WILDTRACK-toolkit

Tools accompanying the WILDTRACK multi-camera dataset.

1. WILDTRACK annotations viewer

Graphical User Interface (GUI) illustration of the WILDTRACK dataset annotations. The script annotations_viewer.py draws the annotations for each of the views and displays the synchronized frames simultaneously. Input arguments (or symbolic links as the defaults):

1. dir_annotations. Directory which contains the JSON multi-view annotation files. Each JSON file lists the annotations of a single multi-view frame, as provided when downloading the dataset.
2. dir_frames. The root directory which contains subdirectories. Each such subdirectory contains the frames per camera (as provided when downloading the dataset). For example:

   frames/
   ├── view0/
   ...
   └── view6/
   

It warns the user if less or more then 7 subdirectories (views) are given. The script works if there exists at least one subdirectory in dir_frames. Provides basic functionality such as, go to the previous/next multi-view frame with step 1 or 10. Run: python annotations_viewer.py --help for details.

Dependencies

• Python 3.6 or later
• Tkinter
• Python JSON (JavaScript Object Notation) module
• OpenCV

Installing dependencies

Please ensure you have the latest Python and NumPy. We recommend the Anaconda package manager.

$ conda create --name wildtrack-toolbox
$ source activate wildtrack-toolbox
$ conda install -c anaconda opencv=3.4.1  # might take a while
$ conda install -c anaconda pillow

Use $ source deactivate when done.

2. Camera calibration - example code

Code that uses the calibration files of the WILDTRACK dataset, and the OpenCV library. intersecting_area.py draws the area considered for annotating the persons, in the WILDTRACK dataset, which lies approximately in the intersection between the fields of view of the seven cameras. Primarily, it generates 3D points on the ground plane (z-axis is 0) using a grid of size 1440x480, origin at (-300, -90, 0) in centimeters (cm), with a step/offset of 2.5 cm for both the axis. It then projects these points in each view. Finally, it stores the marked frames in the given output directory. Run: python intersecting_area.py --help for details.

Dependencies

• Python 3.6 or later
• OpenCV
• Python ElementTree package

Publication

WILDTRACK: A Multi-Camera HD Dataset for Dense Unscripted Pedestrian Detection. Tatjana Chavdarova, Pierre Baqué, Stéphane Bouquet, Andrii Maksai, Cijo Jose, Timur Bagautdinov, Louis Lettry, Pascal Fua, Luc Van Gool, François Fleuret. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 5030-5039 http://openaccess.thecvf.com/content_cvpr_2018/html/Chavdarova_WILDTRACK_A_Multi-Camera_CVPR_2018_paper.html

License

Copyright (c) 2008 Idiap Research Institute.

GNU GENERAL PUBLIC LICENSE, Version 3, 29 June 2007. WILDTRACK toolkit is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License version 3 as published by the Free Software Foundation. WILDTRACK toolkit is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with WILDTRACK toolkit. If not, see http://www.gnu.org/licenses/.

See also

• Multi Camera Calibration Suite
• Annotation tool