Lin Yuan

MSc Semester Project

Title

Automatic multi-view image data generation and its interactive rendering on android platform mobile device

Candidate

Lin Yuan

Supervisor

Prof. Dr. Touradj Ebrahimi

Assistants: Martin Rerabek and Philippe Hanhart

Expected

June 8, 2012

Place

EPFL

Description

One of the alternative methods to display 3D content on the conventional 2D screen is based on the monocular cue called motion parallax. One of the ways how to employ motion parallax based method is to use multi-view image or video content and then to display relevant view for each position of viewer face. The goal of this work is to develop the python script which allows the user to generate set of cameras in different setup (1D, circular, grid) and render the scene form Blender with each of them. Another goal is to develop an android platform application which will use the frontal camera for the face detection and display relevant view from multi-view dataset.

More specifically the following tasks should be performed:

  • Research and review the recent developments about face detection algorithm.
  • Research and investigate the properties of Blender 3D rendering software.
  • Develop the python script which allows automatic generation of N-cameras in different setup (1D parallel setup, grid setup, circle setup).
  • Install and setup tools for android applications development.
  • Develop the android application to display relevant view of the multi-view set based on viewer face position – assuming only the tilt movement i.e. 1D camera setup.
  • Optionally, enhance the application for circular camera setup and for grid camera setup.
  • Document all the development process and final version of all codes.