The Toulouse Vanishing Points Dataset

We present the Toulouse Vanishing Points Dataset, a public photographs database of Manhattan scenes taken with an iPad Air 1. The purpose of this dataset is the evaluation of vanishing points estimation algorithms. Its originality is the addition of Inertial Measurement Unit (IMU) data synchronized with the camera under the form of rotation matrices. Moreover, contrary to existing works which provide vanishing points of reference in the form of single points, we computed uncertainty regions.

Some photos of the dataset with their ground truth line segments:

• V. Angladon, S. Gasparini and V. Charvillat (2015). The Toulouse Vanishing Points Dataset. Proceedings of the 6th ACM Multimedia Systems Conference. With associated slides of the oral presentation.
• Orasis'15 article (French conference) to come.

The dataset containing the photographs and the ground truths can be downloaded here: dataset.zip

We developed an iOS application to record the orientation of the camera at the moment a photo is taken.
This application is compatible with iPhone and iPad running on iOS 7.
The orientation is provided in the form of change of basis matrices from the world reference frame to the camera frame and is stored in the EXIF UserComment field.
This application is not present on the Apple App Store, however we provide its source code released under the BSD license.
Download the source code of the iOS application.

This web application enables to accurately draw line segments on an image and to store their endpoints in the JSON format and the Matlab .mat format.
Download the source code of the ground truth editor.
We are currently modifying this editor to make it work on a production web server.
A web hosted version of this ground truth editor will be available in the coming weeks.

We provide Matlab source code to easily load the ground truth data associated to the images and to compute the double wedges intersections on your own images.
Download the source code.
This archive contains a subset of our source code.
Beware: the source code is not compatible with GNU Octave due to the use of classes.

• Loading and displaying the data with Matlab
• Computing loading and displaying the uncertainty regions with Matlab

We would like to thank the following people who took part in this project:
Simone Gasparini, Vincent Charvillat, Marie-Anne Bauda, Axel Carlier, Bastien Durix and Yvain Queau.