Giskar - Distance estimation through single camera features applied to Neural Networks

From AIRWiki
Jump to: navigation, search

Part 1: project profile

Project name

Giskar.

Project short description

Distance estimation through single camera features fed into a neural network.

Dates

Start date: 2008/10/01

End date: 2009/04/20

Website(s)

Work in progress.

People involved

Alessandro Stranieri User:AlessandroStranieri

Project head(s)

Pier Luca Lanzi - User:PierLucaLanzi

Students currently working on the project

Alessandro Stranieri User:AlessandroStranieri

External personnel:

dott. Martin Felder (Postdoc researcher at the chair of Robotics and Embedded Systems, TUM) - felder (at) in (dot) tum (dot) de

Laboratory work and risk analysis

Laboratory work for this project will be mainly performed at AIRLab/DEI. There won't be any risk as it involves only software development.

Part 2: project description

The approach is developed as a C++ library and a set of Python scripts. The C++ library is used to extract features from a set of images. The features and a set of distances read through a laser scanner are then fed into a supervised learning algorithm to train a neural network. The neural network is then used to perform real-time distance estimation on images acquired through a camera.