Visual Sweden Pose Estimation

Dnr:

SNIC 2016/5-43

Type:

SNAC Small

Principal Investigator:

Per-Erik Forssén

Affiliation:

Linköpings universitet

Start Date:

2016-09-11

End Date:

2017-10-01

Primary Classification:

10207: Datorseende och robotik (autonoma system)

Webpage:

http://www.visualsweden.se

Allocation

Abstract

Visual Sweden needs common tools for visualisation and arrangement of camera data in 3D environments. This project will train convolutional neural networks to predict camera pose from image content in a limited environment. Training data will be generated by rendering views from a 3D model.