Naman Kohli

Self-similarity representation of Weber faces for kinship classification

Establishing kinship using images can be utilized as context information in different applications including face recognition. However, the process of automatically detecting kinship in facial images is a challenging and relatively less explored task. The reason for this includes limited availability of datasets as well as the inherent variations amongst kins. This paper presents a kinship classification algorithm that uses the local description of the pre-processed Weber face image. A kinship database is also prepared that contains images pertaining to 272 kin pairs. The database includes images of celebrities (and their kins) and has four ethnicity groups and seven kinship groups. The proposed algorithm outperforms an existing algorithm and yields a classification accuracy of 75.2%.

Unraveling the Effect of Textured Contact Lenses on Iris Recognition

The presence of a contact lens, particularly a textured cosmetic lens, poses a challenge to iris recognition as it obfuscates the natural iris patterns. The main contribution of this paper is to present an in-depth analysis of the effect of contact lenses on iris recognition. Two databases, namely, the IIIT-D Iris Contact Lens database and the ND-Contact Lens database, are prepared to analyze the variations caused due to contact lenses. We also present a novel lens detection algorithm that can be used to reduce the effect of contact lenses. The proposed approach outperforms other lens detection algorithms on the two databases and shows improved iris recognition performance.

Revisiting iris recognition with color cosmetic contact lenses

Over the years, iris recognition has gained importance in the biometrics applications and is being used in several large scale nationwide projects. Though iris patterns are unique, they may be affected by external factors such as illumination, camera-eye angle, and sensor interoperability. The presence of contact lens, particularly color cosmetic lens, may also pose a challenge to iris biometrics as it obfuscates the iris patterns and changes the inter and intra-class distributions. This paper presents an in-depth analysis of the effect of contact lens on iris recognition performance. We also present the IIIT-D Contact Lens Iris database with over 6500 images pertaining to 101 subjects. For each subject, images are captured without lens, transparent (prescription) lens, and color cosmetic lens (textured) using two different iris sensors. The results computed using VeriEye suggest that color cosmetic lens significantly increases the false rejection at a fixed false acceptance rate. Also, the experiments on four existing lens detection algorithms suggest that incorporating lens detection helps in maintaining the iris recognition performance. However further research is required to build sophisticated lens detection algorithm that can improve iris recognition.

Hierarchical Representation Learning for Kinship Verification

Kinship verification has a number of applications such as organizing large collections of images and recognizing resemblances among humans. In this research, first, a human study is conducted to understand the capabilities of human mind and to identify the discriminatory areas of a face that facilitate kinship-cues. The visual stimuli presented to the participants determines their ability to recognize kin relationship using the whole face as well as specific facial regions. The effect of participant gender and age and kin-relation pair of the stimulus is analyzed using quantitative measures such as accuracy, discriminability index d1, and perceptual information entropy. Utilizing the information obtained from the human study, a hierarchical Kinship Verification via Representation Learning (KVRL) framework is utilized to learn the representation of different face regions in an unsupervised manner. We propose a novel approach for feature representation termed as filtered contractive deep belief networks (fcDBN). The proposed feature representation encodes relational information present in images using filters and contractive regularization penalty. A compact representation of facial images of kin is extracted as an output from the learned model and a multi-layer neural network is utilized to verify the kin accurately. A new WVU Kinship Database is created which consists of multiple images per subject to facilitate kinship verification. The results show that the proposed deep learning framework (KVRL-fcDBN) yields stateof- the-art kinship verification accuracy on the WVU Kinship database and on four existing benchmark datasets. Further, kinship information is used as a soft biometric modality to boost the performance of face verification via product of likelihood ratio and support vector machine based approaches. Using the proposed KVRL-fcDBN framework, an improvement of over 20% is observed in the performance of face verification.

Effect of illicit drug abuse on face recognition

Over the years, significant research has been undertaken to improve the performance of face recognition in the presence of covariates such as variations in pose, illumination, expressions, aging, and use of disguises. This paper highlights the effect of illicit drug abuse on facial features. An Illicit Drug Abuse Face (IDAF) database of 105 subjects has been created to study the performance on two commercial face recognition systems and popular face recognition algorithms. The experimental results show the decreased performance of current face recognition algorithms on drug abuse face images. This paper also proposes projective Dictionary learning based illicit Drug Abuse face Classification (DDAC) framework to effectively detect and separate faces affected by drug abuse from normal faces. This important pre-processing step stimulates researchers to develop a new class of face recognition algorithms specifically designed to improve the face recognition performance on faces affected by drug abuse. The highest classification accuracy of 88.81% is observed to detect such faces by the proposed DDAC framework on a combined database of illicit drug abuse and regular faces.

Multiple Projective Dictionary Learning to Detect Plastic Surgery for Face Verification

Researchers have shown that the changes in face features due to plastic surgery can be modeled as a covariate that reduces the ability of algorithms to recognize a persons identity. Traditional dictionary learning methods learn a sparse representation using I0 and I1 norms that are computationally expensive. This paper presents a multiple projective dictionary learning (MPDL) framework that does not require the computation of I0 and I1 norms. We propose a novel solution to discriminate plastic surgery faces from regular faces by learning representations of local and global plastic surgery faces using multiple projective dictionaries and by using compact binary face descriptors. Experimental results on the plastic surgery database show that the proposed MPDL framework is able to detect plastic surgery faces with a high accuracy of 97.96%. To verify the identity of a person, the detected plastic surgery faces are divided into local regions of interest (ROIs) that are likely to be altered by a particular plastic surgery. The cosine distance between the compact binary face descriptors is computed for each ROI in the detected plastic surgery faces. In addition, we compute the human visual system feature similarity score based on phase congruency and gradient magnitude between the same ROIs. The cosine distance scores and the feature similarity scores are combined to learn a support vector machine model to verify if the faces belong to the same person. We integrate our proposed MPDL framework for face verification with two commercial systems to demonstrate an improvement in verification performance on a combined database of plastic surgery and regular face images.

Detecting medley of iris spoofing attacks using DESIST

Human iris is considered a reliable and accurate modality for biometric recognition due to its unique texture information. However, similar to other biometric modalities, iris recognition systems are also vulnerable to presentation attacks (commonly called spoofing) that attempt to conceal or impersonate identity. Examples of typical iris spoofing attacks are printed iris images, textured contact lenses, and synthetic creation of iris images. It is critical to note that majority of the algorithms proposed in the literature are trained to handle a specific type of spoofing attack. These algorithms usually perform very well on that particular attack. However, in real-world applications, an attacker may perform different spoofing attacks. In such a case, the problem becomes more challenging due to inherent variations in different attacks. In this paper, we focus on a medley of iris spoofing attacks and present a unified framework for detecting such attacks. We propose a novel structural and textural feature based iris spoofing detection framework (DESIST). Multi-order dense Zernike moments are calculated across the iris image which encode variations in structure of the iris image. Local Binary Pattern with Variance (LBPV) is utilized for representing textural changes in a spoofed iris image. The highest classification accuracy of 82.20% is observed by the proposed framework for detecting normal and spoofed iris images on a combined iris spoofing database.

Face Presentation Attack with Latex Masks in Multispectral Videos

Face recognition systems are susceptible to presentation attacks such as printed photo attacks, replay attacks, and 3D mask attacks. These attacks, primarily studied in visible spectrum, aim to obfuscate or impersonate a persons identity. This paper presents a unique multispectral video face database for face presentation attack using latex and paper masks. The proposed Multispectral Latex Mask based Video Face Presentation Attack (MLFP) database contains 1350 videos in visible, near infrared, and thermal spectrums. Since the database consists of videos of subjects without any mask as well as wearing ten different masks, the effect of identity concealment is analyzed in each spectrum using face recognition algorithms. We also present the performance of existing presentation attack detection algorithms on the proposed MLFP database. It is observed that the thermal imaging spectrum is most effective in detecting face presentation attacks.

Region-specific fMRI dictionary for decoding face verification in humans

This paper focuses on decoding the process of face verification in the human brain using fMRI responses. 2400 fMRI responses are collected from different participants while they perform face verification on genuine and imposter stimuli face pairs. The first part of the paper analyzes the responses covering both cognitive and fMRI neuro-imaging results. With an average verification accuracy of 64.79% by human participants, the results of the cognitive analysis depict that the performance of female participants is significantly higher than the male participants with respect to imposter pairs. The results of the neuro-imaging analysis identifies regions of the brain such as the left fusiform gyrus, caudate nucleus, and superior frontal gyrus that are activated when participants perform face verification tasks. The second part of the paper proposes a novel two-level fMRI dictionary learning approach to predict if the stimuli observed is genuine or imposter using the brain activation data for selected regions. A comparative analysis with existing machine learning techniques illustrates that the proposed approach yields at least 4.5% higher classification accuracy than other algorithms. It is envisioned that the result of this study is the first step in designing brain-inspired automatic face verification algorithms.