Filters
cris logo
-
New search Filters
Select from list
Displayed from Show more
You have reached the maximum number of displayed search results.
All results displayed
No results are available for the search performed
Displayed from
You have reached the maximum number of displayed search results.
All results displayed
Loading results
Obtaining statistical data

Projects / Programmes source: ARIS

Face deidentification with generative deep models (FaceGEN)

Edit
Research activity

Code Science Field Subfield
2.06.02  Engineering sciences and technologies  Systems and cybernetics  System theory and control systems 

Code Science Field
P175  Natural sciences and mathematics  Informatics, systems theory 

Code Science Field
2.02  Engineering and Technology  Electrical engineering, Electronic engineering, Information engineering 
Keywords
deidentification, deep learning, generative models, artificial intelligence, face syntehsis, privacy protection
Evaluation (rules)
source: COBISS
Evaluation of bibliographic research performance indicators according to ARIS methodology
Citations Citations for bibliographic records in COBIB.SI that are linked to records in citation databases
source: WoS source: Scopus source: COBISS
Researchers (15)
no. Code Name and surname Research area Role Period No. of publications
1.  22472  PhD Borut Batagelj  Computer science and informatics  Researcher  2019 - 2022  192 
2.  54426  Tea Brašanac  Computer science and informatics  Researcher  2020 - 2021  0 
3.  11805  PhD Simon Dobrišek  Computer science and informatics  Researcher  2019 - 2022  284 
4.  53820  PhD Žiga Emeršič  Computer science and informatics  Researcher  2019 - 2022  84 
5.  38118  PhD Klemen Grm  Systems and cybernetics  Researcher  2019 - 2022  45 
6.  31985  PhD Janez Križaj  Systems and cybernetics  Researcher  2019 - 2022  39 
7.  19226  PhD Peter Peer  Computer science and informatics  Researcher  2019 - 2022  408 
8.  51910  Martin Pernuš  Computer science and informatics  Junior researcher  2019 - 2022  14 
9.  21310  PhD Janez Perš  Systems and cybernetics  Researcher  2019 - 2022  238 
10.  53724  Peter Rot  Computer science and informatics  Researcher  2019 - 2022  21 
11.  39512  Robert Sedevčič    Technical associate  2019 - 2020  0 
12.  09581  PhD Franc Solina  Computer science and informatics  Researcher  2019 - 2022  639 
13.  28458  PhD Vitomir Štruc  Systems and cybernetics  Head  2019 - 2022  360 
14.  11380  PhD Mario Žganec  Metrology  Researcher  2019 - 2022  98 
15.  12000  PhD Jerneja Žganec Gros  Computer science and informatics  Researcher  2019 - 2022  290 
Organisations (3)
no. Code Research organisation City Registration number No. of publications
1.  1538  University of Ljubljana, Faculty of Electrical Engineering  Ljubljana  1626965  27,752 
2.  1539  University of Ljubljana, Faculty of Computer and Information Science  Ljubljana  1627023  16,234 
3.  1986  ALPINEON R & D  Ljubljana  1820931  387 
Abstract
The wide-spread availability of mobile devices with their imaging capabilities have transformed the social conventions and expectations around the appropriate use of images and videos. While it is easier than ever to conduct and archive video calls, capture images and share them online or collect and process large-scale imagery, appropriate measures need to be taken to ensure that this data is not misused and the privacy of people is protected. Especially exposed here are vulnerable demographic groups, such as children or teenagers, who are often unaware of the potential risks of recklessly sharing their personal information, images and videos online. The need for technology capable of ensuring a high level of privacy was already identified by the European Union and expressed in the form of the EU Data Protection Directive (95/46/EC). This directive was later replaced by the General Data Protection Regulation (GDPR), which doubles down on the need for privacy enhancing technologies (PET) and, thus, directly supports initiatives aimed at developing technology that positively affects the privacy of people. Both the EU data Protection Directive and its recent successor GDPR point to an evident gap between the current state-of-technology and the imminent need within our environment for so-called PET technology capable of ensuring suitable levels of privacy protection, while not affecting the utility and usability of the data. A possible solution to address this gap can be found in de-identification technology. De-identification is defined as the process of concealing or replacing personal identifiers in personal information, in order to prevent the disclosure and use of data for purposes unrelated to the purpose for which the data was originally collected (i.e., to prevent function creep effects). In image and video data, this process is related to de-identification of the facial region, which carries most of the information related to the identity of the individuals in the imagery. A common approach to facial de-identification is to conceal the original facial appearance through filtering and masking or in case of sufficiently high-resolution images, with surrogate faces typically generated by Active Appearance Models. While such techniques are the standard in the literature, they suffer from several drawbacks: 1) they often destroy too much of the information content of the imagery and consequently affect the utility of the data after deidentification, 2) they struggle with low-quality and low-resolution images, 3) they depend on several potentially error-prone processing steps and 4) they often make restricting assumptions that limit their applicability in real-life applications. Within the proposed fundamental research project “Face deidentification with Generative Deep Models” (FaceGEN), we will strive to address these shortcomings and conduct research on deidentification technology with a particular focus on deep learning, which has recently been shown to be a highly effective tool for various computer vision and machine learning problems. Our goal is to develop deep generative models and conditional face synthesis techniques that can be used for deidentification with still images and formal privacy protection schemes, such as k-anonymity, but also with video, where multiple faces in cluttered and unconstrained scenes may appear in the data. The main tangible results of the project will be novel and highly robust deidentification technology as well as new generative deep models and input-conditioned image synthesis techniques that are able to deidentify all parts of the facial data photo-realistically (including soft-biometric cues), while still preserving non-identity related information, such as demographics, gender, and the like. We expect the developed models to exhibit unprecedented deidentification performance and provide convincing, naturally appearing deidentification results, far beyond what is achievable w
Significance for science
The deidentification-related goals of the project are in line with the recent GDPR regulative, the outreach by NIST and efforts from private companies towards privacy enhancing technology. The planned project will provide tangible technology and theoretically well-founded procedures that are in line with the expectations expressed in the listed documents. Private companies interested in PET technologies will have the necessary groundwork for developing commercial products enabling them to compete in the international AI and PET market. The goals related to the deep-learning and AI are expected to be even more far reaching with state-of-the-art generative models being useful across numerous sectors (IT, electronics, medicine, arts, etc.), industries and applications (e.g., smart assistants, self-driving cars, IoT, computer graphics, recognition technologies, mobile applications) that are classified by major companies as crucial for future development. The expertise will contribute to strengthening Europe’s R&D capacities in a critical sector with high added value.  FaceGEN will advance science in the areas of computer vision, machine learning and biometrics by generating Key Enabling Technologies (KETs) with a highly specialized research program. The planned FaceGEN tasks will contribute to progress beyond the state-of-the-art in deidentification and generative networks. Below are a few points where we expect our project to make a lasting impact: • Scientific impact:  o on broader areas of AI and deep learning, o on new ideas in unsupervised vision problems, o on the development of new research directions and new approaches to generative deep models. • Technological impact:  o on the development of contemporary state-of-the-art de-identification techniques for image data, o on guidelines and recommendations for new standards for de-identification, o on the development of standardized evaluation methodologies for de-identification technology,  o on novel ideas in related areas, such as biometrics and forensics. • Socio-economic impact:  o on imminent availability of PET technology for vulnerable demographic groups, o on improvement of quality of life of people due to technology that provides reasonable trade-offs between privacy protection and data utility/usability, o on the business sector by providing the necessary groundwork and ensuring the relevant know-how to develop commercial products capable of privacy – preserving image/video acquisition.   o on standards and examples of good practices in ICO certified technology complying with the principles of privacy-by-design. • Micro impact (at the institutional level): o on training the next generation of researchers in a hot area with a high applicability and potential market value,  o on the expertise of the project partners and competitiveness in Slovenia’s smart specialization strategy and EU funded project on PET technologies.
Significance for the country
The deidentification-related goals of the project are in line with the recent GDPR regulative, the outreach by NIST and efforts from private companies towards privacy enhancing technology. The planned project will provide tangible technology and theoretically well-founded procedures that are in line with the expectations expressed in the listed documents. Private companies interested in PET technologies will have the necessary groundwork for developing commercial products enabling them to compete in the international AI and PET market. The goals related to the deep-learning and AI are expected to be even more far reaching with state-of-the-art generative models being useful across numerous sectors (IT, electronics, medicine, arts, etc.), industries and applications (e.g., smart assistants, self-driving cars, IoT, computer graphics, recognition technologies, mobile applications) that are classified by major companies as crucial for future development. The expertise will contribute to strengthening Europe’s R&D capacities in a critical sector with high added value.  FaceGEN will advance science in the areas of computer vision, machine learning and biometrics by generating Key Enabling Technologies (KETs) with a highly specialized research program. The planned FaceGEN tasks will contribute to progress beyond the state-of-the-art in deidentification and generative networks. Below are a few points where we expect our project to make a lasting impact: • Scientific impact:  o on broader areas of AI and deep learning, o on new ideas in unsupervised vision problems, o on the development of new research directions and new approaches to generative deep models. • Technological impact:  o on the development of contemporary state-of-the-art de-identification techniques for image data, o on guidelines and recommendations for new standards for de-identification, o on the development of standardized evaluation methodologies for de-identification technology,  o on novel ideas in related areas, such as biometrics and forensics. • Socio-economic impact:  o on imminent availability of PET technology for vulnerable demographic groups, o on improvement of quality of life of people due to technology that provides reasonable trade-offs between privacy protection and data utility/usability, o on the business sector by providing the necessary groundwork and ensuring the relevant know-how to develop commercial products capable of privacy – preserving image/video acquisition.   o on standards and examples of good practices in ICO certified technology complying with the principles of privacy-by-design. • Micro impact (at the institutional level): o on training the next generation of researchers in a hot area with a high applicability and potential market value,  o on the expertise of the project partners and competitiveness in Slovenia’s smart specialization strategy and EU funded project on PET technologies.
Confirmation required
Views history
Favourite