Workshops : ICSPIS 2021 - 7th Conference on Signal Processing and Intelligent Systems

Workshops

Participants in the workshops will be received the "Certificate of Attendance at the Workshop"

To register for the workshops, please visit this page

Num.	Workshop title	Presenter	Date and Time
1	The new generations of the Mobile Networks: Challenges and opportunities	Dr. Abolfazl Diyanat	December 31, 2021, 8 A.M. – 12 P.M.
2	Signal and Image Decomposition and its applications: From Fourier transform to sparse and Deep models	Dr. Aboozar Ghaffari	December 31, 2021, 8 A.M. – 12 P.M.
3	Deep Learning for Computer Vision with Keras	Dr. Mohammad Reza Mohammadi	December 31, 2021, 13 P.M – 17 P.M.
4	Theory, Recording and Processing of Physiological Signals (EEG, EMG, ECG and EOG) using openbci modules	Dr. Ali Farzamnia	December 31, 2021, 13 P.M – 17 P.M.
Non-Payed Workshops
5	Edge Processing, Challenges, and Opportunities	Mohammad Hossein Ghaeminia	December 30, 2021, 9 A.M. – 11 P.M.

*The registration fee for each of the following workshops is normally set at 150 thousand Tomans and for students at 100 thousand Tomans, which in case of group registration (more than 4 people), a 25% discount will be considered for each person.

**********************************************************

************************************

*******************

(Deep Learning for Computer Vision with Keras)

Presenter:

Dr. Mohammad Reza Mohammadi— Assistant professor, Iran University of Science and Technology

Faculty Members - School of Computer Engineering - Iran University of Science & Technology (IUST)

Abstract

Deep learning is the state of the art in most computer vision tasks. In this workshop, we first present the basics of deep learning for computer vision. Then, we introduce the Keras framework and write a simple deep neural network code for image classification in the Google Colab environment. We then discuss some modern deep convolutional neural networks and implement one of them for image classification. It also discusses how to convert these networks into fully convolutional networks for object detection and semantic segmentation. Finally, we introduce AutoDL to reduce the need for highly-educated data scientists to build, train and maintain deep learning algorithms.

************************************************************************************************

*****************************************************************************

**************************************

*************

(The new generations of the Mobile Networks: Challenges and opportunities)

Presenter:

Dr. Abolfazl Diyanat— Assistant professor, Iran University of Science and Technology

Abstract

Today, Cellular Mobile Networks are one of the most lucrative and engaging industries in most of the countries in the world. It can be discerned that design, implementation, or any advancement in the new generations of mobile networks is considered as a major criterion in the development of the countries. In this workshop, the importance of this industry, research perspective, and the main features in different generations of Mobile Networks will be discussed along with a look at the features of the next generation of mobile networks. The growth and advancement of mobile networks from 1G to the fifth generation, and now, the sixth generation, is due to the development of technologies like Machine Learning, MIMO, IoT, SDN, NFV, SDR, Cloud Computing, and so on. This is especially true in the fifth and sixth-generation of Mobile Networks.

************************************************************************************************

*****************************************************************************

**************************************

*************

(Signal and Image Decomposition and its applications: From Fourier transform to sparse and Deep models)

Presenter:

Dr. Aboozar Ghaffari— Assistant professor, Iran University of Science and Technology

Abstract

Signal and image representation has a critical role in signal and image processing, such as denoising, classification, compression, and source separation. In this workshop, we present a review of signal decomposition approaches based on different viewpoints. At first, the linear decomposition based on Fourier and wavelet as deterministic transforms is illustrated. Then, Statistical decomposition based on multi-channel data such as principal component analysis PCA, Singular Spectrum Analysis (SSA), and independent component analysis ICA and their relations with deterministic transforms are investigated. Recently, two models based on sparse presentation and Deep neural networks have been used in many applications. In this workshop, these models and their relationships are described. Finally, some applications of these approaches have been described in many applications, such as classification, denoising, and vital sign estimation based on face video.

************************************************************************************************

*****************************************************************************

**************************************

*************

(Theory, Recording and Processing of Physiological Signals (EEG, EMG, ECG and EOG) using openbci modules)

Presenter:

Dr. Ali Farzamnia— Senior Lecturer, Universiti Malaysia Sabah

Abstract

Due to the large number of studies on the structure and function of the human brain, familiarity with the methods and modalities of recording information of this very important organ of the body is of great importance for researchers in relevant fields. One of the most important and widely used tools for recording brain activity is the electroencephalogram signal. This signal has been welcomed by researchers much more than other signal recording modalities due to its non-invasiveness, availability and portability of its recording devices as well as reasonable cost. For this reason, standard databases are available from this signal. Therefore, familiarity with the principles and methods of recording and processing this signal can be an important gateway to a large part of research in the field of brain function, including the study of various diseases of the central nervous system such as sleep disorders, movement disorders, Open up hearing and vision as well as areas of technology and rehabilitation such as emotion recognition, driver fatigue detection and other applications of this signal.

In this workshop, participants are first introduced to the basic concepts, nature, production and recording of electroencephalogram signals and various applications of this signal in medicine and engineering, and the principles of theory related to quantitative measurement and processing methods of this signal. Also, in this section, the principles of recording all physiological signals based on the Open BCI module and the ganglion and cyton boards will be presented. In the second part of the workshop, participants are introduced to how to record physiological signals based on the Open BCI module in various scenarios (including driver fatigue, automatic emotion detection, etc.) and practically recorded EEG, EMG, ECG and EOG data and learning the relevant details. In the third part of the workshop, participants are introduced to the EEG Lab toolbox, which is an important toolbox in the field of signal processing, and theoretical and practical topics related to the signal recorded in the environment of this toolbox.

Topics presented in the workshop:

Introduce the EEGLAB Toolkit and how to attach it to MATLAB
Importing EEG data in the toolbox and display it
Change channel montages and settings
Import and export different data types
Obtaining frequency spectrum of channels and brain mapping

Introducing pre-processing tools

Change the sampling rate of the signal
Data filtering
Change the reference channel
Independent component analysis (ICA)
Signal decomposition using ICA method
Two-dimensional and three-dimensional display of components on the skull
Check and remove the components
Noise reduction using various methods provided in the EEGLAB toolbox
Enter a collection of data from a study in a toolbox
***********************************************************************************************

*****************************************************************************

**************************************

***********

(Title: Video Processing on Edge)

Presenter:

Dr. Mohammad Hossein Ghaeminia— Hamrah Aval (MCI) company, Tehran,

Abstract

Computer vision is widely used to detect a variety of abnormalities and objects in video. The use of deep neural networks (DNNs) in computer vision can achieve high accuracy and performance, but applying DNNs require huge computing resources, storage space, and large video data. Thus, DNN-based video analysis has been further developed based on the cloud infrastructure with video data received from a set of still cameras. Thus, there are three major limitations in developing the edge services: First, the transmission of large amounts of video data from cameras leads to high bandwidth consumption and high latency. Second, when DNNs are deployed on equipment with limited resources (such as edge), it is difficult to achieve high detection accuracy. Third, still cameras can only collect a limited amount of video data corresponding to small region. It is therefore limited the use of applications on edge devices. In this workshop, we introduce deep neural networks and edge processing. Then suitable models and algorithms for light and executable processing on edge devices will be introduced. Recently, the edge-based methods reduce processing costs and time by up to 50%, while able to maintain the accuracy of common methods to an acceptable level.

Conference Countdown

----