PEViD: Privacy Evaluation Video Dataset
Description
PEViD is a video dataset for objective and subjective evaluations of different types of privacy protection tools. The dataset consists of 20 video sequences (16 seconds each) of full HD resolution covering different video surveillance scenarios: walking, fighting, stealing, and dropping bag, in outdoor and indoor environments, as well as during day and night conditions. Most of the scenes are captured by two Canon HD cameras simultaneously from different angles to provide a clearer view of people in video sequences. All the participant are of various gender, race, dressed differently, and carrying various personal accessories. They have also read and signed the consent form, allowing free usage of these video sequences for research purposes. The dataset was annotated using Open Source ViPER-GT annotation tool with primary (face and body silhouette) and secondary (personal items and features, such as hair, skin regions, accessories, etc.) privacy sensitive regions. Additional personal information, such as gender, race, hair color, etc., was also recorded in annotation files, which are stored in XML format and can be easily processed in an automatic way.
Access
PEViD dataset is publicly available for research. We are still in the process of making the dataset available online for everyone to download. But until then, the interested parties can contact Pavel Korshunov (pavel.korshunov@epfl.ch) directly and will be provided with the access instructions.
Broken link!
The original link was: http://mmspg.epfl.ch/page-58275.html
License
The data is distributed under the CClicence - Creative Commons Attribution-Noncommercial-ShareAlike (CC BY-NC-SA 4.0). The full licence info can be viewed here (https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode).
References and Citation
If you use this Dataset in your work, please, cite the following paper in the references: [KE2013]
References
KE2013: Pavel Korshunov and Touradj Ebrahimi, PEViD: Privacy Evaluation Video Dataset, in proceedings of proc. SPIE Applications of Digital Image Processing XXXVI, volume 8856, 2013.