Contact
| Job Type
| Description
| Advisor
| Subject
| date
|
stern@bgu.ac.il
| Joint project with Rafael (research student position availiable)
| Imaging through the atmosphere may be severely degraded by turbulence and aerosols. The atmospheric degradation can be mitigated by means of Adaptive Optics (AO). The classical AO method relies on wavefront sensors such as the Shack–Hartmann sensor. This research will explore algorithms that estimate the waveform from conventional images and use them to control the AO system.
| Prof. Adrian Stern
| Deep Learning wavefront sensing for Adaptive Optics
| 1.8.2024
|
stern@bgu.ac.il
| Joint project with Rafael (research student position availiable)
| Phase information is lost when images are captured with a conventional camera. However, the phase may carry valuable information. For example, light passes through scatterers, and the properties of the scatterers are encoded in the phase of the image. Thus, recovering the phase can help to mitigate the image degradation. Traditional phase retrieval techniques use iterative algorithms. This research aims to use the Deep learning base method for this task and compare or combine them with traditional methods.
| Prof. Adrian Stern
| Deep Learning method for phase retrieval in highly scattering imaging conditions
| 1.8.2024
|
stern@bgu.ac.il
| Joint project with Rafael (research student position availiable)
| Conventional image formation through turbulence was extensively investigated, yielding useful analytical and statistical models. In this research, we wish to derive a model for the influence of turbulence on event images captured with the rapidly evolving neuromorphic cameras (see for example, Nature, vol. 629, no. 8014, p. 1034–1040, 2024)
| Prof. Adrian Stern
| Event imaging through scattering media
| 1.8.2024
|
stern@bgu.ac.il
| With Samsung
| Event Cameras are an emerging imaging modality (see, for example, Nature, vol. 629, no. 8014, p. 1034–1040, 2024). This research aims to implement a new patent we submitted that involves opto-algorithmic solution for converting an event camera to color, 3D and polarimetric camera. The research involves optical design and algorithms
| Prof. Adrian Stern
| Developing multi-dimesional event camera
| 1.8.2024
|
alinak@bgu.ac.il
| MSc./PhD.
| As global warming affects us all, green energy and clean air are urgent topics these days. Submicron particles can be grouped in a flow with controlled oscillations. The model led to the establishment of several systems that take advantage of this effect, among them a laboratory system that groups particles from the room's light to increase the capture of particles and viruses. However, particles of tens nanometers in size are challenging to trap. In framework of the project the optical forces will be generated to trap and manipulate a particle without an actual contact! This will be performed in addition to activating the fluid vibrations and acoustic vibrations to full optomechanical manipulation. The subject is innovative and suitable for a master's or third degree. Required background: background in wave propagation and electromagnetic fields. Preferred background: AI. Skills: Programming skills in standard languages (C, Python, etc.), construction of set-ups. Project overview: MSc./PhD. Scholarships are offered. In collaboration with Environmental Engineering
| Prof Alina Karabchevsky
| Optomechanical manipulation with particles in gas and water
| 16.1.2023 |
alinak@bgu.ac.il
| MSc./PhD.
| Required background: background in probability and algorithms. Prefer background: AI search/planning algorithms and Markov decision problems. Skills: Programming skills in standard languages (C, Python, etc.) Project overview: To be useful in applications involving the control of physical hardware or interaction with humans, intelligent systems must be robust and responsive. Currently, this conflicts with being general-purpose since the response time of a traditional planner cannot be guaranteed. In this project, we enable robust time-aware planning, thus significantly broadening the range of applications that can be addressed by intelligent systems. The research is in collaboration with Department of Computer Sciences.
| Prof Alina Karabchevsky
| Artificial intelligence to control the hardware
| 16.01.2023
|
ytshak@bgu.ac.il
| MSc/PhD.
| Vision
prostheses for blind people can be improved by analyzing and segmenting regions
and objects in a 3D scene. We use a novel camera array system that we recently
developed for 3D imaging, and Deep Learning methods for the scene analysis.
This project is supported by an ISF grant.
| Prof.
Yitzhak Yitzhaky
| 3D Scene
analysis using computer vision and array imaging, for vision rehabilitation
| 3.10.22
|
ytshak@bgu.ac.il
| MSc/PhD.
| We
recently developed a new approach for true emotion recognition (i.e., not
relying on facial expressions that may be faked), using multi-spectral imaging
and machine learning. This research aims to extend this approach and develop
applications such as lie detection. In this approach, we sense and analyze in
the video mild changes in facial blood flow due to emotional state. The
research is in cooperation with the Psychology Dept.
| Prof.
Yitzhak Yitzhaky
| True
emotion recognition via deep learning and multi-spectral imaging
| 3.10.2022
|
stern@bgu.ac.il
| MSc/PhD.
| An imaging technology developed in our lab was chosen to be installed on Beresheet 2 mission to the Moon. The research aims to optimize the sensing process and develop an algorithm for reconstructing compressively sensed spectral images of the Moon. For this purpose, a deep learning methodology will be used, similar to the one we developed for satellite spectral imaging of the Earth.
| Prof. Adrian Stern
| Learned compressive spectral imaging for Beresheet2 mission
| 15.09.2022
|
stern@bgu.ac.il
| MSc/PhD.
|
By
combining Deep Learning and Compressive Sensing concepts, we aim to develop
and demonstrate a camera that captures an image with no more than 10x10 pixel
sensors
|
| Prof. Adrian Stern
| Developing an imaging camera with extremely few sensors
| 15.09.2022
|
| MSc.
| A project funded by the EU involves multiple parties (Technion, Shiba hospital, a startup company, and us) develops a novel hyperspectral scanning system for pathological applications. The system uses simultaneous measurements of multiple fluorescent biomarkers and AI algorithms for image analysis, cell classification and drug response prognosis. The research conducted by our group aims to optimize interferometer spectral imaging by utilizing physics-informed deep learning methods.
| Prof. Adrian Stern
| Optimization of hyperspectral imaging for precision medicine in cancer diagnostics
| 15.9.2022
|
stern@bgu.ac.il
| MSc/PhD.
| Deep Learning (DL) algorithms have evolved to exhibit state-of-the-art performance for analyzing and processing captured data, such as images. However, most of the DL algorithms have been found to be vulnerable to so-called adversarial attacks that hamper their utilization for applications that require high reliability. Recently we have introduced a new defense paradigm for defending DL algorithms from adversarial attacks. The new paradigm is based on encryption in the optical domain and exhibits advantages unmet by software defense algorithms. The research aims to explore this new defense paradigm further.
| Prof. Adrian Stern
| Defending Deep Learning algorithms by Optical Image Encryption
| 15.9.2022
|