Visual Processing and Understating

Intel Domain Leader - Ronny Ronen
In addition to the ICRI-CI 3-layered capstone research, ICRI-CI host several research project in the visual processing and understanding domain. These research projects were added because of their potential high value to Intel, the industry, or the academia.



Saliency estimation in video

Intel Mentor: Vered Bar-Bracha (Intel WINS)

Academia Researcher(s): 
Prof. Ayellet Tal, Technion 
Prof. Lihi Zelnik-Manor, Technion

Participating Student(s): 
Yoav Liberman
Yonatan Dishon

Research Project Summary:
This research project deals with saliency estimation in video. The project aims to develop efficient algorithms that could possibly lead to real-time implementation in the future. The main application we’ll have in mind is video-conferencing as was requested by Intel. Current solutions to video saliency prediction provide either poor quality or require extensive runtime. We intend to develop solutions that are both efficient and provide high-quality, sufficient for application use. Such algorithms could be useful for many applications, and in particular to improve quality of transmitted videos.

Statistics of depth images

Academia Researcher(s):  Prof. Yair Weiss, HUJI

Participating Student(s): Dan Rosenbaum

Research Project Summary:
Cameras that can capture both RGB and Depth (RGB-D) are becoming widely available.  Due to the nature of the depth sensor, the resolution and quality of the depth channel is far inferior to the RGB. We seek to use machine learning to estimate the joint statistics of RGB-D and use these statistics to improve depth resolution and quality.

Yair Weiss – Publications

 

Mental phenotyping from 3D cameras

Academia Researcher(s): Prof. Daphna Weinshall (HUJI)

Participating Student(s): 
Talia Tron
Nomi Vinokurov 
Daniel Hadar

Research Project Summary:
Recent technological advances in 3D cameras are primarily being used for human computer interactions and gaming. In this research project we will use computer vision and machine learning techniques to develop tools which will allow us to interpret human’s non-verbal behavior (including bodily gestures and facial expressions) for the purpose of monitoring people’s mental state, in particular as related to such pathologies as Parkinson’s disease and Schizophrenia.

Daphna Weinshall – Publications

 

Blind Video: Video without photographers

Academia Researcher(s): Prof. Shmuel Peleg (HUJI)

Research Project Summary

Blind Video, video without photographers, includes surveillance video and video from wearable cameras. Most data in such video is irrelevant. This research project will improve approaches that give people a fast and efficient access to the important information in such videos. In particular the multi-camera scenario will be explored.

Publications