiv preprint arXiv:1611.06565, 2016.
[13] Z. Jia, A. Zlateski, F. Durand, and K. Li, \Optimizing n-dimensional, winograd-based convolution for manycore cpus,» in Proceedings of the 23rd ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, pp. 109{123, ACM, 2018.
[14] I. Pitas, Digital image processing algorithms and applications. John Wiley & Sons, 2000.
[15] D. B. Kirk andW. H.Wen-Mei, Programming massively parallel processors: a hands-on approach. Morgan kaufmann, 2016.
[16] V. Sze, Y.-H. Chen, T.-J. Yang, and J. S. Emer, \E_cient processing of deep neural networks: A tutorial and survey,» Proceedings of the IEEE, vol. 105, no. 12, pp. 2295{2329, 2017.
[17] V. Strassen, \Gaussian elimination is not optimal,» Numerische mathematik, vol. 13, no. 4, pp. 354-356, 1969.
[18] J. Cong and B. Xiao, \Minimizing computation in convolutional neural networks,» in international conference on arti_cial neural networks, pp. 281- 290, Springer, 2014.
[19] I. Pitas and M. Strintzis, \Algorithms and structures for convolutions over galois _elds,» in 9-me Colloque sur le traitement du signal et des images, FRA, 1983, GRETSI, Groupe d’Etudes du Traitement du Signal et des Images, 1983.
[20] R. Twogood, M. Ekstrom, and S. Mitra, \Optimal sectioning procedure for the implementation of 2-d digital _lters,» IEEE Transactions on Circuits and Systems, vol. 25, no. 5, pp. 260-269, 1978.
[21] R. Agarwal and J. Cooley, \New algorithms for digital convolution,» IEEE
Transactions on Acoustics, Speech, and Signal Processing, vol. 25, no. 5, pp. 392{410, 1977.
[22] N. I. Mademlis, V. Mygdalis and I.Pitas, \Challenges in autonomous uav cinematography: An overview,» IEEE International Conference on Multimedia and Expo (ICME), pp. 392-410, 1977.
[23] P. Ribenboim, Classical theory of algebraic numbers. Springer Science & Business Media, 2013.

Relevant links: