Skip to content

lzb863/GN-Net-Benchmark

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GN-Net Benchmark

For more information see https://vision.in.tum.de/gn-net

This code allows to generate pixel-correspondences for the GN-Net benchmark, presented in

"GN-Net: The Gauss-Newton Loss for Multi-Weather Relocalization".
L. von Stumberg, P. Wenzel, Q. Khan and D. Cremers. RA-L 2020

Install Dependencies

pip install opencv-python
pip install scipy==1.1.0
pip install scikit-image

Running

See example.py how to generate matches. At the moment matches can only be generated for the Carla sequences. For generating the matches either the groundtruth-depths or depths exported from Stereo DSO can be used. To specify this you can adjust the parameter use_dso_depths of the generate_matches_carla

For a start you can simply run

python example.py

About

Additional Code for the GN-Net benchmark. The data is available at http://vision.in.tum.de/gn-net

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%