Large Language Models (LLMs) have the potential to enhance decision-making significantly in core military operational contexts that support training, readiness, and mission execution under low-risk conditions. Still, their implementation must be approached carefully, considering the associated risks. This paper examines the integration of LLMs into military decision-making, emphasizing the LLM’s ability to improve intelligence analysis, enhance situational awareness , support strategic planning, predict threats, optimize logistics, and strengthen cybersecurity . The paper also considers misinterpretation, bias, misinformation , or overreliance on AI-generated suggestions, potentially leading to errors in routine but critical decision-making processes. Our work concludes by proposing solutions and promoting the responsible implementation of LLMs to ensure their effective and ethical use in military operations. To build trust in LLMs, this paper advocates for developing cybersecurity frameworks , transparency, and ethical oversight. It further suggests using machine unlearning (MU) to selectively remove outdated or compromised data from LLM training datasets, preserving the integrity of the insights they generate. The paper underscores the imperative for integrating LLMs in low-risk military contexts, coupled with sustained research efforts to mitigate potential hazards.

A YOLOv10-based measurement software to analyze agricultural vacuum seed meters was developed to be widely utilized by growers to plant peanuts. The software can process videos recorded on a test stand or from a performing planter under field conditions. Initially, images under various light conditions and camera angles were captured and manually annotated to create a dataset. Advanced image augmentation techniques were applied to enhance the dataset, which was utilized to train a YOLOv10 object detection model, achieving an impressive mAP50 score of 99.4%. Paired with a counting algorithm, the model accurately identified singulated peanuts, empty cells (missing peanuts/spots), and multiple transported peanuts (doubles). Thus, the software automatically calculated and exported detection percentages for each category. The data enables the assessment of seed meter performance across different variables, offering valuable insights for optimizing seed plate and singulator design, vacuum pressure, and rotational speed. The software demonstrated a low error rate of 1.91% and ensured reliable performance measurement metrics.

Developing deep learning tools for automated quantification of arbuscular mycorrhizal fungi (AMF) necessitates large, high-quality datasets capturing the diversity of AMF-root interactions under varied environmental conditions. We present MycorrhiSEE, a 15-Terabyte (Tb), high-resolution image dataset comprising approximately 137,500 whole-slide images (WSIs) of root segments from 5,500 sorghum plants representing 337 genotypes, grown under diverse agronomic treatments. Images were acquired using the ZEISS AxioScan 7 high-throughput robotic microscope equipped with a 5-megapixel resolution color camera. This dataset provides unprecedented detail and diversity, offering a valuable resource for developing deep learning applications in AMF research.

This paper focuses on thermal-to-visible face matching through image synthesis. Most of the legacy face image datasets are composed of visible band data. Thermal band as well as dual band, i.e. visible and thermal face datasets, are limited. Operating in the thermal band and therefore working on visible thermal face recognition (FR) systems can be beneficial in various scenarios. The challenge is cross-spectral matching, i.e. matching gallery, visible band, face images against thermal ones. To address this problem, we train and test two of the most popular image-to-image translation Generative Adversarial Networks (GANs). These are Pix2pix and StarGAN2. In this work, the two aforementioned GAN trained models are tested, and the visible face images generated are matched against the ground truth visible faces using one of the most powerful visible-to-visible face matching algorithms, namely Facenet. We also perform an ablation study where the original thermal and visible images are photometrically normalized before training the image synthesis-specific models. The main outcomes of our study are that the FR accuracy from the pix2pix model did not vary significantly; when using StarGAN2, the original face images yield much higher accuracy compared to the photometrically normalized ones; finally, we observe that, when using the pix2pix model for image synthesis, bearded and non-Caucasian generated face images suffer the most from different noise factors. Specifically, the FR accuracy when using pix2pix after image synthesis yields a face verification area under curve (AUC) of 58.3%, while the same models when tested on data excluding bearded and non-Caucasian faces yields an accuracy of 68.6%.

Convolutional neural networks in the task of object detection and localization have been evolving in the last few years. Various convolutional network models have been proposed such as Faster Region-Based, Mask Region-Based, Single Shot Detection, and "You Only Look Once" models (with different versions). Although instance segmentation has been explored with many models, the Mask Region-Based Convolutional Neural Network has been one of the most competitive models in terms of overall object detection performance. Its widespread use in many different applications encouraged us to take a closer look at model performance in a unique object detection task, namely the detection of spacecraft images in the wild. The main research question in this paper, is whether this off-the-shelf proposed architecture can effectively detect and localize a spacecraft when using the Spacecraft Pose Estimation Dataset, under a variety of different image degradation factors and at various degradation levels. The inspiration for this investigation is the effect of deep space environments on charge-coupled device image sensors, and other imaging hardware. The capability of detection and localization to continue in the face of pixel loss and Gaussian noise are explored. The effects of training augmentation on object detection performance is another task that has also been studied. Some of our main findings include that supplementing training on degraded images improve significantly the detection results. In low degradation scenarios, the improvement is better than the baseline results. Also, the proposed models is able to detect and localize properly on spacecraft images degraded by pixel loss. The proposed model continues to perform close to baseline in conditions even up to 80% pixel loss for the black background experiments.

This paper proposes a novel approach to automatically detect the eye centers in challenging thermal (Long-Wave Infrared; 8 – 14 µm) face images so they can be geometrically normalized, which is an important part of face recognition systems. Developing a new LWIR based eye center detection model would require many thermal images, which is not possible due to the unavailability of publicly available large scale LWIR face datasets. An alternative solution would be to use a pre-trained, visible band based facial landmark detection model and test it on LWIR face images. In the latter case, the challenges are the significant differences between the visible and thermal face images. In this paper, we focus on addressing this research gap by proposing a solution that is based on the following approach. First, we synthesize visible from thermal face images. Then, we exploit the existing robust visible band facial landmark detection models to detect eye centers in the synthesized visible band face images. While we empirically test different image synthesis models, we determine that StarGAN2 (an image-to-image translation generative adversarial network model that learns a mapping between the different visual domains) yields the highest eye center detection accuracy when compared to the other state-of-the-art models. Thus, we train a StarGAN2 model to be able to synthesize good quality visible band images from their thermal band counterparts. Next, we use an efficient visible band based facial landmark detection model to detect the eye centers in the synthesized visible band face images. Finally, we map these coordinates to the original LWIR face images, which are used for geometric normalization and, finally run a set of face recognition experiments. Compared to the baseline model, our approach increases the eye detection accuracy by 14%, 50%, and 30% when the normalized error (e) is set to be ≤ 0.05, ≤ 0.10, and ≤ 0.25 respectively compared to the baseline model. Our approach yields up to a 15 % and 7% increase in face recognition accuracy when using advanced deep-learning based matchers, namely Facenet and VGGFace respectively, and by 30% when using other conventional face matching techniques such as LBP-LTP matchers.

In recent years, visible-spectrum face verification systems have been shown to match the performance of experienced forensic examiners. However, such systems are ineffective in low-light and nighttime conditions. Thermal face imagery, which captures body heat emissions, effectively augments the visible spectrum, capturing discriminative facial features in scenes with limited illumination. Due to the increased cost and difficulty of obtaining diverse, paired thermal and visible spectrum datasets, not many algorithms and large-scale benchmarks for low-light recognition are available. This paper presents an algorithm that achieves state-of-the-art performance on both the ARL-VTF and TUFTS multi-spectral face datasets. Importantly, we study the impact of face alignment, pixel-level correspondence, and identity classification with label smoothing for multi-spectral face synthesis and verification. We show that our proposed method is widely applicable, robust, and highly effective. In addition, we show that the proposed method significantly outperforms face frontalization methods on profile-to-frontal verification. Finally, we present MILAB-VTF(B), a challenging multi-spectral face dataset that is composed of paired thermal and visible videos. To the best of our knowledge, with face data from 400 subjects, this dataset represents the most extensive collection of publicly available indoor and long-range outdoor thermal-visible face imagery. Lastly, we show that our end-to-end thermal-to-visible face verification system provides strong performance on the MILAB-VTF(B) dataset.

Liveness Detection (LivDet)-Face is an international competition series open to academia and industry. The competition’s objective is to assess and report state-of-the-art in liveness / Presentation Attack Detection (PAD) for face recognition. Impersonation and presentation of false samples to the sensors can be classified as presentation attacks and the ability for the sensors to detect such attempts is known as PAD. LivDet-Face 2021 * will be the first edition of the face liveness competition. This competition serves as an important benchmark in face presentation attack detection, offering (a) an independent assessment of the current state of the art in face PAD, and (b) a common evaluation protocol, availability of Presentation Attack Instruments (PAI) and live face image dataset through the Biometric Evaluation and Testing (BEAT) platform. The competition can be easily followed by researchers after it is closed, in a platform in which participants can compare their solutions against the LivDet-Face winners.

An efficient iris sensor identification algorithm can be used in certain forensic applications, i.e. detecting mislabeled iris data at large scale iris datasets, and verifying the validity of the data origin of collected iris datasets that are available to be shared. Such knowledge can potentially increase the overall iris recognition system accuracy by offering the operator the option to match same-sensor or cross-sensor iris images. In either case the knowledge of the origin of the sensor used to collect these data, when not available, or the correction of mislabeled data, is expected to result in higher iris matching accuracy. Another benefit of iris sensor identification is that it can assist in improving the detection of fake iris data, i.e. when knowing the iris sensor, we can apply more appropriate models for fake detection that are tuned for a specific iris sensor. In this paper we propose an efficient deep learning-based iris recognition algorithm that is sensor inter-operable. Our approach utilizes a moderate amount of data and is adaptable to learning rate variations as well as variations of the amount of data used for training per class. Our proposed approach uses a set of iris datasets that include iris images captured at different standoff distances. We are using the original captured, dual eye, or periocular images rather than the iris itself, after detecting, segmenting, and normalizing the iris. Thus, the algorithm is efficient, fast, and less depended on additional algorithmic processes that can add computational complexity. Our proposed process includes transfer learning using iris images of higher quality via the utilization of a set of image quality metrics and achieves close to a hundred percent accuracy after cross-validation.

In this paper, we propose a Faster R-CNN based model for detection of firearms in surveillance and Closed-Circuit Television (CCTV) video. Due to the lack of a publicly available benchmark surveillance video database containing firearms, we created our own firearm database composed of only real surveillance footage from scenes with people holding firearms. Portable firearms such as rifles and pistols, often referred to as small arms, have a high degree of variation in surveillance footage in terms of shape, illumination, scale, pose, and occlusion. To deploy a fast and accurate firearm detection system for real world scenarios, it is important to detect all types of common small arms. To accomplish this, we collected images for our database that contain both handguns (pistols and revolvers) and long guns (rifles and shotguns) to capture as much variation of commonly used small arms as possible. To train our firearm detector we first, assess 18 common geometric and photometric data augmentation techniques in order to identify which ones improve detection performance. Next, we identify deep learning based small arm detection techniques that improve detection performance. We reuse these techniques in another more comprehensive augmentation assessment in order to identify the most efficient combination in terms of detection performance. Finally, we optimize the hyper-parameters of all the tested firearm detection models and apply cross-validation. Our proposed model can accurately detect firearms in video frames taken from real surveillance footage close to real-time, yielding precision and recall scores of 93.9% and 96.4% for handguns, and 95.2% and 94.6% for long guns.

Advances in iris recognition discuss the impact of pupil size variations on iris matching accuracy, thus promoting the need to model these covariates that are present in the near infrared (NIR) spectrum to measure their level of authenticity. This work incorporates these principles to propose a novel methodology that automatically distinguishes subject-specific variations of light-adapted pupil size behavior in iris video sequences consisting of two main steps. In the first step, the sinuous nature of the light-adapted pupil size is characterized, which depends on the dilation extrema, the dilative rate of change, and midpoint behavior. In the second and final step, these aspects are fed into a classification framework that distinguishes subject-specific light-adapted pupil size responses. Experimental results, when using the West Virginia University Pupillary Light Reflex Ramp (WVU-PLR Ramp) dataset, convey the efficacy of this approach with classification accuracies ranging from 92% - 100% when support vector machines (SVMs) are used and 100% when neural networks (NNs) are applied. The results of this work can be used to effectively describe light-adapted variations in pupil size. Additionally, these results indicate that the changes in light-adapted pupil size can potentially be a soft iris biometric trait.

Despite the apparent benefits of modern social coding paradigm such as Stack Overflow, its potential security risks have been largely overlooked (e.g., insecure codes could be easily embedded and distributed). To address this imminent issue, in this paper, we bring a significant insight to leverage both social coding properties and code content for automatic detection of insecure code snippets in Stack Overflow. To determine if the given code snippets are insecure, we not only analyze the code content, but also utilize various kinds of relations among users, badges, questions, answers and code snippets in Stack Overflow. To model the rich semantic relationships, we first introduce a structured heterogeneous information network (HIN) for representation and then use meta-path based approach to incorporate higher-level semantics to build up relatedness over code snippets. Later, we propose a novel hierarchical attention-based sequence learning model named CodeHin2Vec to seamlessly integrate node (i.e., code snippet) content with HIN-based relations for representation learning. After that, a classifier is built for insecure code snippet detection. Integrating our proposed method, an intelligent system named iTrustSO is accordingly developed to address the code security issues in modern software coding platforms. Comprehensive experiments on the data collections from Stack Overflow are conducted to validate the effectiveness of our developed system iTrustSO by comparisons with alternative methods.

In this work, we propose an optimal solution for face detection when operating in the thermal and visible bands. Our aim is to train, fine tune, optimize and validate preexisting object detection models using thermal and visible data separately. Thus, we perform an empirical study to determine the most efficient band specific DeepFace detection model in terms of detection performance. The original object detection models that were selected for our study are the Faster R-CNN (Region based Convolutional Neural Network), SSD (Single-shot Multi-Box Detector) and R-FCN (Region-based Fully Convolutional Network). Also, the dual-band dataset used for this work is composed of two challenging MWIR and visible band face datasets, where the faces were captured under variable conditions, i.e. indoors, outdoors, different standoff distances (5 and 10 meters) and poses. Experimental results show that the proposed detection model yields the highest accuracy independent of the band and scenario used. Specifically, we show that a modified and tuned Faster R-CNN architecture with ResNet 101 is the most promising model when compared to all the other models tested. The proposed model yields accuracy of 99.2% and 98.4% when tested on thermal and visible face data respectively. Finally, while the proposed model is relatively slower than its competitors, our further experiments show that the speed of this network can be increased by reducing the number of proposals in RPN (Region Proposal Network), and thus, the computational complexity challenge is significantly minimized.

Facial attribute analysis is an important step in many biometric algorithms, including face recognition based human authentication. Detecting the state of face attributes is becoming even more popular in mobile based applications, where a biometric authentication application is used to quickly and accurately verify the claimed identity of the owners device, and also, keep the device secure if an intruder attempts to gain unauthorized access to it. While there is a large number of facial attributes that can be automatically detected to support a face recognition system, in this paper we focus on detecting three specific ones that are useful during both an enrollment and authentication process: (1) determining whether the eyes of a subject are open or closed, (2) determining whether a subject is wearing glasses or not, and, finally, (3) detecting whether a subjects facial pose is either frontal or non-frontal. These attributes are associated with face image quality control, which is a very useful component of modern FR systems under the following context: during live authentication, by limiting low quality face image data, we can enhance the face-based authentication accuracy, while at the same time improve user satisfaction via improved system efficiency (i.e. less false match attempts during authentication). Thus, to automatically and efficiently detect all of the aforementioned facial attributes, we developed both conventional and deep learning based models. These models are trained and tested on diverse and challenging face image datasets, using data captured from traditional cameras and mobile devices, when operating at multiple standoff distances, in either indoor or outdoor conditions. Our proposed attribute-specific detection models are robust, yielding up to 100% accuracy (in terms of F1 score) depending on the attribute tested, as well as the model and dataset(s) used for training and testing.

Ear recognition has its advantages in identifying non-cooperative individuals in unconstrained environments. Ear detection is a major step within the ear recognition algorithmic process. While conventional approaches for ear detection have been used in the past, Faster Region-based Convolutional Neural Network (Faster R-CNN) based detection methods have recently achieved superior detection performance in various benchmark studies, including those on face detection. In this work, we propose an ear detection system that uses Faster R-CNN. The training of the system is performed on two stages: First, an AlexNet model is trained for classifying ear vs. non-ear segments. Second, the unified Region Proposal Network (RPN) with the AlexNet, that shares the convolutional features, are trained for ear detection. The proposed system operates in real-time and accomplishes 98 % detection rate on a test set, composed of data coming from different ear datasets. In addition, the system's ear detection performance is high even when the test images are coming from un-controlled settings with a wide variety of images in terms of image quality, illumination and ear occlusion.

Online underground forums have been widely used by cybercriminals to trade the illicit products, resources and services, which have played a central role in the cybercrim-inal ecosystem. Unfortunately, due to the number of forums, their size, and the expertise required, it's infeasible to perform manual exploration to understand their behavioral processes. In this paper, we propose a novel framework named iDetector to automate the analysis of underground forums for the detection of cybercrime-suspected threads. In iDetector, to detect whether the given threads are cybercrime-suspected threads, we not only analyze the content in the threads, but also utilize the relations among threads, users, replies, and topics. To model this kind of rich semantic relationships (i.e., thread-user, thread-reply, thread-topic, reply-user and reply-topic relations), we introduce a structured heterogeneous information network (HIN) for representation, which is capable to be composed of different types of entities and relations. To capture the complex relationships (e.g., two threads are relevant if they were posted by the same user and discussed the same topic), we use a meta-structure based approach to characterize the semantic relatedness over threads. As different meta-structures depict the relatedness over threads at different views, we then build a classifier using Laplacian scores to aggregate different similarities formulated by different meta-structures to make predictions. To the best of our knowledge, this is the first work to use structural HIN to automate underground forum analysis. Comprehensive experiments on real data collections from underground forums (e.g., Hack Forums) are conducted to validate the effectiveness of our developed system iDetector in cybercrime-suspected thread detection by comparisons with other alternative methods.

Electrocardiography (ECG) is the process of recording the electrical activity of the human heart over time using electrodes that are placed over the skin. While the primary usage of electrocardiograms, the recorded signals, has been focused on the check of signs of heart-related diseases, recent studies have moved also toward their usage for human authentication. Thus, an ECG signal can be unique enough to be used independently as a biometric modality. In addition to its inherent liveness detection, it is easy to collect and can be easily captured either via sensors attached to the human body (fingertips, chest, wrist) or even passively using wireless sensors. In this paper, we propose a novel approach that exploits the spectro-temporal dynamic characteristics of the ECG signal to establish personal recognition system using both short-time Fourier transform (STFT) and generalized Morse wavelets (CWT). This process results in enriching the information extracted from the original ECG signal that is inserted in a 2D convolutional neural network (CNN) which extracts higher level and subject-specific ECG-based features for each individual. To validate our proposed CNN model, we performed nested cross-validation using eight different ECG databases. These databases are considered challenging since they include both normal and abnormal heartbeats as well as a dynamic number of subjects. Our proposed algorithms yield superior performance when compared to other state-ofart approaches discussed in the literature, i.e. the STFT-based one achieves an average identification rate, equal error rate (EER), and area under curve (AUC) of 97.86%, 0.0268, and 0.9933 respectively, whereas the CWT achieves comparable to STFT results in 97.5%, 0.0386, and 0.9882 respectively.

Arrhythmia is an abnormal heart rhythm that occurs due to the improper operation of the electrical impulses that coordinate the heartbeats. It is one of the most well-known heart conditions (including coronary artery disease, heart failure etc.) that is experienced by millions of people around the world. While there are several types of arrhythmias, not all of them are dangerous or harmful. However, there are arrhythmias that can often lead to death in minutes (e.g, ventricular fibrillation and ventricular tachycardia) even in young people. Thus, the detection of arrhythmia is critical for stopping and reversing its progression and for increasing longevity and life quality. While a doctor can perform different heart-monitoring tests specific to arrhythmias, the electrocardiogram (ECG) is one of the most common ones used either independently or in combination with other tests (to only detect, e.g. echocardiogram, or trigger arrhythmia and, then, detect, e.g. stress test). We propose a machine learning approach that augments the traditional arrhythmia detection approaches via our automatic arrhythmia classification system. It utilizes the texture of the ECG signal in both the temporal and spectro-temporal domains to detect and classify four types of heartbeats. The original ECG signal is first preprocessed, and then, the R-peaks associated with heartbeat estimation are identified. Next, 1D local binary patterns (LBP) in the temporal domain are utilized, while 2D LBPs and texture-based features extracted by a grayscale co-occurrence matrix (GLCM) are utilized in the spectro-temporal domain using the short-time Fourier transform (STFT) and Morse wavelets. Finally, different classifiers, as well as different ECG lead configurations are examined before we determine our proposed time-frequency SVM model, which obtains a maximum accuracy of 99.81%, sensitivity of 98.17%, and specificity of 99.98% when using a 10 cross-validation on the MIT-BIH database. Our approach yields competitive accuracy when compared to other methods discussed in the literature.

Because of the explosive growth of Android malware and due to the severity of its damages, the detection of Android malware has become an increasing important topic in cybersecurity. Currently, the major defense against Android malware is commercial mobile security products which mainly use signature-based method for detection. However, attackers can easily devise methods, such as obfuscation and repackaging, to evade the detection, which calls for new defensive techniques that are harder to evade. In this paper, resting on the analysis of Application Programming Interface (API) calls extracted from the smali files, we further categorize the API calls which belong to the some method in the smali code into a block. Based on the generated API call blocks, we then explore deep neural networks (i.e., Deep Belief Network (DBN) and Stacked AutoEncoders (SAEs)) for newly unknown Android malware detection. Using a real sample collection from Comodo Cloud Security Center, a comprehensive experimental study is performed to compare various malware detection approaches. The experimental results demonstrate that (1) our proposed feature extraction method (i.e., using API call blocks) outperforms using API calls directly in Android malware detection; (2) DBN works better than SAEs in this application; and (3) the detection performance of deep neural networks is better than shallow learning architectures.

In this paper, we propose a convolutional neural network (CNN) based, scenario-dependent and sensor (mobile device) adaptable hierarchical classification framework. Our proposed framework is designed to automatically categorize face data captured under various challenging conditions, before the FR algorithms (pre-processing, feature extraction and matching) are used. First, a unique multi-sensor database (using Samsung S4 Zoom, Nokia 1020, iPhone 5S and Samsung S5 phones) is collected containing face images indoors, outdoors, with yaw angle from -90° to +90° and at two different distances, i.e. 1 and 10 meters. To cope with pose variations, face detection and pose estimation algorithms are used for classifying the facial images into a frontal or a non-frontal class. Next, our proposed framework is used where tri-level hierarchical classification is performed as follows: Level 1, face images are classified based on phone type; Level 2, face images are further classified into indoor and outdoor images; and finally, Level 3 face images are classified into a close (1m) and a far, low quality, (10m) distance categories respectively. Experimental results show that classification accuracy is scenario dependent, reaching from 95 to more than 98% accuracy for level 2 and from 90 to more than 99% for level 3 classification. A set of experiments is performed indicating that, the usage of data grouping before the face matching is performed, resulted in a significantly improved rank-1 identification rate when compared to the original (all vs. all) biometric system.

Performing a direct match between images from different spectra (i.e., passive infrared and visible) is challenging because each spectrum contains different information pertaining to the subject's face. In this work, we investigate the benefits and limitations of using synthesized visible face images from thermal ones and vice versa in cross-spectral face recognition systems. For this purpose, we propose utilizing canonical correlation analysis (CCA) and manifold learning dimensionality reduction (LLE). There are four primary contributions of this work. First, we formulate the cross-spectral heterogeneous face matching problem (visible to passive IR) using an image synthesis framework. Second, a new processed database composed of two datasets consistent of separate controlled frontal face subsets (VIS-MWIR and VIS-LWIR) is generated from the original, raw face datasets collected in three different bands (visible, MWIR and LWIR). This multi-band database is constructed using three different methods for preprocessing face images before feature extraction methods are applied. There are: (1) face detection, (2) CSU's geometric normalization, and (3) our recommended geometric normalization method. Third, a post-synthesis image denoising methodology is applied, which helps alleviate different noise patterns present in synthesized images and improve baseline FR accuracy (i.e. before image synthesis and denoising is applied) in practical heterogeneous FR scenarios. Finally, an extensive experimental study is performed to demonstrate the feasibility and benefits of cross-spectral matching when using our image synthesis and denoising approach. Our results are also compared to a baseline commercial matcher and various academic matchers provided by the CSU's Face Identification Evaluation System.

In this work, our objective is to study the impact of statistical hypothesis tests for the purpose of improving heterogeneous face recognition (FR). A series of tests are conducted to find the most suitable type of statistical analysis test (parametric vs. non-parametric). To conduct the experiments, we used a multi-spectral face database (visible and Near-IR) collected under challenging conditions, i.e. at night time and at four different standoff distances, namely 30, 60,90 and 120 meters. Next, the selected statistical analysis test is used to find the statistical significance of; (i) image restoration, (ii) fusion of scores. First, Gabor Wavelets, Histogram of gradients (HOG) and Local binary patterns (LBP) feature descriptors are empirically selected. Then the statistical analysis reveals which descriptors result in higher recognition performance. Finally, statistical hypothesis tests are performed to explore the impact of data stratification (grouping of gallery and probe sets) in terms of ethnicity, gender. A set of face identification studies are performed. Experimental results suggest that our proposed image restoration approach, fusion schemes and the usage of stratification result in a significantly better performance results than the baseline, e.g. the rank-one score is improved from 50% to 71% when using image restoration, to 73% when using fusion of scores and to 75% (i.e. in the case of testing FR accuracy only on the female Asian class) when employing database stratification.

Surveillance systems are very important for law enforcement and military applications. Capturing a biometric modality at a distance and under difficult conditions is a very challenging process. While face or gait can be used to identify an individual in such application, tattoos can also help in the identification process whenever available. Tattoos are considered a soft biometric and in some scenarios may be the only clue that can be used to verify the identity of a suspect or to rule out a suspect. One of the major challenges in tattoo recognition systems is image registration, i.e. the alignment of one tattoo image to a reference image. Accurate registration can greatly improve recognition accuracy. In this paper, we propose a two-level automatic tattoo registration and correction system based on SIFT descriptors and the RANSAC algorithm with a homography model. By using image quality index techniques and a postprocessing step (where we refine our original registration results by an automated correction process where outliers are first identified and then re-processed), our system is able to demonstrate accurate registration results. We tested our registration system using two tattoo image databases. The first one is the NIST-Tatt-C database with 109 subjects collected under uncontrolled condition, and the second one is the new WVU tattoo database (WVU-Tatt) with 79 subjects, which is collected under controlled conditions. Experimental results show that, first, we obtained 100% registration accuracy in both databases. Then, the effect of our registration process on tattoo recognition performance was assessed when using both the NIST-Tatt-C database where the accuracy improved from 54.13% (no registration) to 100% (with registration) and the WVU-Tatt database where the accuracy improved from 86.08% (no registration) to 98.73% (with registration).

Currently, binary biclustering algorithms are too slow and non-specific to handle biological datasets that have a large number of attributes, which is essential for the computational biology problem of microarray analysis. Specialized computers may be needed to execute an algorithm, and may fail to produce a solution, due to its large resource needs. The biclusters also include too many false positives, the type I error, which hinders biological discovery. We propose an algorithm that can analyze datasets with a large attribute set at different densities, and can operate on a laptop, which makes it accessible to practitioners. EMFP produces biclusters that have a very low Root Mean Squared Error and false positive rate, with very few type II errors. Our binary biclustering algorithm is a hybrid, axis-parallel, pattern-based algorithm that finds multiple, non-overlapping, near-constant, deterministic, binary submatricies, with a variable confidence threshold, and the novel use of local density comparisons versus the standard global threshold. EMFP introduces a new, and intuitive way to calculate internal measures for binary biclustering methods. We also introduce a framework to ease comparison with other algorithms, and compare to both binary and general biclustering algorithms using two real, and 80 synthetic databases.

In this paper we study the problems of intra-spectral and cross-spectral face recognition (FR) in homogeneous and heterogeneous environments. Specifically we investigate the advantages and limitations of matching (i) short wave infrared (SWIR) face images to visible images under controlled or uncontrolled conditions, (ii) mid-wave infrared (MWIR) to MWIR or visible images under controlled conditions, and (iii) intra-distance near infrared (NIR) to NIR images and cross-distance, cross-spectral NIR to visible images. All NIR images were captured night-time, outdoors and at mid-ranges (from 30 up to 120 meters). We utilized both commercial and academic face matchers and performed a set of experiments indicating that our cross-photometric score level fusion rule can be utilized to improve SWIR cross-spectral matching performance across all FR scenarios investigated. We also show that intra-spectral matching results, using either MWIR or NIR images, are comparable to the baseline results, i.e., when comparing visible to visible face images. Our experiments also indicate that the level of improvement in recognition performance is scenario dependent. Experiments also show that cross-spectral matching (the heterogeneous problem, where gallery and probe sets have face images acquired in different spectral bands) is a very challenging problem and it requires further investigation to address real-world law enforcement or military situations.

We investigate the question of whether facial metrology can be exploited for reliable gender prediction. A new method based solely on metrological information from facial landmarks is developed. Here, metrological features are defined in terms of specially normalized angle and distance measures and computed based on given landmarks on facial images. The performance of the proposed metrology- based method is compared with that of a state-of-the-art appearance-based method for gender classification. Results are reported on two standard face databases, namely, MUCT and XM2VTS containing 276 and 295 images, respectively. The performance of the metrology-based approach was slightly lower than that of the appearance- based method by only about 3.8% for the MUCT database and about 5.7% for the XM2VTS database.

In this paper we study the problem of cross spectral face recognition in heterogeneous environments. Specifically we investigate the advantages and limitations of matching short wave infrared (SWIR) face images to visible images under controlled or uncontrolled conditions. The contributions of this work are three-fold. First, three different databases are considered, which represent three different data collection conditions, i.e., images acquired in fully controlled (indoors), semi-controlled (indoors at standoff distances ≥ 50m), and uncontrolled (outdoor operational conditions) environments. Second, we demonstrate the possibility of SWIR cross-spectral matching under controlled and challenging scenarios. Third, we illustrate how photometric normalization and our proposed cross-photometric score level fusion rule can be utilized to improve cross-spectral matching performance across all scenarios. We utilized both commercial and academic (texture-based) face matchers and performed a set of experiments indicating that SWIR images can be matched to visible images with encouraging results. Our experiments also indicate that the level of improvement in recognition performance is scenario dependent.

We study the problem of restoring severely degraded face images such as images scanned from passport photos or images subjected to fax compression, downscaling, and printing. The purpose of this paper is to illustrate the complexity of face recognition in such realistic scenarios and to provide a viable solution to it. The contributions of this work are two-fold. First, a database of face images is assembled and used to illustrate the challenges associated with matching severely degraded face images. Second, a preprocessing scheme with low computational complexity is developed in order to eliminate the noise present in degraded images and restore their quality. An extensive experimental study is performed to establish that the proposed restoration scheme improves the quality of the ensuing face images while simultaneously improving the performance of face matching.

The problem of face verification across the short wave infrared spectrum (SWIR) is studied in order to illustrate the advantages and limitations of SWIR face verification. The contributions of this work are two-fold. First, a database of 50 subjects is assembled and used to illustrate the challenges associated with the problem. Second, a set of experiments is performed in order to demonstrate the possibility of SWIR cross-spectral matching. Experiments also show that images captured under different SWIR wavelengths can be matched to visible images with promising results. The role of multispectral fusion in improving recognition performance in SWIR images is finally illustrated. To the best of our knowledge, this is the first time cross-spectral SWIR face recognition is being investigated in the open literature.

In this paper the problem of eye detection across three different bands, i.e., the visible, multispectral, and short wave infrared (SWIR), is studied in order to illustrate the advantages and limitations of multi-band eye localization. The contributions of this work are two-fold. First, a multi-band database of 30 subjects is assembled and used to illustrate the challenges associated with the problem. Second, a set of experiments is performed in order to demonstrate the possibility for multi-band eye detection. Experiments show that the eyes on face images captured under different bands can be detected with promising results. Finally, we illustrate that recognition performance in all studied bands is favorably affected by the geometric normalization of raw face images that is based on our proposed detection methodology. To the best of our knowledge this is the first time that this problem is being investigated in the open literature in the context of human eye localization across different bands.

A new multimodal biometric database designed and acquired within the framework of the European BioSecure Network of Excellence is presented. It is comprised of more than 600 individuals acquired simultaneously in three scenarios: 1) over the Internet, 2) in an office environment with desktop PC, and 3) in indoor/outdoor environments with mobile portable hardware. The three scenarios include a common part of audio/video data. Also, signature and fingerprint data have been acquired both with desktop PC and mobile portable hardware. Additionally, hand and iris data were acquired in the second scenario using desktop PC. Acquisition has been conducted by 11 European institutions. Additional features of the BioSecure Multimodal Database (BMDB) are: two acquisition sessions, several sensors in certain modalities, balanced gender and age distributions, multimodal realistic scenarios with simple and quick tasks per modality, cross-European diversity, availability of demographic data, and compatibility with other multimodal databases. The novel acquisition conditions of the BMDB allow us to perform new challenging research and evaluation of either monomodal or multimodal biometric systems, as in the recent BioSecure Multimodal Evaluation campaign. A description of this campaign including baseline results of individual modalities from the new database is also given. The database is expected to be available for research purposes through the BioSecure Association during 2008.

As biometric technology is rolled out on a larger scale, it will be a common scenario (known as cross-device matching) to have a template acquired by one biometric device used by another during testing. This requires a biometric system to work with different acquisition devices, an issue known as device interoperability. We further distinguish two subproblems, depending on whether the device identity is known or unknown. In the latter case, we show that the device information can be probabilistically inferred given quality measures (e.g., image resolution) derived from the raw biometric data. By keeping the template unchanged, cross-device matching can result in significant degradation in performance. We propose to minimize this degradation by using device-specific quality-dependent score normalization. In the context of fusion, after having normalized each device output independently, these outputs can be combined using the naive Bayes principal. We have compared and categorized several state-of-the-art quality-based score normalization procedures, depending on how the relationship between quality measures and score is modeled, as follows: 1) direct modeling; 2) modeling via the cluster index of quality measures; and 3) extending 2) to further include the device information (device-specific cluster index). Experimental results carried out on the Biosecure DS2 data set show that the last approach can reduce both false acceptance and false rejection rates simultaneously. Furthermore, the compounded effect of normalizing each system individually in multimodal fusion is a significant improvement in performance over the baseline fusion (without using any quality information) when the device information is given.

Automatically verifying the identity of a person by means of biometrics (e.g., face and fingerprint) is an important application in our day-to-day activities such as accessing banking services and security control in airports. To increase the system reliability, several biometric devices are often used. Such a combined system is known as a multimodal biometric system. This paper reports a benchmarking study carried out within the framework of the BioSecure DS2 (Access Control) evaluation campaign organized by the University of Surrey, involving face, fingerprint, and iris biometrics for person authentication, targeting the application of physical access control in a medium-size establishment with some 500 persons. While multimodal biometrics is a well-investigated subject in the literature, there exists no benchmark for a fusion algorithm comparison. Working towards this goal, we designed two sets of experiments: quality-dependent and cost-sensitive evaluation. The quality-dependent evaluation aims at assessing how well fusion algorithms can perform under changing quality of raw biometric images principally due to change of devices. The cost-sensitive evaluation, on the other hand, investigates how well a fusion algorithm can perform given restricted computation and in the presence of software and hardware failures, resulting in errors such as failure-to-acquire and failure-to-match. Since multiple capturing devices are available, a fusion algorithm should be able to handle this nonideal but nevertheless realistic scenario. In both evaluations, each fusion algorithm is provided with scores from each biometric comparison subsystem as well as the quality measures of both the template and the query data. The response to the call of the evaluation campaign proved very encouraging, with the submission of 22 fusion systems. To the best of our knowledge, this campaign is the first attempt to benchmark quality-based multimodal fusion algorithms. In the presence of changing image quality which may be due to a change of acquisition devices and/or device capturing configurations, we observe that the top performing fusion algorithms are those that exploit automatically derived quality measurements. Our evaluation also suggests that while using all the available biometric sensors can definitely increase the fusion performance, this comes at the expense of increased cost in terms of acquisition time, computation time, the physical cost of hardware, and its maintenance cost. As demonstrated in our experiments, a promising solution which minimizes the composite cost is sequential fusion, where a fusion algorithm sequentially uses match scores until a desired confidence is reached, or until all the match scores are exhausted, before outputting the final combined score.

The proposed demonstration is based on the work performed as part of the NEAT-o-Games project. NEAT-o-Games is a suite of games that runs on mobile terminals such as cell phones. Unlike other games, NEAT-o-Games' primary goal is to become part of people's everyday routines and attack the behavioral aspect of the sedentary lifestyle. Their main characteristic is that they are not carried out in short bouts, but are being played continuously and are interwoven in the daily routine of the players. Data from wearable accelerometers are logged wirelessly to a cell phone and control the animation of the player in a virtual race game (avatar) with other players over the cellular network. Players can use their excess of activity points earned from the race game to get hints in mental games of the suite, like Sudoku. Initial studies indicate that NEAT-o-Games may bring a positive physical, psychological, and social impact on players.

The optimisation of a smart card face verification system (SCFVS) design is a complex task. As the parameters involved are not independent, the search space is of exponential complexity. We investigate simplified optimisation strategies and demonstrate that both system performance and speed of access can be improved by jointly optimised parameter setting and level of probe compression. Experimental results suggest that the choice of one strategy over another is a matter of the amount of time available for the system design, system performance and response time.