Semantic Drone Dataset

This dataset focuses on semantic understanding of urban scenes for increasing the safety of autonomous drone flight and landing procedures. For more information have a look at the dataset page.

Stereo Training Data Generation Tool

This tool is allows you to generate training data for confidence learning in stereo vision. As input the tool requires depth maps of a query algorithm together with known stereo camera poses. The tool will analyze contradictions and consistencies between depth maps from different view points and thus automatically label images that can be used for training.

If you use our algorithm in any of your publications, please cite the following paper:
Using Self-Contradiction to Learn Confidence Measures in Stereo Vision 
Christian Mostegel and Markus Rumpler and Friedrich Fraundorfer and Horst Bischof,
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016.


Camera Calibration Tool

This software package enables the calibration of a camera according to the basic principles of planar target based calibration. In the calibration routine the focal length, camera center and distortion parameters are optimized for a simplified camera model, where the aspect ratio is one and the skew factor zero. Note that this approximation is valid for almost all standard consumer and industrial digital cameras. 

If you use our software in any of your publications, please cite the following paper:

Flexible and User-Centric Camera Calibration using Planar Fiducial Markers
Shreyansh Daftry, Michael Maurer, Andreas Wendel, Horst Bischof.
In Proceedings of the British Machine Vision Conference (BMVC), Bristol (UK), 2013.

[Download] (Windows7 x64 / Windows8 x64)

Line3D++, Multi View Stereo using Line Segments

Line3D++ is an SfM post-processing tool which enables you to generate abstract line-based 3D models from arbitrary urban or indoor scenes. It is designed as an easy to use C++ library, which can be easily integrated into the SfM pipeline of your choice. For your convenience, executables which can process bundler, VisualSfM, Pix4D, OpenMVG, and mavmap results are included.

If you use our algorithm in any of your publications, please cite the following paper:
Efficient 3D Scene Abstraction Using Line Segments
Manuel Hofer, Michael Maurer, Horst Bischof,
In Computer Vision and Image Understanding (CVIU), 2016.