High performance computing in data and image processing A.A. Lukianitsa 1,2, A.G. Shishkin 1,2, F.S. Zaitsev 1,2 1 Fusion Advanced Research Group Ltd.,

Slides:



Advertisements
Similar presentations
Examples of ITER CODAC requirements for diagnostics
Advertisements

Road-Sign Detection and Recognition Based on Support Vector Machines Saturnino, Sergio et al. Yunjia Man ECG 782 Dr. Brendan.
Sensor-Based Abnormal Human-Activity Detection Authors: Jie Yin, Qiang Yang, and Jeffrey Junfeng Pan Presenter: Raghu Rangan.
Visual Event Detection & Recognition Filiz Bunyak Ersoy, Ph.D. student Smart Engineering Systems Lab.
Object Inter-Camera Tracking with non- overlapping views: A new dynamic approach Trevor Montcalm Bubaker Boufama.
Benny Neeman Leon Ribinik 27/01/2009. Our Goal – People Tracking We would like to be able to track and distinguish the different people in a movie.
SKELETON BASED PERFORMANCE PREDICTION ON SHARED NETWORKS Sukhdeep Sodhi Microsoft Corp Jaspal Subhlok University of Houston.
Measurement of magnetic island width by using multi-channel ECE radiometer on HT-7 tokamak Han Xiang( 韩翔 ), Ling Bili( 凌必利 ), Gao Xiang( 高翔 ), Liu Yong(
ICIP 2000, Vancouver, Canada IVML, ECE, NTUA Face Detection: Is it only for Face Recognition?  A few years earlier  Face Detection Face Recognition 
Motion Tracking. Image Processing and Computer Vision: 82 Introduction Finding how objects have moved in an image sequence Movement in space Movement.
Motion Detection And Analysis Michael Knowles Tuesday 13 th January 2004.
Rodent Behavior Analysis Tom Henderson Vision Based Behavior Analysis Universitaet Karlsruhe (TH) 12 November /9.
Object Detection and Tracking Mike Knowles 11 th January 2005
2007Theo Schouten1 Introduction. 2007Theo Schouten2 Human Eye Cones, Rods Reaction time: 0.1 sec (enough for transferring 100 nerve.
Imaging Diagnostics at the H-1 National Plasma Fusion Research Facility Left: The coherence tomography system Above: Plasma emission reconstructions compared.
55:148 Digital Image Processing Chapter 11 3D Vision, Geometry Topics: Basics of projective geometry Points and hyperplanes in projective space Homography.
Computing Platform Benchmark By Boonyarit Changaival King Mongkut’s University of Technology Thonburi (KMUTT)
Digital Sound and Video Chapter 10, Exploring the Digital Domain.
Fast imaging of global eigenmodes in the H-1 heliac ABSTRACT We report a study of coherent plasma instabilities in the H-1 plasma using a synchronous gated.
An Automatic System for CME Detection and Source Region Identification Jie Zhang Art Poland, Harry Wechsler Kirk Borne George Mason.
Nils P. Basse Plasma Science and Fusion Center Massachusetts Institute of Technology Cambridge, MA USA ABB seminar November 7th, 2005 Measurements.
Massively Parallel Magnetohydrodynamics on the Cray XT3 Joshua Breslau and Jin Chen Princeton Plasma Physics Laboratory Cray XT3 Technical Workshop Nashville,
Multimedia Databases (MMDB)
6 th Japan-Korea Workshop on Theory and Simulation of Magnetic Fusion Plasmas Hyunsun Han, G. Park, Sumin Yi, and J.Y. Kim 3D MHD SIMULATIONS.
1. Introduction Motion Segmentation The Affine Motion Model Contour Extraction & Shape Estimation Recursive Shape Estimation & Motion Estimation Occlusion.
 In electrical engineering and computer science image processing is any form of signal processing for which the input is an image, such as a photograph.
Database structure for the European Integrated Tokamak Modelling Task Force F. Imbeaux On behalf of the Data Coordination Project.
Object Based Processing for Privacy Protected Surveillance Karl Martin Kostas N. Plataniotis University of Toronto Dept. of Electrical and Computer Engineering.
PROTO-SPHERA Diagnostics PROTO-SPHERA WORKSHOP Frascati March 18-19, 2002.
Introduction EE 520: Image Analysis & Computer Vision.
An Information Fusion Approach for Multiview Feature Tracking Esra Ataer-Cansizoglu and Margrit Betke ) Image and.
Academic and pedagogical options in CIM laboratory CIM in universities.
Forward-Scan Sonar Tomographic Reconstruction PHD Filter Multiple Target Tracking Bayesian Multiple Target Tracking in Forward Scan Sonar.
Kevin Cherry Robert Firth Manohar Karki. Accurate detection of moving objects within scenes with dynamic background, in scenarios where the camera is.
December 9, 2014Computer Vision Lecture 23: Motion Analysis 1 Now we will talk about… Motion Analysis.
J M Lopez 1 (of 17) 7 th Workshop on Fusion.., Frascati March, Simulator of the JET real-time disruption predictor J.M. Lopez*, S. Dormido-Canto,
Vehicle Segmentation and Tracking From a Low-Angle Off-Axis Camera Neeraj K. Kanhere Committee members Dr. Stanley Birchfield Dr. Robert Schalkoff Dr.
Tracking and event recognition – the Etiseo experience Son Tran, Nagia Ghanem, David Harwood and Larry Davis UMIACS, University of Maryland.
Lei Li Computer Science Department Carnegie Mellon University Pre Proposal Time Series Learning completed work 11/27/2015.
By Dirk Hekhuis Advisors Dr. Greg Wolffe Dr. Christian Trefftz.
Segmentation of Vehicles in Traffic Video Tun-Yu Chiang Wilson Lau.
Human Activity Recognition at Mid and Near Range Ram Nevatia University of Southern California Based on work of several collaborators: F. Lv, P. Natarajan,
Virtual Application Profiler (VAPP) Problem – Increasing hardware complexity – Programmers need to understand interactions between architecture and their.
By Naveen kumar Badam. Contents INTRODUCTION ARCHITECTURE OF THE PROPOSED MODEL MODULES INVOLVED IN THE MODEL FUTURE WORKS CONCLUSION.
RFX workshop / /Valentin Igochine Page 1 Control of MHD instabilities. Similarities and differences between tokamak and RFP V. Igochine, T. Bolzonella,
QAS Design of the DEMO Reactor
Steady State Discharge Modeling for KSTAR C. Kessel Princeton Plasma Physics Laboratory US-Korea Workshop - KSTAR Collaborations, 5/19-20/2004.
Multimedia Systems and Communication Research Multimedia Systems and Communication Research Department of Electrical and Computer Engineering Multimedia.
IEEE International Conference on Multimedia and Expo.
Situation We now accept that grammar is not restricted to writing but is present in speech. Problem This can lead to assumptions that there is one kind.
Suspicious Behavior in Outdoor Video Analysis - Challenges & Complexities Air Force Institute of Technology/ROME Air Force Research Lab Unclassified IED.
WHAT IS DATA MINING?  The process of automatically extracting useful information from large amounts of data.  Uses traditional data analysis techniques.
Company Confidential Konnet ViAn Konnet ViAn P. Ltd. 102, Shiv Shakti Compex Baner road, Pune , India /5 Computing.
Long Pulse High Performance Plasma Scenario Development for NSTX C. Kessel and S. Kaye - providing TRANSP runs of specific discharges S.
Motion tracking TEAM D, Project 11: Laura Gui - Timisoara Calin Garboni - Timisoara Peter Horvath - Szeged Peter Kovacs - Debrecen.
Visual Information Processing. Human Perception V.S. Machine Perception  Human perception: pictorial information improvement for human interpretation.
Generalized and Hybrid Fast-ICA Implementation using GPU
Proposing Data Mining for Plasma Diagnosis
Parallel Plasma Equilibrium Reconstruction Using GPU
S. N. Simrock, A. Aallekar, L. Abadie, L. Bertalot, M. Cheon, C
Utilizing AI & GPUs to Build Cloud-based Real-Time Video Event Detection Solutions Zvika Ashani CTO.
An Overview of Some Concepts in the World of Computing
Fast and Robust Object Tracking with Adaptive Detection
Vehicle Segmentation and Tracking in the Presence of Occlusions
Chapter 1 Introduction.
Deep Neural Networks for Onboard Intelligence
Computer Evolution and Performance
Computer Graphics Lecture 15.
Development High-Speed Visible Diagnostics for Real-Time Plasma Boundary Reconstruction on EAST By: Biao Shen 8/27/2019.
Presentation transcript:

High performance computing in data and image processing A.A. Lukianitsa 1,2, A.G. Shishkin 1,2, F.S. Zaitsev 1,2 1 Fusion Advanced Research Group Ltd., Slovakia. 2 Moscow State University, Russia th Workshop on Fusion Data Processing Validation and Analysis Invitation. Frascati 2012

Contents 1.Motivation. 2.Previous experience in fusion data mining. 3.High performance computing (HPC). 4.Examples of HPC applications. 5.“Virtual Tokamak” technology. 6.Summary. 2

1. Motivation Huge amounts of collected and stored data from various sources: –Large physical experiments. –Video surveillance. –DNA sequencing. New challenges in various scientific areas: –Real-time thermonuclear pulses control. –Genomic sequence analysis and gene-finding. –Real-time video analysis and indexing. New computational algorithms. Internet transmission speeds and usage keep growing. Cloud computing. Urgent need for high performance computing. 3

L,H-mode classification with SVM for the Global confinement database 10 4 pulses [Plas.Phys.Contr. Fus. 50(2008)065013]. 2. Previous experience in fusion data mining Linear separating hyperplane was found Forecast accuracy %. Deep H-mode is predicted for ITER. Power loss threshold is more optimistic for ITER ~0 (was ~40) MW. Allows checking L or H-mode for a particular set of 8 parameters, including prediction and evaluation how deep is the set in a mode. 4

Code VIP (Video Image Processing). Plasma boundary reconstruction in (R,Z,η). Dynamic programming algorithm. [32 EPS Conf. on Plasma Phys. 2005, P ] 5

Actual plasma boundary is quite different from the image and can be different from the EFIT reconstruction. Code VIP was installed in MAST for regular use. 6

3D processing: plasma boundary reconstruction. 7

Detection and tracking of snow flakes in JET plasma video. [Frontiers in Diagnostic Technologies.1st Int. Conf. Frascati 2009] Complicated algorithm: camera image stabilizing, image intensity equa- lizing, background modeling, foreground particles separation, snowflakes tracking Frame=78Frame=79Frame=80 Snowflakes tracking over the time Trajectories 8

Demo of snow flakes detecting and tracking. 9 The moment of appearance, trajectory, size, shape and color can give understanding what is happening.

Code FIRe (Fluorescence Intensity Reconstruction) - reconstructing distribution of the source of light using high resolution photo or video images. [1st Korean-Russian Workshop on Data Mining. 2007, p ] Tikhonov regularization algorithm. Treating several cameras. Removing the lens distortions. Options for given number of mirror or diffusive reflections from surfaces. Ray tracing. Description of each particular surface using triangular grid. Reconstruction for MAST plasmas One camera in the equatorial plane 10

Processing of magnetic diagnostics data using Hidden Markov Models (HMMs). Integrated data analyses. [8th Int. FLINS Conference "Computational Intelligence in Decision and Control“. Madrid 2008, p ] Well suited, since magnetic frequencies are similar to the speech ones. HMMs allow compression (from giga to kilo bytes) of time dependent data of different length, without substantial loss of information, to a standardized form, which can be used for different purposes, e.g. pattern recognition. Possible applications (almost all plasma processes change magnetic field): Navigation in the magnetic databases: finding similar or distant patterns. Instabilities recognition and prediction. Disruption prediction. Recognition of L, H-transition during a hybrid scenario pulse. Regular oscillations recognition. Determination of symmetry or rotation characteristics in poloidal or toroidal angles, harmonics numbers, unstable modes. Integrated analyses of data from different diagnostics (~100 in JET). Correlating with experimental (X-ray, optical, neutral, neutron, currents in coils, etc.) and/or numerical (safety factor, current density, type of instability, plasma shape, temperature and density profiles, etc). 11

Transformation of magnetic oscillations to the sound frequencies. 12

3. HP computing CPU vs. GPU: CPU –Fast caches. –Branching adaptability. –High performance. GPU –Multiple ALUs (Arithmetic Logic Units). –Fast onboard memory. –High throughput on parallel tasks/ Executes program on each fragment/vertex. –Easily scalable. CPUs are good for task parallelism. GPUs are good for data parallelism. 13

FARG experience with CUDA for commercial applications: Acceleration 14 log 10 S detecting & tracking of moving or left objects, monitoring car parking spaces, face detection & recognition, smoke & fire detection in videos, noise auto suppression, genetic information transfer, art authentication, emotion recognition by speech, biomedical signal analyses, etc.

4. Examples of HPC applications Demo of hyperplane construction with SVM. Accelertion ~

Key Features –Move to real-time Optical Flow on high video resolution and high frame rate. –Edge-preserving post processing algorithm to enhance quality. –More accurate occlusion estimation. –Scalable, modular, versatile architecture. –Low system’s resources consumption. Real-time НPC optical flow processing

GPU/CPU speed up, times Frame resolution Real-time НPC optical flow processing

Real-time НPC optical flow processing: the original video, background, blobs (compact moving objects), foreground (isolated blobs), blobs boundaries. 18

5. Virtual Tokamak Modeling plasma pulses and diagnostics within the universal toolbox “Virtual Tokamak” framework. Singling out the most advanced and adequate approaches and algorithms in direct and inverse problems in fusion. Their modification and standardization. Allowing remote usage of the software by a usual Web-browser. Distributed computations, cloud and GRID-like technologies. Integrated data analyses. Diagnostics benchmarking and co- ordination. Design support. Pilot URL: leader.ic.msu.su/~fusion 19

Demo of the “Vertual Tokamak” GUI. 20

6. Summary Accurate and robust data and image processing codes are described with HPC advancement of the algorithms HPC algorithms can allow real time plasma data analyses, reconstruction, modeling and 3D processing as in commercial applications. Possible application in advanced real-time feed-back plasma shape and position control and instabilities suppression. Advancement of the plasma and chamber state alarm systems. A reasonable direction of diagnostics advancements is application of FARG’s commercial HPC developments as modules in the existing software, e.g. used at JET. CUDA HPC has high potential in numerical codes: e-nets, Monte- Carlo. 21