הפקולטה להנדסה ע"ש איבי ואלדר פליישמן

EE Seminar: Handling drift induced by background clutter in on-line visual tracking

Speaker: Shaul Oron

Ph.D. student under the supervision of Prof. Shai Avidan

Sunday, March 19^th, 2017 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

Handling drift induced by background clutter in on-line visual tracking

Abstract

Visual tracking is a challenging task that attracts considerable interest in the field of computer vision. Despite recent advances many open questions still remain. One of the most basic challenges is handling tracking drift and, in particular, drift induced by background clutter. We address this challenge from several different angles.

We first consider a constrained setup of real-time tracking of objects from a known class (i.e., vehicles). In such a setup, one can train, offline, a non-real-time object class detector. Such a detector can help reduce the effects of background clutter. We discuss how this can be done, and focus on integrating the detector in a real-time tracking system. Our research leads to several novel fusion strategies that outperform individual system components as well as naive fusion techniques. We demonstrate the advantages of our method in a wide range of experiments.

We then turn our attention to model-free visual tracking in which the object class is unknown. In order to combat clutter induced drift in this setup we propose a novel, unified tracking framework. In this framework a parametric rigid transformation is assumed, and the likelihood of pixels belonging to either foreground or background is explicitly modeled. The resulting tracking algorithm provides an extension to a well known tracking method. Competitive performance is demonstrated when evaluated on a standard tracking benchmark.

Finally, we propose a novel similarity measure between point sets that can be used for template matching. This similarity measure alleviates the need for explicit background modeling and the constraints of a rigid geometric transformation. We investigate key properties of this similarity measure and show that it can be

applied successfully to visual tracking. The resulting tracking algorithm is able to cope with background clutter and non-rigid deformations in a fully unsupervised manner. Tracking performance is demonstrated on several benchmarks and compared to many leading tracking methods, with promising results.

19 במרץ 2017, 15:00

חדר 011, בניין כיתות-חשמל

סמינר מחלקתי ביה"ס להנדסה מכאנית Lior Medina

05 ביוני 2017, 14:00

וולפסון 206

סמינר מחלקתי ביה"ס להנדסה מכאנית Lior Medina

למפגש בוגרות וסטודנטיות תעשייה וניהול על כוס יין

15 במרץ 2017, 18:00

אולם 011 בניין כיתות בפקולטה להנדסה

למפגש בוגרות וסטודנטיות תעשייה וניהול על כוס יין

ארגון הבוגרים וועד הנדסה שמחים להזמינכן

למפגש בוגרות וסטודנטיות על כוס יין

המפגש מיועד לסטודנטיות ובוגרות ממגמת תעשייה וניהול

סטודנטיות בואו לשמוע על ניהול קריירה ולשאול מה שבאמת מעניין אתכן

בוגרות מוזמנות לשתף מניסיונכן בנושאי כיווני קריירה והתפתחות אישית

האירוע יתקיים ביום רביעי 15.03.17 בין השעות 18:00-20:30

אולם 011 בניין כיתות

18:00-18:30 התכנסות (על כוס יין וכיבוד קל)

18:30-19:00 פתיחה + הרצאה בנושא בנייה וניהול קריירה

19:00-20:00 פאנל של בוגרות מכיווני קריירה שונים

20:00-20:30 שאלות, תשובות, פיזור

EE SEminar: Optimal thresholding of singular values and eigenvalues

(The talk will be given in English)

Speaker: Prof. Matan Gavish
School of Computer Science and Engineering, Hebrew University

Wednesday, March 15^th, 2017
15:00 - 16:00

Room 011, Kitot Bldg., Faculty of Engineering

Optimal thresholding of singular values and eigenvalues

Abstract

It is common practice in multivariate and matrix-valued data analysis to reduce dimensionality by performing a Singular Value Decomposition or Principal Component Analysis, and keeping only r singular values or principal components, the rest being presumably associated with noise. However, the literature does not propose a disciplined criterion to determine r; most practitioners still look for the ``elbow in the Scree Plot'', a 50-years-old heuristic performed by eye. I'll review a line of work which develops a systematic approach to eigenvalue and singular value thresholding. This approach assumes that the signal is low-rank and that the noise is rotationally invariant. Recent results derive optimal thresholds in the presence of quite general noise distributions.

Joint work with David Donoho, Iain Johnstone and Edgar Dobriban (Stanford).

Bio
Matan is an assistant professor at the Hebrew University of Jerusalem School of Computer Science and Engineering. He received the dual B.Sc. degree in Mathematics and Physics from Tel Aviv University (TAU) in 2006, the M.Sc. degree in Mathematics from the Hebrew University of Jerusalem in 2008 (supervised by Hillel Furstenberg) and the Ph.D. degree in Statistics from Stanford University in 2014 (supervised by David Donoho and Ronald Coifman). His research interests include applied harmonic analysis, high-dimensional statistics, computing and machine learning. He was in the Adi Lautman Interdisciplinary Program for outstanding students at TAU from 2002 to 2006 and held a William R. and Sara Hart Kimball Stanford Graduate Fellowship from 2009 to 2012.

‏

15 במרץ 2017, 15:00

חדר 011, בניין כיתות-חשמל

סמינר מחלקתי

14 במרץ 2017, 14:00

Soft logic and numbers

Moshe Klein - Ph.D student

Abstract:

Marcelo Dascal (2008) wrote about the mathematician and philosopher Gottfried Wilhelm Leibniz, that as a young researcher he aspired to develop a universal language with a single symbol. Spencer-Brown (1969) fulfilled this vision in his book Laws of Form. As suggested by Dascal, Leibniz converted his first vision into a new one: to discover and develop a mathematical language that will demonstrate a softer logic that will overcome the limitations of the dichotomy of truth and falsehood. Leibniz had an ambitious plan to construct a universal language, which will prevent misunderstandings between people as well as serve as a scientific language that reflects thought. According to Dascal, language is a tool for thinking and influences thinking. Precise formal language, precise expression and thinking, are necessary to reduce the number of errors and increase certainty, thus allowing for the resolution of disputes. However, Dascal argues that Leibniz knew that no rational thinking and no “soft rationality” could be described by a formal computational model of rationality and computational language. In fact, Leibniz wrote in many occasions that the logic of two states is insufficient to grasp the full meaning of reason.

In this work ,we propose to see the Necker cube phenomenon as a basis for the development of a mathematical language in accordance with Leibniz’s vision of soft logic. By the development of a new coordinate system, we make a distinction between -0 and +0. This distinction enables us to present points on the Mobius strip in a natural way.

We established a connection between "Recursive Distinctioning" and soft logic, and use it as a basis for a new computational model. This model has a potential to change the current

computational paradigm.

This work was performed under the supervision of Prof. Oded Maimon.

ההרצאה תתקיים ביום שלישי 14.3.17, בשעה 14:00 בחדר 438, בנין וולפסון הנדסה, הפקולטה להנדסה, אוניברסיטת תל-אביב

EE Seminar: On massively multithreaded DSP architecture

08 במרץ 2017, 15:30

חדר 011, בניין כיתות-חשמל

Speaker: Ben Perach

M.Sc. student under the supervision of Prof. Shlomo Weiss

Wednesday, March 8^th 2017 at 15:30

Room 011, Kitot Bldg., Faculty of Engineering

On massively multithreaded DSP architecture

Abstract

Processor designers attempt to gain performance by introducing new, more efficient architectures. Digital signal processors (DSP) are designed to process very efficiently digital signal applications. This focus allows DSP designers to select tradeoffs that may not be appropriate for general purpose processors. For example, typical power consumption of DSPs is measured in milliwatts while the power used by an Intel state-of-the-art processor is in the range of tens of watts. In recent years we have seen a growing use of GPUs by non-graphical scientific applications because of their computation capabilities.

In this work we introduce a new DSP architecture, designed on the basis of the same parallel processing principles used in GPU architectures.

We implement this new architecture on FPGA, show evaluation results for widely used DSP algorithms and evaluate die area and power consumption.

We achieve similar computation speeds as in GPUs, with reduced die area and power consumption due to the use of hardware adaptations for DSP applications.

EE Seminar: Resource Allocation For LDPC-coded multi-carrier Downlink Channels

08 במרץ 2017, 15:00

חדר 011, בניין כיתות-חשמל

Speaker: Max Bluvshtein,

M.Sc. student under the supervision of Dr. Ofer Amrani

Wednesday, March 8^th, 2017 at 15:00

Room 011, Kitot Bldg., Faculty of Engineering

Resource Allocation For LDPC-coded multi-carrier Downlink Channels

Abstract

Various techniques have been proposed to address the problem of resource allocation for multi-carrier communications, i.e. Orthogonal frequency-division multiplexing (OFDM) systems. They can be categorized by the type of criteria they aim at optimizing. These criteria, for the most part, arise either from information-theoretic measures such as channel capacity, or from measures that describe the error performance of a channel, which does not necessarily model the coding being employed.

In this work, we propose an optimization technique that is tailored for low density parity-check (LDPC) coded OFDM systems, by employing the so-called general stability condition introduced by Richardson et. al. ("Design of capacity-approaching irregular low-density parity-check codes"). The latter formulates a necessary condition for the Belief-Propagation (BP) decoder to perfectly decode a received vector with no errors.

We re-formulate the general condition so as to model practical multi-carrier/multi-user systems such as OFDMA and OFDRMA (OFDM with random multiple-access). By doing so, a general framework for resource allocation is laid down and employed herein to optimize the transmitted power based on the characteristics of LDPC-coded systems. Fortunately, a convex optimization problem is obtained, whose closed-form solution is elaborated upon in this work. The proposed optimization technique is utilized for performing power allocation in OFDMA and OFDMRA systems in a way that minimizes the transmitted power while guaranteeing reliable decoding.

To validate this technique, it is compared to information-theoretic methods that aim at optimizing the mutual-information between the transmitter and receiver. It is shown that both provide almost identical performance for all the scenarios addressed in this work.

EE Seminar: Coresets for Kinematic Data: From Theorems to Autonomous Toy-Drones

(The talk will be given in English)

Speaker: Dr. Dan Feldman
Robotics & Big Data Lab, Computer Science Department, University of Haifa

Monday, March 13^th, 2017
15:00 - 16:00

Room 011, Kitot Bldg., Faculty of Engineering

Coresets for Kinematic Data: From Theorems to Autonomous Toy-Drones

Abstract

A coreset (or core-set) of a dataset is its semantic compression with respect to a set of queries, such that querying the (small) coreset provably yields an approximate answer to querying the original (full) dataset. However, we are not aware of real-time systems that compute coresets in a rate of dozens of frames per second.

I will suggest a framework to turn theorems to such systems using coresets. This is by maintaining such a coreset for kinematic (moving) set of n points, and run algorithms on the small coresets, instead of the n points, in realtime using weak devices.

This also enabled my group to implement a low-cost (< $100) mini-computer with a wireless system that tracks a toy (and harmless) quadcopter which guides guests to a desired room (in a hospital, mall, hotel, museum, etc.) with no help of additional human or remote controller. I will present the system as well as extensive experimental results.

A joint work with Soliman Nasser and Ibrahim Jubran

13 במרץ 2017, 15:00

חדר 011, בניין כיתות-חשמל

EE Seminar: Computational aspects of communication

(The talk will be given in English)

Speaker: Dr. Elad Haramaty
School of Engineering and Applied Sciences, Harvard University

Monday, March 20^th, 2017
15:00 - 16:00

Room 011, Kitot Bldg., Faculty of Engineering

Computational aspects of communication

Abstract

Communication has played a growing role in our lives over the past decade. It often becomes a bottleneck in the performance of our daily tasks. This motivates the pursuit for more efficient communication. However, efficiency is becoming more challenging from the computational aspect, due to several of its characteristics in modern communications.

One such characteristic is the interactivity of the protocols in today's noisy communication environments. One natural approach to overcome such a challenge is called Interactive Coding. Interactive coding is an efficient black-box mechanism to convert any interactive protocol that performs well under noiseless environment into one that is also resilient to errors, while preserving the efficiency of the protocol.

Another characteristic which challenges today’s communications is the dynamic and interactive nature of modern protocols. This may lead to desynchronization between the communicating parties. However, until now, almost all designed systems assume that both parties are perfectly synchronized and all context is shared perfectly by the communicating agents. Thus, any violation of those protocols leads to a breakdown in communication.

This talk will address both of the aforementioned challenges.The first part of the talk will be devoted to interactive coding and in the second part, we will focus on designing protocols under uncertainty of the shared context.

Bio

Elad Haramaty is a postdoctoral researcher at Harvard University in the School of Engineering and Applied Sciences with Madhu sudan.
Previously, he was a postdoc in Northeastern University with Emanuelle Viola. He earned his PhD from the Technion under the guidance of Amir Shpilka, and completed undergraduate work at the Technion in the mathematics department. His research interests cover a broad range of theoretical computer science, especially algebraic complexity and communication related questions.

20 במרץ 2017, 15:00

חדר 011, בניין כיתות-חשמל

EE Seminar: Secure Group Testing

(The talk will be given in English)

Speaker: Dr. Asaf Cohen
Department of Communication Systems Engineering,Ben Gurion University

Monday, May 15^th, 2017
15:00 - 16:00

Room 011, Kitot Bldg., Faculty of Engineering

Secure Group Testing

Abstract

The principal mission of Group Testing (GT) is to identify a small subset of defective items from a large population, by grouping items into as little as possible test pools. The test outcome of a pool is positive if it contains at least one defective item, and is negative otherwise. GT algorithms are utilized in numerous applications, and in most of them the privacy of the tested subjects, namely, whether they are defective or not, is critical. In this talk, we consider a scenario where there is an eavesdropper (Eve) which is able to observe a subset of the GT outcomes (pools). We propose a new non-adaptive Secure Group Testing (SGT) algorithm based on information theoretic principles, which keeps the eavesdropper ignorant regarding the items’ status. Specifically, when the fraction of tests observed by Eve is 0 ≤ δ < 1, via random coding, ML decoding and a matching converse, we prove that the number of tests required for both correct reconstruction at the legitimate user (with high probability) and negligible mutual information at Eve’s side is 1/(1−δ) times the number of tests required with no secrecy constraint. We conclude with a practical SGT decoding algorithm, and asses its performance as well.

Based on joint work with Alejandro Cohen, Omer Gurewitz and Sidharth Jaggi.

15 במאי 2017, 15:00

חדר 011, בניין כיתות-חשמל

עמודים