Frédéric Devernay: Computer Vision Research

Job offers / internships

Plese email me (frederic.devernay at inria.fr) if you are interested in any of the following open positions.

PhD positions

• High resolution 3D panoramic video processing / Vidéo panoramique 3D haute résolution

Master / M2R projects

• Multi-view video using unsynchronized cameras
• Optimizing the plenoptic camera for digital refocussing
• Parallax compensation on 360° panoramic video

Engineering degree projects

• Plenoptic camera simulator

Research activities

Here are a few highlights on my current and past research activities. Please note that some of the older stuff may be broken. Feel free to use any media found here for teaching.

Publications for all of these are available for download.

• (2005-) 3D Stereoscopic Cinema Meets Computer Vision
• (2004-2009) A Variational Method for Scene Flow Estimation from Stereo Sequences (media and source code)
• (1994) From Projective to Euclidean Reconstruction (media)
• (1994) Computing Differential Properties of 3-D Shapes from Stereoscopic Images without 3-D Model (media)

Sample code

Here is some sample code I did to better understand some published algorithms. Please tell me if you think something is wrong in these sample implementations.

• A PROSAC implementation written in C (previous versions: 1.5, 1.4, 1.3, 1.2, 1.1).
• All RANSAC derivations must choose a random k-subset of {0,1,...n-1}. Most implementations use inefficient algorithms to do so (take a look at OpenCV's getSubset from cvfundam.cpp...). Here's the right way to do so: deal.c. If you find an even better solution, please tell me.
• Generating a Chi-squared table is easy.