Speaker: Anna Darzi

M.Sc. student under the supervision of Prof. Shai Avidan

Wednesday, March 6^th, 2019 at 15:30

Room 011, Kitot Bldg., Faculty of Engineering

Co-occurrence Based Texture Synthesis

Abstract

We model local texture patterns using the co-occurrence statistics of pixel values. We then train a conditional generative adversarial network (cGAN) to synthesize new textures from the co-occurrence statistics and a random seed noise. Co-occurrences have long been used to measure similarity between textures. That is, two textures are considered similar if their corresponding co-occurrence matrices are similar. By the same token, we show that multiple textures generated from the same co-occurrence matrix are similar to each other. This gives rise to a new texture synthesis algorithm.

We use co-occurrence based texture synthesis in various settings. For example, we generate variations on the input texture by using the same co-occurrence statistics with different seed noise, or we merge two co-occurrence matrices to smoothly
interpolate between different textures.

In another case, we synthesize a dynamic texture sequence by interpolating between two co-occurrence matrices. Yet another option is to create a sequence that
summarizes the various local texture patterns in a given texture image. And because co-occurrence statistics have clear and intuitive meaning we develop a tool that lets users modify them directly and hence influence the local characteristics of the synthesized texture image.

Speaker: Elad Tzoreff

Ph.D. student under the supervision of Prof. Anthony J. Weiss

Monday, March 4^th, 2019 at 15:00

Room 011, Kitot Bldg., Faculty of Engineering

Single Sensor Trajectory Optimization for Best Emitter Localization

Abstract

Passive emitter localization has many civilian, commercial and military applications. The rapidly increasing utilization of smartphones and, therefore mobile applications, has created a high demand for location based services, both in commercial applications and social networking, for multiple and varied uses. Location based services are also critical to many businesses and government organizations to derive real insight from data tied to specific locations where activities take place. The spatial patterns that location-related data and services can provide is one of the most powerful and useful aspects when location is a common denominator in all of these activities and can be leveraged to better understand patterns and relationships. Accordingly, precise, and personalized localization solutions become a fundamental requirement of any commercial/social service.

In this presentation I will address the problem of a single platform trajectory optimization, aims to provide a targeted localization solution for a given emitter based on TOA measurements (i.e., minimizing the localization error of the emitter). The problem of trajectory optimization is a constraint non-convex optimization problem. Constraints arise due to physical limitation of the platforms, and geographical constrains such as restricted areas and safety zones in which the receiver is not allowed to travel. I will discuss two use-cases, a pre-mission design in which the entire trajectory is optimized based on prior knowledge on the emitter location. The second use-case is a real-time path design, in which the receiver begins with a coarse estimation of the emitter location, and searches for the next best way-point to travel to. In this case, the uncertainty in the estimation is incorporated into the optimization problem, in order to avoid over-optimistic steps in preliminary stages of the process. For both use-cases, we propose convex relaxation solutions based on Semi-definite relaxation methods and demonstrate their impressive results in terms of performance and robustness. Next, I will discuss the trajectory optimization of a pair of sensors which cooperate to localize an emitter based on TDOA observations. The presence of more than a single sensor imposes additional constraints on the pairwise distances between the sensors. We derive a solution based on the alternating direction of multipliers (ADMM) with intermediate steps carried out using the majorization minimization (MM) and SDR methods. The algorithm is demonstrated to outperform global optimizers such as genetic and the basin-hoping algorithms, both in terms of performance (better localization error) and speed of convergence.

As a final step, in order to provide an algorithmic solution that is capable of operating in real time environments, we introduce a differential dynamic programming (DDP) solution that is demonstrated to converge quadratically to good local optima, exploiting the desired properties of Newton method.

Speaker:  Moti Kadosh

M.Sc. student under the supervision of Prof. Ariel Shamir and Prof. Daniel Cohen Or

Wednesday, February 27^th, 2019 at 15:00

Room 011, Kitot Bldg., Faculty of Engineering

On the Role of Geometry in Geo-Localization

Abstract

            Humans can build a mental map of a geographical area to find their way and recognize places. The basic task we consider is finding the pose (position & orientation) of a camera in a large 3D scene from a single image. Our goal is to examine whether such a capability can be learned by a neural network. In particular, we aim to explore the role of geometry alone in geo-localization using CNNs, while ignoring the often available texture of the scene. We therefore deliberately avoid using texture or rich geometric details and use images projected from a simple 3D model of a city, which we term lean images. Lean images contain mostly information that relates to the geometry of the area viewed (edges, faces, or relative depth). We find that the network is capable of estimating the camera pose from the lean images, not by memorization but by some measure of geometric learning of the geographical area. The main contributions of this thesis are: (i) demonstrating the power of CNNs for recovering camera pose using lean images; and (ii) providing insight into the role of geometry in the CNN learning process.

You are invited to attend a lecture

Localized microwave-heating of basalts –
Melting, lava-like eruption, and dusty-plasma ejection
By
Yoav Shoshani
MSc student under the supervision of Prof. Eli Jerby

Abstract
Localized microwave-heating (LMH) and the consequent formation of hotspots by thermal-runaway instability may cause local melting, and even plasma ejection by various materials. The LMH phenomenon, generated by the microwave-material interaction, enables a rapid heating within a localized zone (a hotspot, much smaller than the microwave wavelength). This local heating process evolves up to the occurrence of a phase transition of the solid material into liquid, gas or plasma. In this study, LMH phenomena of basalt melting, lava-like eruptions and dusty-plasma ejection are presented. This study includes experimental and numerical investigations of the LMH interactions of basalts with microwaves, characterization of the dusty plasma and its nano-particle products by various diagnostic means (in- and ex-situ). Additionally, these experimental results are considered in various scientific and practical aspects, namely their potential relevance to natural ball-lightning and volcanic effects, as well as their significance for several applications, such as construction, mining, powder generation, microwave-induced breakdown spectroscopy (MIBS), and mineral extraction.

On Thursday, February 28, 2019, 11:00
Room 234, Wolfson Building