A5076371381- Face recognition and analysis
- Biometric Identification and Security
- Advanced Image Fusion Techniques
- Remote-Sensing Image Classification
- Face and Expression Recognition
- Multimodal Machine Learning Applications
- Image and Signal Denoising Methods
- Advanced Image and Video Retrieval Techniques
- Natural Language Processing Techniques
- Topic Modeling
- Privacy-Preserving Technologies in Data
- Image Enhancement Techniques
- Imbalanced Data Classification Techniques
- Human Pose and Action Recognition
- User Authentication and Security Systems
- Dermatoglyphics and Human Traits
- Image Retrieval and Classification Techniques
- AI in cancer detection
- Retinal Imaging and Analysis
- Fusion materials and technologies
- Gaze Tracking and Assistive Technology
- Aging and Gerontology Research
- Constraint Satisfaction and Optimization
- Nuclear Materials and Properties
- Artificial Intelligence in Healthcare and Education
Idiap Research Institute
2019-2025
Zimmer Biomet (Switzerland)
2019-2022
Indian Institute of Technology Bombay
2010-2013
In this paper, we present EdgeFace-a lightweight and efficient face recognition network inspired by the hybrid architecture of EdgeNeXt. By effectively combining strengths both CNN Transformer models, a low rank linear layer, EdgeFace achieves excellent performance optimized for edge devices. The proposed not only maintains computational costs compact storage, but also high accuracy, making it suitable deployment on model achieved top ranking among models with fewer than 2M parameters in...
This paper presents a new approach for hyperspectral image visualization. A bilateral filtering-based is presented fusion to generate an appropriate resultant image. The proposed retains even the minor details that exist in individual bands, by exploiting edge-preserving characteristics of filter. It does not introduce visible artifacts fused hierarchical scheme has also been implementation purposes accommodate large number bands. provides computational and storage efficiency without...
This paper presents the summary of Efficient Face Recognition Competition (EFaR) held at 2023 International Joint Conference on Biometrics (IJCB 2023). The competition received 17 submissions from 6 different teams. To drive further development efficient face recognition models, submitted solutions are ranked based a weighted score achieved verification accuracies diverse set benchmarks, as well deployability given by number floating-point operations and model size. evaluation is extended to...
Makeup is a simple and easy instrument that can alter the appearance of person's face, hence, create presentation attack on face recognition (FR) systems. These attacks, especially ones mimicking ageing, are difficult to detect due their close resemblance with genuine (non-makeup) appearances. Makeups also degrade performance systems various algorithms use human as an input. The detection facial makeups effective prohibitory measure minimize these problems. This work proposes deep...
Demographic bias in deep learning-based face recog-nition systems has led to serious concerns. Several ex-isting works attempt mitigate by incorporating demographic-specific processing during inference, which requires knowledge or learning of demographic attribute with an additional cost. We propose regularize training the recognition CNN, for fairness, im-posing constraints on distributions matching scores. Our regularization term enforces score from different groups respect a pre-defined...
Demographic bias in face recognition (FR) has emerged as a critical area of research, given its impact on fairness, equity, and reliability across diverse applications. As FR technologies are increasingly deployed globally, disparities performance demographic groups -- such race, ethnicity, gender have garnered significant attention. These biases not only compromise the credibility systems but also raise ethical concerns, especially when these employed sensitive domains. This review...
With advancements in hardware, high-quality head-mounted display (HMD) devices are being developed by numerous companies, driving increased consumer interest AR, VR, and MR applications. This proliferation of HMD opens up possibilities for a wide range applications beyond entertainment. Most commercially available equipped with internal inward-facing cameras to record the periocular areas. Given nature these captured data, many such as biometric authentication gaze analysis become feasible....
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...
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...
In this paper we propose a new approach for visualization-oriented fusion of hyperspectral image bands. The proposed technique has been devised to generate the fused with certain set desired properties better visualization. should provide resultant high local contrast without driving individual pixels into over- or under-saturation. We focus on these image, and formulate multi-objective cost function same. have shown how can incorporate constraint spatial smoothness weight vectors, as...
The field of Vascular Biometric Recognition has drawn a lot attention recently with the emergence new computer vision techniques. different methods using Deep Learning involve understanding deeper features from vascular network. specific architecture veins needs complex model capable comprehending pattern. In this paper, we present an image enhancement method Convolutional Neural Network. For task, residual convolutional auto-encoder been trained in supervised way to enhance vein patterns...
With advancements in hardware, high-quality HMD devices are being developed by numerous companies, driving increased consumer interest AR, VR, and MR applications. In this work, we present a new dataset, called VRBiom, of periocular videos acquired using Virtual Reality headset. The targeted at biometric applications, consists 900 short from 25 individuals recorded the NIR spectrum. These 10s long have been captured internal tracking cameras Meta Quest Pro 72 FPS. To encompass real-world...
The fusion of hyperspectral images is an important area in research and applications. Several techniques have been developed the literature for visualization hyper-spectral data. amount computation needed such directly related to volume Most these involve a significant due high data, making processes slow. We analyze statistical characteristics this data order develop technique faster fusion. image bands represent response scene collected over contiguous narrow wavelength. adjacent being...
Shantipriya Parida, Idris Abdulmumin, Shamsuddeen Hassan Muhammad, Aneesh Bose, Guneet Singh Kohli, Ibrahim Said Ahmad, Ketan Kotwal, Sayan Deb Sarkar, Ondřej Bojar, Habeebah Kakudi. Findings of the Association for Computational Linguistics: ACL 2023.
Presentation attacks using 3D masks pose a serious threat to face recognition systems. Automatic detection of these is challenging due hyper-realistic nature masks. In this work, we consider presentations acquired in near infrared (NIR) imaging channel for mask-based attacks. We propose patch pooling mechanism learn complex textural features from lower layers convolutional neural network (CNN). The proposed layer can be used conjunction with pretrained CNN without fine-tuning or adaptation....
The accuracy of finger vein recognition systems gets degraded due to low and uneven contrast between veins surroundings, often resulting in poor detection patterns. We propose a finger-vein enhancement technique, ResFPN (Residual Feature Pyramid Network), as generic preprocessing method agnostic the pipeline. A bottom-up pyramidal architecture using novel Structure Detection block (SDBlock) facilitates extraction varied widths. Using feature aggregation module (FAM), we combine these...
Large occlusions result in a significant decline image classification accuracy. During inference, diverse types of unseen introduce out-of-distribution data to the model, leading accuracy dropping as low 50%. As encompass spatially connected regions, conventional methods involving feature reconstruction are inadequate for enhancing performance. We LEARN: Latent Enhancing feAture Reconstruction Network -- An auto-encoder based network that can be incorporated into model before its classifier...
Current finger-vein or palm-vein recognition systems usually require direct contact of the subject with apparatus. This can be problematic in environments where hygiene is primary importance. In this work we present a contactless vascular biometrics sensor platform named \sweet which used for hand studies (wrist-, palm- and finger-vein) surface features such as palmprint. It supports several acquisition modalities multi-spectral Near-Infrared (NIR), RGB-color, Stereo Vision (SV) Photometric...