• Make the Bag Disappear: Carrying Status-invariant Gait-based Human Age Estimation using Parallel Generative Adversarial Networks.
  Xiang Li; Yasushi Makihara; Chi Xu; Prof. Yasushi Yagi; Mingwu Ren


• Cosmetic-Aware Makeup Cleanser
  Yi Li; Huaibo Huang; Junchi Yu; Ran He; Tieniu Tan


• A Genetic Algorithm Enabled Similarity-Based Attack on Cancellable Biometrics.
  Xingbo Dong; Zhe Jin; Andrew Beng Jin Teoh


• Unconstrained Thermal Hand Segmentation.
  Ewelina Bartuzi; Mateusz M Trokielewicz


• An End-to-End Convolutional Neural Network for ECG-Based Biometric Authentication.
  João R Pinto; Jaime Cardoso


• The Effect of Broad and Specific Demographic Homogeneity on the Imposter Distributions and False Match Rates in Face Recognition Algorithm Performance.
  John J. Howard; Yevgeniy Sirotin; Arun Vermury


• Hybrid Dictionary Learning and Matching for Video-based Face Verification.
  Jingxiao Zheng; Jun-Cheng Chen; Vishal Patel; Carlos Castillo; Rama Chellappa


• Reliable Age and Gender Estimation from Face Images: Stating the Confidence of Model Predictions.
  Philipp Terhörst; Marco Huber; Jan Kolf; Ines Zelch; Naser Damer; Florian Kirchbuchner; Arjan Kuijper


• Gaze-angle Impact on Iris Segmentation using CNNs.
  Ehsaneddin Jalilian; Andreas Uhl; Mahmut Karakaya


• Rotation Invariant Finger Vein Recognition.
  Bernhard Prommegger; Andreas Uhl


• User profiling using sequential mining over web elements.
  Matan Levi; Itay Hazan


• Palmprint Recognition Using Realistic Animation Aided Data Augmentation.
  Pranjal Swarup; Wai-Kin Adams Kong


• Effects of Postmortem Decomposition on Face Recognition.
  David Bolme; David Cornett; Dawnie Steadman; Kelly Sauerwein; Tiffany Saul


• How to Save Your Face: a Facial Recognition Method Robust Against Image Reconstruction.
  Marcin Plata; Piotr Syga; Marek Klonowski


• Finger-Vein Template Protection based on Alignment-Free Hashing.
  Simon Kirchgasser; Zhe Jin; Yenlung Lai; Andreas Uhl


• Perception of Image Features in Post-Mortem Iris Recognition: Humans vs Machines.
  Adam Czajka; Mateusz M Trokielewicz; Piotr Maciejewicz


• Cross-sensor iris recognition using adversarial strategy and sensor-specific information.
  Jianze Wei; Yunlong Wang; Xiang Wu; Zhaofeng He; Ran He; Zhenan Sun


• Deep Learning-Based Feature Extraction in Iris Recognition: Use Existing Models, Fine-tune or Train From Scratch?.
  Aidan Boyd; Adam Czajka; Kevin Bowyer


• Securing CNN Model and Face Template using Blockchain.
  Akhil Goel; Akshay Agarwal; Mayank Vatsa; Richa Singh; Nalini Ratha


• LC-DECAL: Label Consistent Deep Collaborative Learning for Face Recognition.
  Lamha Goel; Mayank Vatsa; Richa Singh


• Identity-Aware Deep Face Hallucination via Adversarial Face Verification.
  Hadi Kazemi; Fariborz Taherkhani; Nasser Nasrabadi


• A-LINK: Recognizing Disguised Faces via Active Learning based Inter-Domain Knowledge.
  Anshuman Suri; Mayank Vatsa; Richa Singh


• ThirdEye: Triplet Based Iris Recognition without Normalization.
  Sohaib Ahmad; Benjamin Fuller


• MobiFace: A Lightweight Deep Learning Face Recognition on Mobile Devices.
  Chi Nhan Duong; Kha Gia Quach; Ibsa K Jalata; Ngan Le; Khoa Luu


• Realistic Dreams: Cascaded Enhancement of GAN-generated Images with an Example in Face Morphing Attacks.
  Naser Damer; Fadi Boutros; Alexandra Moseguí Saladié; Florian Kirchbuchner; Arjan Kuijper


• Robust Subject-invariant Feature Learning for Ocular Biometrics in Visible Spectrum.
  Sai Narsi Reddy Donthi Reddy; Ajita Rattani; Prof. Reza Derakhshani


• Multi-task Learning For Detecting and Segmenting Manipulated Facial Images and Videos.
  Huy Nguyen; Fuming Fang; Junichi Yamagishi; Isao Echizen


• Attribute-Guided Coupled GAN for Cross-Resolution Face Recognition.
  Veeru Talreja; Fariborz Taherkhani; Nasser Nasrabadi; Matthew Valenti


• Zero-Shot Deep Hashing and Neural Network Based Error Correction for Face Template Protection.
  Veeru Talreja; Matthew Valenti; Nasser Nasrabadi


• Smartphone Camera De-identification while Preserving Biometric Utility.
  Sudipta Banerjee; Arun Ross


• Face Phylogeny Tree: Deducing Relationships Between Near Duplicate Face Images Using Legendre Polynomials and Radial Basis Functions.
  Sudipta Banerjee; Arun Ross


• Facial Attribute Classification: A Comprehensive Study and a Novel Mid-Level Fusion Classifier.
  Abdulaziz A Alorf; A Lynn Abbott; Kadir Peker


• MasterPrint Attack Resistance: A Maximum Cover based Approach for Automatic Fingerprint Template Selection.
  Aditi Roy; Nasir Memon; Arun Ross


• Hierarchical Bloom Filter Framework for Security, Space-efficiency, and Rapid Query Handling in Biometric Systems.
  Sumaiya Shomaji; Fatemah Ganji; Damon L Woodard; Domenic Forte


• Subclass Contrastive Loss for Injured Face Recognition.
  Puspita Majumdar; Saheb Chhabra; Richa Singh; Mayank Vatsa


• Defending Against Adversarial Iris Examples Using Wavelet Decomposition.
  Sobhan Soleymani; Ali Dabouei; Jeremy Dawson; Nasser Nasrabadi


• A Locality Sensitive Hashing Based Approach for Generating Cancelable Fingerprints Templates.
  Debanjan Sadhya; Zahid Akhtar; Dipankar Dasgupta


• On Learning Joint Multi-biometric Representations by Deep Fusion.
  Naser Damer ; Kristiyan Dimitrov ; Andreas Braun; Arjan Kuijper


• IARPA Janus Benchmark Multi-Domain Face.
  Nathan D Kalka; James A Duncan; Jeremy Dawson; Charles Otto


• Morton Filters for Iris Template Protection - An Incremental and Superior Approach Over Bloom Filters.
  Kiran Raja; Raghavendra Ramachandra; Christoph Busch

Robust Tattoo Detection and Retrieval Competition (RTDRC 2019)

Organizers: Prof. Shiguang Shan, A/Prof. Hu Han, Dr Abhijit Das, Dr Antitza Dantcheva

Despite the enormous progress in biometrics-based primary modalities such as the face, iris and fingerprint, unimodal biometrics identification has not been accepted in forensics [5, 6]. Tattoos, which constitute a pertinent and highly distinctive soft biometric trait, can be particularly useful in describing wanted or missing people, or even unidentified bodies. Hence, tattoos are highly instrumental in person identification. Consequently, research on tattoo-based biometrics has gained significant interest in the last few years. In order to explore the potential of tattoos, various research directions have been proposed in the literature. However, existing tattoo search methods or tattoo retrieval techniques mainly focus on the matching of cropped tattoos. Therefore, these topics require more analysis. Open research problems include tattoo detection, as well as localization, i.e., determining whether an image contains a tattoo and if so, segmentation of the tattoo.

In addition, Sketch-Based Tattoo Search can be instrumental in many scenarios, e.g., if the surveillance image of a crime scene is not available, and the query is a tattoo sketch, drawn based a witness description. Therefore, it is important to evaluate the matching performance in similar scenarios.

We note that tattoo detection retrieval techniques can get highly affected by the change in sample quality, acquisition technique, etc. Motivated by the past competition on tattoo biometrics, namely Tatt-C and Tatt-E, and to further advance associated research, we host this competition focusing on the robustness of the evaluation pertained to tattoo detection and retrieval methodologies involving cross-dataset evaluation. The competition will focus on the following three tasks: (i) tattoo detection, (ii) tattoo retrieval, and (iii) tattoo sketch-based retrieval

An Improved Approach to Preprocessing for Fingerprint Recognition Algorithms

Demonstrators: Ashok Patel, Meghna Patel, Satyam Parikh

A demonstration of an experimental algorithm to enhance some of the critical phases of preprocessing to removing the noise, and make clear the fingerprint image for feature extraction. The algorithm also achieves enhancements in the post-processing phases for eliminating falsely extracted minutiae, to extract exact core point detection, and improving the matching of valid minutiae. The algorithm has been tested using FVC2000 and FingerDOS databases for measuring the average FMR = 1% and FNMR = 1.43% and accuracy 98.7% for both databases. There are several reasons, like displacement of finger during scanning, environmental conditions, behavior of user, etc., that cause a reduction in valid fingerprint minutiae and consequently acceptance rate during fingerprint recognition. However, the result and accuracy of fingerprint recognition depends on the presence of valid minutiae.

Spotlight Orals for Doctoral Consortium

Doctoral Consortium (Poster Only)

Tutorial Schedule:

• 9:00 – 12:30 Tutorial I: Recent Advances in Heterogeneous Face Recognition (HFR): Infrared-to-Visible Matching
Speakers: Cunjian Chen (Michigan State University), Shuowen (Sean) Hu (U.S. Army Research Laboratory, Ben Riggan (University of Nebraska-Lincoln), Nathaniel J. Short, (Booz Allen Hamilton), and Vishal M. Patel (Johns Hopkins University)


• 9:00 – 12:30 Tutorial II: Estimation of Soft-biometrics from fingerprints
Speakers: Emanuela Marasco (George Mason University) and Larry Tang (George Mason University)


• 2:00 - 5:30 PM Tutorial III: Face Anti-Spoofing: Past, Present and the Future
Speakers: Yaojie Liu (Michigan State University) and Xiaoming Liu (Michigan State University)

Title: Recent Advances in Heterogeneous Face Recognition (HFR): Infrared-to-Visible Matching

Speakers: Cunjian Chen (Michigan State University), Shuowen (Sean) Hu (U.S. Army Research Laboratory, Ben Riggan (University of Nebraska-Lincoln), Nathaniel J. Short, (Booz Allen Hamilton), and Vishal M. Patel (Johns Hopkins University)

Abstract: This tutorial will address the recent advancements for the emerging research area of heterogeneous face recognition (HFR) from mainly an academic perspective but with insights from government and industry, which seeks to match facial probe imagery acquired in one modality against a gallery database of facial images acquired in another modality. HFR has strong potential to provide new capabilities for law enforcement, the military, and the intelligence community. In this tutorial, we will focus on infrared-to-visible face recognition for low- light and nighttime applications, discussing feature extraction techniques, regression methods, and classification algorithms for matching infrared imagery to gallery databases containing only visible imagery. We will also present recent advances in exploiting sensor technology such as polarimetric imaging for HFR and discuss new algorithms such as generative adversarial network-based approaches for HFR.

Title: Estimation of Soft Biometrics from Fingerprints

Speakers: Emanuela Marasco (George Mason University) and Larry Tang (George Mason University)

Abstract: Accurate gender prediction brings benefits to several applications such as content-based indexing, security, forensics and intelligence applications. A gender recognizer can be combined with the output of primary identifiers to increase the recognition accuracy in challenging scenarios (e.g., partial evidence). In crime investigations, gender classification may minimize the list of suspects. Although the development of reliable gender estimators is needed, existing approaches involving traditional ridge-based fingerprint data are only 80% accurate, and often the processes are not fully automated. Research studies have shown that females exhibit a higher ridge density compared to males, due to finer epidermal ridge details. Thus, the local texture of a fingerprint is expected to offer gender cues because it can encode the ridge density structure that varies between males and females. Local textural descriptors such as Local Binary Patterns (LBP), Local Phase Quantization (LPQ) and Binarized Statistical Image Features (BSIF) have been found to be successful for this task in both high quality and degraded fingerprint images. Differences between males and females were also captured in the frequency domain through Fourier analysis and Discrete Wavelet Transform. Recent studies have explored gender signature on a small and well-defined partition of a fingerprint corresponding to the single minutiae. This methodology is able to deal well with partial fingerprints. Furthermore, given that biometric data is gathered from different sensors, analyzing the sensitivity of the feature set to device changes becomes important. In this regard, local textural descriptors have shown robustness with respect to capture bias. Additionally, logistic regression models were applied to identify the significant features for gender estimation. These models exploit quality-based NFIQ2 features as well as local texture.
This tutorial aims at bringing awareness about the ongoing research in Gender estimation from fingerprints and the challenges faced in improving the prediction accuracy. We will review the various experiments carried out, the several features extracted from fingerprints, classification models used and their efficiency in classifying gender.

Title: Face Anti-Spoofing: Past, Present and the Future

Speakers: Yaojie Liu (Michigan State University) and Xiaoming Liu (Michigan State University)

Abstract: Face is one of the most popular biometric modalities due to its convenience of usage in access control, phone unlock and etc. Despite the high recognition accuracy, face recognition systems are not designed to distinguish between real human faces and fake ones, e.g., photograph, screen. Face spoof attacks, or presentation attacks, are the real-world attacks that use those fake faces to deceive the systems to recognize them as the real live person. Thus, to safely utilizing face recognition systems, face anti-spoofing techniques are crucial in detecting spoof attacks before performing recognition.
This tutorial provides a comprehensive review of the development of face anti-spoofing technologies. We focus on solutions for RGB sensors and include discussions of various spoof attacks (e.g. print attack, replay attack, and 3D mask attack), existing databases (e.g CASIAMFSD, OULU-NPU, HKBU MAR, and SiW), and representative works (e.g. conventional approaches and deep learning-based approaches). We also plan to discuss hardware solutions (e.g. NIR, and depth camera), generalizability to unknown attacks, several practical tips of building a face anti-spoofing system, and future research directions for face anti-spoofing.

Abstract:

Over the last half a century, biometrics provided leadership in many areas of pattern recognition and machine learning through many practical challenges while AI was going through its winter hibernation. Unaware of the challenging problems in biometrics use in large scale deployments and its grave societal impact, biometrics was considered as a solved problem by research community in mid-eighties to early nineties. In this talk, we will show how every aspect of modern AI including detection, feature extraction, matching, classification, extreme-scale classification, security, adversarial attacks, privacy, bias/fairness, ethics, explainability has been widely researched in the biometrics community much in advance before they became burning problems in AI. Many of the problems were first encountered and addressed in biometrics, e.g., scale of classification (starting with PCASYS), identification, detection (face), adoption difficulties in real world, privacy/security, datasets, standardization, enhancement, to just name a few. Successfully overcoming many operational difficulties and teething user acceptance issues, biometrics has matured into handling billions of users through mobile phones and large scale government programs. This talk will review some of the component technologies that enabled the phenomenal success of biometrics recently seen at an unprecedented level as well highlight how some of the revolutionary technologies in terms of blockchain and advanced security techniques like fully homomorphic encryption can bring biometrics even closer to masses. The rest of the machine learning and AI community will benefit as history will repeat again with the biometrics beacon showing the path for AI/ML to move forward.

Dr. Nalini K. Ratha is a Research Staff Member at IBM Thomas J. Watson Research Center, Yorktown Heights, NY. He received his B. Tech. in Electrical Engineering from Indian Institute of Technology, Kanpur, M.Tech. degree in Computer Science and Engineering also from Indian Institute of Technology, Kanpur and Ph. D. in Computer Science from Michigan State University. He has authored more than 100 research papers in the area of biometrics and has been co-chair of several leading biometrics conferences and served on the editorial boards of IEEE Trans. on PAMI, IEEE Trans. on SMC-B, IEEE Trans. on Image Processing and Pattern Recognition journal. He has co-authored a popular book on biometrics entitled “Guide to Biometrics” and also co-edited two books entitled “Automatic Fingerprint Recognition Systems” and “Advances in Biometrics: Sensors, Algorithms and Systems”. He has offered tutorials on biometrics technology at leading IEEE conferences and also teaches courses on biometrics and security. He is Fellow of IEEE, Fellow of IAPR and an ACM Distinguished Scientist. His research interests include biometrics, pattern recognition and computer vision. He has been an adjunct professor at IIIT Delhi, Cooper Union and NYU-Poly. During 2011-2012 he was the president of the IEEE Biometrics Council. At IBM, he has received several awards including a Research Division Award, Outstanding Innovation Award and Outstanding Technical Accomplishment Award along with several patent achievement awards. Recently he has been designated as an IBM Research Master Inventor.

A. Advice for Oral Presenters

Please lookup at the program to find out which day is your presentation scheduled. Your oral presentation should last: 15 mins if you presenting in sessions 1-3 and 18 mins if you are presenting in the “Best Reviewed Papers” Session. The audience will have 2 mins to ask you questions. In addition, you are kindly requested to prepare a poster to present at the poster session that day. Please find below some advice:

Oral Presentation

1. Your session chair will communicate with you to ask
• Who is the presenter
• Learn one thing about the presenter to use for the introduction
2. Please think about how to make your talk engaging and elucidating.
3. Give the audience something they can walk away with: valuable insights, actionable information, and fresh perspectives.
4. Make sure that your review of the related work points out clearly what are the gaps in the literature and how your paper fits in the literature (meaning which gap of these gaps does it eliminate)
5. Make sure that you clearly state what the contributions of this paper are and what are the advantages of your solution.
6. Make sure you state what is the impact of your work (basically who cares).
7. State the limitations of your work and offer a discussion and a reflection on the results of your method/system.
8. Make sure that you state why you compare with specific algorithms and why did you select the databases that you provide results.
9. Mention if your improvement is statistically significant or not.
10. If you processed data under an IRB protocol add the following sentences: “The data were acquired using the IRB protocol titled “<IRB Title>” approved by <Name of your institution> or “The data were processed using the IRB protocol titled “<IRB Title>” approved by <Name of your institution>.
11. Practice, Practice, Practice!
12. Finish on time!

Poster Presentation

1. Prepare a 30 s presentation that presents the overview of your work to the attendee that comes to your poster
2. Prepare a 3 min presentation to share with those that would like to know the details
3. The poster boards will be 8 ft wide and 4 ft tall.

B. Advice for Poster Presenters

Please lookup at the program to find out which day is your presentation scheduled. Along with your poster presentation, you are kindly asked to offer a 5 min presentation to entice the audience to come to your poster. Your oral presentation should last for 5 mins. The audience will have 1 min to ask you questions. Please find below some advice:

Oral Presentation

1. Your session chair will communicate with you to ask
• Who is the presenter
• Learn one thing about the presenter to use for the introduction
2. Please think about how to make your talk engaging and elucidating. You are to provide the reason an attendee should come to your poster.
3. Practice, Practice, Practice!
4. Finish on time!

Poster Presentation

1. Give the audience something they can walk away with: valuable insights, actionable information, and fresh perspectives.
2. Make sure that your review of the related work points out clearly what are the gaps in the literature and how your paper fits in the literature (meaning which gap of these gaps does it eliminate)
3. Make sure that you clearly state what the contributions of this paper are and what are the advantages of your solution.
4. Make sure you state what the impact of your work (basically who cares) is
5. State the limitations of your work and offer a discussion and a reflection on the results of your method/system.
6. Make sure that you state why you compare with specific algorithms and why did you select the databases that you provide results.
7. Mention if your improvement is statistically significant or not.
8. If you processed data under an IRB protocol add the following sentences: “The data were acquired using the IRB protocol titled “<IRB Title>” approved by <Name of your institution> or “The data were processed using the IRB protocol titled “<IRB Title>” approved by <Name of your institution>.
9. Prepare a 30 s presentation that presents the overview of your work to the attendee that comes to your poster
10. Prepare a 3 min presentation to share with those that would like to know the details
11. The poster boards will be 8 ft wide and 4 ft tall.

Best Posters:

Defending Against Adversarial Iris Examples Using Wavelet Decomposition, Sobhan Soleymani; Ali Dabouei; Jeremy Dawson; Nasser Nasrabadi, West Virginia University
Face Phylogeny Tree: Deducing Relationships Between Near-Duplicate Face Images Using Legendre Polynomials and Radial Basis Functions, Sudipta Banerjee and Arun Ross, Michigan State University

Best Student Paper:

Subclass Contrastive Loss for Injured Face Recognition, Puspita Majumdar, Saheb Chhabra, Richa Singh, Mayank Vatsa, IIIT-Delhi, India

Best Paper:

Face Phylogeny Tree: Deducing Relationships Between Near-Duplicate Face Images Using Legendre Polynomials and Radial Basis Functions. Sudipta Banerjee and Arun Ross, Michigan State University