A5019800134- Biometric Identification and Security
- Face recognition and analysis
- Image Retrieval and Classification Techniques
- Advanced Image and Video Retrieval Techniques
- Face and Expression Recognition
- Advanced Statistical Methods and Models
- Digital Media Forensic Detection
- User Authentication and Security Systems
- Statistical Distribution Estimation and Applications
- Advanced Data Compression Techniques
- Video Analysis and Summarization
- Video Surveillance and Tracking Methods
- Image Processing Techniques and Applications
- Advanced Vision and Imaging
- Image Processing and 3D Reconstruction
- Advanced Steganography and Watermarking Techniques
- Handwritten Text Recognition Techniques
- Forecasting Techniques and Applications
- Neural Networks and Applications
- Data Management and Algorithms
- Chaos-based Image/Signal Encryption
- Video Coding and Compression Technologies
- Statistical Methods and Bayesian Inference
- Educational Robotics and Engineering
- Industrial Vision Systems and Defect Detection
Idiap Research Institute
2016-2024
Zimmer Biomet (Switzerland)
2017-2024
Gauhati University
2018
IAP Research (United States)
2018
Fraunhofer Institute for Computer Graphics Research
2018
University of Rajshahi
1991-2014
University of Malaya
2011
École Polytechnique Fédérale de Lausanne
1996-2003
Michigan State University
1992-2003
The digital watermarking schemes of today use pixels (samples in the case audio), frequency or other transform coefficients to embed information. drawback such is that watermark not embedded perceptually significant portions data. We refer techniques as first generation schemes. In this paper we introduce concept second which, unlike schemes, employ notion data features. propose a scheme based on point features images using scale interaction technique 2D continuous wavelets. are used compute...
It is straightforward to apply general schemes for authenticating digital data the problem of images. However, such a scheme would not authenticate images that have undergone lossy compression, even though they may been manipulated otherwise. We propose visual content This robust compression noise, but will detect deliberate manipulation image-data. The proposed based on extraction feature-points from image. These are defined so as be relatively unaffected by compression. set given image...
For face authentication to become widespread on mobile devices, robust countermeasures must be developed for presentation-attack detection (PAD). Existing databases evaluating face-PAD methods do not capture the specific characteristics of devices. We introduce a new database, REPLAY-MOBILE, this purpose.1 This publicly available database includes 1,200 videos corresponding 40 clients. Besides genuine videos, contains variety presentation-attacks. The also provides three non- overlapping...
In recent years, software-based face presentation attack detection (PAD) methods have seen a great progress. However, most existing schemes are not able to generalize well in more realistic conditions. The objective of this competition is evaluate and compare the generalization performances mobile PAD techniques under some real-world variations, including unseen input sensors, instruments (PAI) illumination conditions, on larger scale OULU-NPU dataset using its standard evaluation protocols...
We investigate the vulnerability of convolutional neural network (CNN) based face-recognition (FR) systems to presentation attacks (PA) performed using custom-made silicone masks. Previous works have studied CNN-FR 2D PAs such as print-attacks, or digital- video replay attacks, and rigid 3D This is first study consider flexible Before embarking on research detecting a new variety PA, it important estimate seriousness threat posed by type PA. In this work we demonstrate that custom masks do...
This paper proposes a technique for spatio-temporal segmentation to identify the objects present in scene represented video sequence. processes two consecutive frames at time. A region-merging approach is used scene. Starting from an oversegmentation of current frame, are formed by iteratively merging regions together. Regions merged based on their mutual similarity. We propose modified Kolmogorov-Smirnov test estimating temporal The process weighted, directed graph. Two complementary...
The vulnerability of deep-learning-based face-recognition (FR) methods, to presentation attacks (PA), is studied in this study. Recently, proposed FR methods based on deep neural networks (DNN) have been shown outperform most other by a significant margin. In trustworthy face-verification system, however, maximising recognition-performance alone not sufficient – the system should also be capable resisting various kinds attacks, including PA. Previous experience has that PA systems tends...
High-quality custom-made 3D masks are increasing becoming a serious threat to face-recognition systems. This is driven, in part, by the falling cost of manufacturing such masks. Research face presentation-attack detection (PAD) general, and also specifically for 3D-mask based attacks, has mostly concentrated on imagery visible-light range wavelengths (RGB). We look beyond spectrum find potentially easier solutions challenge (PAD). In particular, we explore use near-infrared (NIR) thermal...
The vulnerability of face recognition systems towards evolving presentation attacks has drawn significant interest in the last decade. In this paper, we present an empirical study on both analysis and attack detection for commercial (FRS) using custom 3D silicone masks corresponding to real subjects. To end, a new database is collected consisting 8 together with bona fide presentations subjects three different devices (smartphones). FRS effectively evaluated two well-known...
This work focuses on detecting presentation attacks (PA) mounted using custom silicone masks. Face recognition (FR) systems have been shown to be highly vulnerable PAs based such Here we explore the use of multispectral data (color imagery, near infrared (NIR) imagery and thermal imagery) for face attack detection (PAD), specifically against mask attacks. Using a new dataset (XCSMAD) representing 21 made masks, establish baseline performance several commonly used face-PAD methods, different...
With face-recognition (FR) increasingly replacing fingerprint sensors for user-authentication on mobile devices, presentation attacks (PA) have emerged as the single most significant hurdle manufacturers of FR systems. Current machine-learning based attack detection (PAD) systems, trained in a data-driven fashion, show excellent performance when evaluated intra-dataset scenarios. Their typically degrades significantly cross-dataset evaluations. This lack generalization current PAD systems...
For the automotive industry moving towards personalized applications and experiences, identification of person inside vehicle is necessary; it must be carried out in a secure manner. In this paper, we propose unique face presentation attack detection (PAD) system for operation passenger vehicle. A typical <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">in-vehicular</i> PAD required to function with several constraints such as bounded sensing...
For content based image retrieval using shape descriptors, most approaches so far extract information from a segmentation of the image. Shape features derived on specific are not suitable for images containing complex structures. Further, static useful only small set queries. In this paper we discuss limitations such boundary features, and propose an alternative characterization technique orientation radiograms. A working system approach is described sample results presented full-image query.
A multichannel filtering-based texture segmentation method is applied to a variety of document image processing problems: text-graphics separation, address-block location, and bar code localization. In each these problems, the text context or in considered define unique texture. Thus, all three analysis problems can be posed as problems. Two-dimensional Gabor filters are used compute features. Both supervised unsupervised methods identify regions images. The performance classification scheme...
Normal 0 false EN-US X-NONE /* Style Definitions */ table.MsoNormalTable {mso-style-name:"Table Normal"; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-noshow:yes; mso-style-priority:99; mso-style-parent:""; mso-padding-alt:0in 5.4pt 0in 5.4pt; mso-para-margin:0in; mso-para-margin-bottom:.0001pt; mso-pagination:widow-orphan; font-size:10.0pt; font-family:"Times New Roman","serif";} The use of logistic regression modeling has exploded during the past decade for prediction and...
Presentation attack detection (PAD) is now considered critically important for any face-recognition (FR) based access-control system. Current deep-learning PAD systems show excellent performance when they are tested in intra-dataset scenarios. Under cross-dataset evaluation the of these drops significantly. This lack generalization attributed to domain-shift. Here, we propose a novel method that leverages large variability present FR datasets induce invariance factors cause Evaluation...
This paper presents the Swiss Federal Institute of Technology (EPFL) proposal to MPEG-4 video coding standardization activity. The proposed technique is based on a novel approach audio-visual data compression entitled dynamic coding. newly born multimedia environment supports plethora applications which cannot be covered adequately by single technique. Dynamic offers opportunity combine several techniques and segmentation strategies. Given particular application, these two degrees freedom...
In this paper, we present two filters that simulate the behavior of biological end-stopped cells. Both are zero-mean filters, and well located in spatial as frequency domains, is, these admissible wavelets. We refer to ES1 ES2. The filter responds ends linear structures which have a specific orientation, ES2 line- segments length within range. show sample results demonstrate proposed wavelets, also discuss scale-space wavelets briefly.
A region merging technique for spatio-temporal segmentation of scenes is presented. The proposed a bottom-up method and expects an initial set regions. These regions are compared on the basis similarity measure that integrates both spatial temporal information. unsupervised procedure based weighted, directed graph updated dynamically. Two clustering rules used to cluster into ensembles represent meaningful objects present in scene. Experimental results demonstrate efficiency method.
The influence of explanatory variables in a logistic model is considered when the goal to determine predictive probability future observation Bayesian approach. We consider whether distribution will change greatly subsets are omitted from data set or observation.