 |
De-identification for Privacy Protection in Surveillance Systems - DePPSS
HRZZ Research Project No. 6733
|
 |
| |
| |
De-identification for Privacy Protection in Surveillance Systems - DePPSS |
Project summary |
|
 |
 |
 Owing to advances of video surveillance systems it is possible, besides detection of people's behavioural anomalies, by using of reliable recognition software to perform identification and simultaneously trace people in the real time. Whilst it is recognized that there are justified reasons for acquisition and sharing videos in manners such as security, bio-terrorism surveillance applications, low enforcement and forensics, there is also a strong need for protecting the privacy of the guiltless individuals who are inevitably captured in the recordings. There are no doubts that video surveillance is privacy intrusive because it allows the observation of certain information that is considered privacy intrusive. Face has central role in the process of human recognition and identification in videos, thus the special attention has to be devoted to the face de-identification methods for privacy protection. De-identification addresses, in context of our project, the automated methods of concealing or/and removing face identifiers of individuals captured in videos.
The importance of privacy protection is mirrored in documents such as UN Universal Declaration of Human Rights, Article 12. and EU’s Data Protection Directive (95/46/EC) as well as commissioned review of the Data Protection Directive (95/46/EC) by the EU Information Commissioner’s Office (ICO) from July 2008. This review (ICO, July 2008) takes into account that in 13 years since the Directive 95/46/EC came into force, the world has seen dramatic changes in the way personal data is accessed, processed and used. At the same time, the general public has become increasingly aware of the potential danger for their personal data to be abused.
Beside scientific objectives and goals (robust face localization, novel methods for face de-identification in videos, preserving data utility and naturalness in de-identification videos, privacy protection), the main objectives of the proposed project are human resources (early stage researchers and PhD students) development by transfer of knowledge in the fields of advanced technologies to carry out problem-oriented research, and increasing a competitiveness level of the research team for Horizon 2020 project applications.
|
|
|
|
|
|
- Detailed review and analyses of previous approaches to the problem of face de-identification for still images, and videos of individuals or groups of people at places such as airports, railway and bus terminals and public transportation vehicles
- Development of robust face localization methods in videos adopted to de-identification process
- Development of novel algorithms and methods for automatic concealing of face identifiers with preserving the data utility and naturalness in videos
- Development of novel algorithms and methods for reversible de-identification
- Set up the experimental camera surveillance system with inbuilt face de-identification
- Evaluation of privacy protection solutions, data utility and naturalness of de-identified video in video surveillance
- Dissemination of the results of research
|
|
|
|
|
|
Journals:
- 4. Meden, Blaž;Malli, Refik Can; Fabijan, Sebastjan;K. Ekenel, Hazim;Štruc, Vitomir;Peer, Peter Face Deidentification with Generative Deep Neural Networks: IET Signal Processing Special Issue on Deidentification, July (2017); 1-17
Link
- 3. Ribaric, Slobodan;Ariyaeeinia, Aladdin; Pavesic, Nikola De-identification for privacy protection in multimedia content: A survey, Signal processing. Image communication (0923-5965) 47 (2016); 131-151
Link
- 1. Ivasic-Kos, Marina;Pobar, Ivan;Ribaric, Slobodan Two-tier image annotation model based on a multi-label classifier and fuzzy-knowledge representation scheme. // Journal of Pattern Recognition 52, pp.287-305, 2016
Link
- 2. Ivasic-Kos, Marina;Ipsic, Ivo;Ribaric, Slobodan A knowledge-based multi-layered image annotation system. // Expert Systems with Applications 42, pp.9539-9553, 2015
Link
Proceedings articles:
- 1. Soldic, Martin;Marcetic, Darijan;Maracic Marijo;Mihalic Darko;Ribaric, Slobodan Real Time Face Tracking under Long-Term Full Oclussions // 10th International Symposiumon Image and on Signal Processing and Analysis (ISPA 2017), 2017 pp. 147-152
Link
- 2. Marcetic, Darijan;Soldic Martin; Ribaric, Slobodan Hybrid Cascade Model for Face Detection in the Wild Based on Normalized Pixel Difference and a Deep Convolutional Neural Network // Lecture Notes in Computer Science,2017. pp. 379-390
Link
- 3. Meden, B.;Emersic, Z.;Struc, V.;Peer, P. k-Same-Net:Neural Network Based Face Deidentification // In Bioinspired Intelligence (IWOBI), 2017 pp. 1-7
Link
- 1. Marcetic, Darijan;Hrkac, Tomislav; Ribaric, Slobodan Two-stage cascade model for unconstrained face detection // First International Workshop on Sensing, Processing and Learning for Intelligent Machines (SPLINE), 2016 pp. 2016. 21-24
Link
- 2. Hrkac, Tomislav;Brkic, Karla; Ribaric, Slobodan;Marcetic, Darijan Deep Learning Architectures for Tattoo Detection and De- identification, Sensing, Processing and Learning for Intelligent Machines (SPLINE), 2016, pp. 45-49
Link
- 3. Marcetic, Darijan; Ribaric, Slobodan Deformable Part-based Robust Face Detection under Occlusion by Using Face Decomposition into Face Components // Proceedings of the 39th International Convention MIPRO 2016, BiForD 2016 pp. 1365-1370
Link
- 4. Marcetic, Darijan;Samarzija, Branko;Soldic, Martin;Ribaric, Slobodan Face De-identification for Privacy Protection in Surveillance Systems. // ROSUS Maribor :Slovenian Section IEEE, 2016. 85-91
Link
- 5. Ribaric, Slobodan;Pavesic, Nikola An Overview of Face De-identification in Still Images and Videos. // 11th IEEE International Conference on Automatic Face and Gesture Recognition FG 2015
Link
Invitations & invited speeches:
- 1. De-identification for Privacy Protection – COST Action IC1206
Link
Research projects:
Tisak:
- Zitto, lo smartphone ti ascolta //Corriere Della Sera, 13.12.2015. Link
Training school:
- 1. Ribaric, Slobodan De-identification for Privacy Protection in Multimedia Content: A Survey. // COST ACTION IC1206, The First Training School, Limassol, 7th-11th October, 2015
Link
Kvalifikacijski doktorski ispit:
- 1. Soldic, Martin Robusna detekcija lica
Link
|
|
|
|
|
|
|
|
|
|
|
|
|
Some of the current and previous projects (both national and international) the Laboratory was involved in are given in the list below:
|
|
|
|
|
|
|
| |
|