Skip to content

Unsupervised Generation of a Viewpoint Annotated Car Dataset from Videos

License

Notifications You must be signed in to change notification settings

lmb-freiburg/unsup-car-dataset

Repository files navigation

Code for the paper "Unsupervised Generation of a Viewpoint Annotated Car Dataset from Videos", ICCV 2015

Terms of use

The code is provided for research purposes only. Any commercial use is prohibited. If you are interested in a commercial use, please contact the copyright holder.

Please cite the following paper if you use this code or its parts in your research:

@InProceedings{SB15,
author       = "N. Sedaghat and T. Brox",
title        = "Unsupervised Generation of a Viewpoint Annotated Car Dataset from Videos",
booktitle    = "IEEE International Conference on Computer Vision (ICCV)",
year         = "2015",
url          = "http://lmb.informatik.uni-freiburg.de//Publications/2015/SB15"
}

Please report bugs and problems to Nima Sedaghat ( nima@cs.uni-freiburg.de )

Installation Requirements

External software/libraries -- not included:

External software/libraries -- included in the 'extern' directory:

Data

Download input car videos from here -- videos of 52 static cars, covering almost the whole 360 degrees around the car. Download resulting dataset from here -- viewpoint- and bounding box-annotated car images.

(Non-)Deterministic Behaviour

Some external components show random behaviour, which result in non-deterministic results that cannot be exactly reproduced across multiple runs. The closest you can get to a deterministic behaviour is by:

  1. setting 'settings.deterministic = 1' in the file setup.m
  2. setting 'param_deterministic_behaviour 1' in the file [VisualSFM Path]/bin/nv.ini

However, the results can still vary noticeably -- unless you use another SFM package, where you can control its behaviour completely.

About

Unsupervised Generation of a Viewpoint Annotated Car Dataset from Videos

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published