engintola.com/research

recent work



I'm currently occupied with my start-up Aurvis - please check out the company page for my most recent work.


I attended EPFL - Computer Vision Laboratory for my PhD Studies and most of the research work presented here is accomplished towards that goal. During my PhD work, I focused on finding efficient methods on the multiple view geometry problem for large image sets containing very high resolution images.

To reach my publications go to here.

2010/ efficient multiple view stereo


efficient multiple view stereo We present a new approach for large scale multi-view stereo matching, which is designed to operate on ultra high resolution image sets and efficiently compute dense 3D point clouds. We show that, by using a robust descriptor for matching purposes and high resolution images, we can skip the computationally expensive steps other algorithms require. As a result, our method has low memory requirements and low computational complexity while producing 3D point clouds containing virtually no outliers.For more details visit here.

2009/ deformable surface reconstruction


deformable surface
                                                                                reconstruction it has recently been shown that deformable 3d surfaces could be recovered from single video streams. however, existing techniques either require a reference view in which the shape of the surface is known a priori, which often may not be available, or require tracking points over long sequences, which is hard to do. in this work we overcome these limitations and reconstruct a deformable surface using only a video sequence of the surface without using any reference configuration. see our results in here.

2009/ virtual view generation with a hybrid camera array


vvg hybrid system I recently did an internship at microsoft research-redmond with communication and collaboration systems group where I built a hybrid camera array with one infrared pulse based depth camera and 4 regular cameras. To see the details of the study, click here.

2008/ dyvine: dynamic visual networks


dyvine i was part of the dyvine - dynamic visual networks consortium. dyvine was a european union fp6 project aiming "to design, develop and test a representative version of a surveillance network based on visual sensors (images and video, in situ or airborne) which can be configured as a function of the requirements and events." I worked on the mobile camera localization part of the project. for more details...

2008/ daisy: a fast local descriptor for dense matching


daisy In this work, we present a new descriptor which we call DAISY and a matching algorithm to estimate depth maps from 2 wide baseline images. we show that computing a descriptor at every point of an image is feasible and can be accomplished very efficiently. additionally, using descriptors in stereo matching improves the state-of-the art results and enables us to work on images taken from very different view-points.
for details...

past research


In the past, I have worked on structure from motion for static scenes and dynamic scenes. for the details of these works you can visit:

+ structure from motion in static scenes
+ structure from motion in dynamic scenes

you can also inspect my master's thesis: multi-view 3d reconstruction of a scene containing independently moving objects


general interests


Although I am working on dense reconstruction and image based rendering oriented topics right now, I have a broad range of interests like all things Bayesian, machine learning, object detection and tracking, elusive occlusions and all 3-d related stuff.


Updated: Thursday, July 24, 2014 10:29:21 +0200