Past projects

SLAM

Goal: Enable real time, high accuracy pose tracking from one or two (stereo) cameras.

Integrating markers for absolute referencing and trajeectory optimization.

Next

Large Scale 3D

Goal: Storage and efficient visualization of billions of 3D points.

Mobile sensors provide huge amounts of 3D data, which need to be stored and processed. Often in real time. Standard techniques are not sufficient to achieve this goal. A super efficient 3D structure has been implemented relying on at compile time optimized BSP trees. No GPU required.

Semantic 3D fusion

Object recognition combined with incremental 3D reconstruction.

Semantic information about the surrounding is crutial for all automated processes dealing with the real world environments. The 3D structure for huge scenes has been re-used here and extended to additional information such as object class and color.

Next

Low power scene analysyis

Football game tracking and automated game recording on a smart phone.

The smartphone is used as a game recording camera, which automatically follows the game and rotates accordingly. No user interaction is required.

Next

Get in touch

me@ {this domain}