The first layer of a deep CNN backbone applies filters to an image to extract the basic features available to later layers. During training, some filters may go inactive, meaning all weights in the filter are near zero. An inactive filter in the final model represents a missed opportunity to extract a useful feature. This phenomenon is especially prevalent in specialized CNNs such as for face recognition (as opposed to, e.g., ImageNet). For example, in one of the most widely-used face recognition models (ArcFace), about half of the filters in the first layer are inactive. We propose a novel approach designed and tested specifically for face recognition networks, known as “CRAFT: Contextual Re-Activation of Filters for Face Recognition Training”. Additionally, CRAFT achieves statistically significant improvements in accuracy over standard training on face recognition benchmarks such as AgeDB-30, CPLFW, LFW, CALFW, CFP-FP, IJBB, and IJBC where accuracy has largely saturated. Notable improvements are observed, with significant gains on the highly challenging Hadrian and Eclipse datasets.

Understanding how facial affect analysis (FAA) systems perform across different demographic groups requires reliable measurement of sensitive attributes such as ancestry, often approximated by skin tone, which itself is highly influenced by lighting conditions. This study compares two objective skin tone classification methods: the widely used Individual Typology Angle (ITA) and a perceptually grounded alternative based on Lightness $\left(L^{*}\right)$ and Hue $\left(H^{*}\right)$. Using AffectNet and a MobileNetbased model, we assess fairness across skin tone groups defined by each method. Results reveal a severe underrepresentation of dark skin tones ($\sim \mathbf{2} \%$), alongside fairness disparities in F1-score (up to 0.08) and TPR (up to 0.11) across groups. While ITA shows limitations due to its sensitivity to lighting, the $H^{*}-L^{*}$ method yields more consistent subgrouping and enables clearer diagnostics through metrics such as Equal Opportunity. Grad-CAM analysis further highlights differences in model attention patterns by skin tone, suggesting variation in feature encoding. To support future mitigation efforts, we also propose a modular fairness-aware pipeline that integrates perceptual skin tone estimation, model interpretability, and fairness evaluation. These findings emphasize the relevance of skin tone measurement choices in fairness assessment and suggest that ITA-based evaluations may overlook disparities affecting darker-skinned individuals.

Given a facial matcher, in explainable face verification, the task is to answer: how relevant are the parts of a probe image to establish the matching with an enrolled image. In many cases, however, the trained models cannot be manipulated and must be treated as "black-boxes". In this paper, we present six different saliency maps that can be used to explain any face verification algorithm with no manipulation inside of the face recognition model. The key idea of the methods is based on how the matching score of the two face images changes when the probe is perturbed. The proposed methods remove and aggregate different parts of the face, and measure contributions of these parts individually and in-collaboration as well. We test and compare our proposed methods in three different scenarios: synthetic images with different qualities and occlusions, real face images with different facial expressions, poses, and occlusions and faces from different demographic groups. In our experiments, five different face verification algorithms are used: ArcFace, Dlib, FaceNet (trained on VGGface2 and CasiaWebFace), and LBP. We conclude that one of the proposed methods achieves saliency maps that are stable and interpretable to humans. In addition, our method, in combination with a new visualization of saliency maps based on contours, shows promising results in comparison with other state-of-the-art art methods. This paper presents good insights into any face verification algorithm, in which it can be clearly appreciated which are the most relevant face areas that an algorithm takes into account to carry out the recognition process.

With the recent surge in threats to public safety, the security focus of several organizations has been moved towards enhanced intelligent screening systems. Conventional X-ray screening, which relies on the human operator is the best use of this technology, allowing for the more accurate identification of potential threats. This paper explores X-ray security imagery by introducing a novel approach that generates realistic synthesized data, which opens up the possibility of using different settings to simulate occlusion, radiopacity, varying textures, and distractors to generate cluttered scenes. The generated synthetic data is effective in the training of deep networks. It allows better generalization on training data to deal with domain adaptation in the real world. The extensive set of experiments in this paper provides evidence for the efficacy of synthetic datasets over human-annotated datasets for automated X-ray security screening. The proposed approach outperforms the state-of-the-art approach for a diverse threat object dataset on mean Average Precision (mAP) of region-based detectors and classification/regression-based detectors.

Matching live images (“selfies”) to images from ID documents is a problem that can arise in various applications. A challenging instance of the problem arises when the face image on the ID document is from early adolescence and the live image is from later adolescence. We explore this problem using a private dataset called Chilean Young Adult (CHIYA) dataset, where we match live face images taken at age 18-19 to face images on scanned ID documents created at ages 9 to 18. State-of-the-art deep learning face matchers (e.g., ArcFace) have relatively poor accuracy for document-to-selfie face matching. To achieve higher accuracy, we fine-tune the best available open-source model with triplet loss for a few-shot learning. Experiments show that our approach achieves higher accuracy than the DocFace+ model recently developed for this problem. Our fine-tuned model was able to improve the true acceptance rate for the most difficult (largest age span) subset from 62.92% to 96.67% at a false acceptance rate of 0.01%. Our fine-tuned model is available for use by other researchers.

To follow the attendance of students is a major concern in many educational institutions. The manual management of the attendance sheets is laborious for crowded classrooms. In this paper, we propose and evaluate a general methodology for the automated student attendance system that can be used in crowded classrooms, in which the session images are taken by a smartphone camera. We release a realistic full-annotated dataset of images of a classroom with around 70 students in 25 sessions, taken during 15 weeks. Ten face recognition algorithms based on learned and handcrafted features are evaluated using a protocol that takes into account the number of face images per subject used in the gallery. In our experiments, the best one has been FaceNet, a method based on deep learning features, achieving around 95% of accuracy with only one enrollment image per subject. We believe that our automated student attendance system based on face recognition can be used to save time for both teacher and students and to prevent fake attendance.

Although face recognition systems have achieved impressive performance in recent years, the low-resolution face recognition task remains challenging, especially when the low-resolution faces are captured under non-ideal conditions, which is widely prevalent in surveillance-based applications. Faces captured in such conditions are often contaminated by blur, non-uniform lighting, and non-frontal face pose. In this paper, we analyze the face recognition techniques using data captured under low-quality conditions in the wild. We provide a comprehensive analysis of the experimental results for two of the most important applications in real surveillance applications, and demonstrate practical approaches to handle both cases that show promising performance. The following three contributions are made: (i) we conduct experiments to evaluate super-resolution methods for low-resolution face recognition; (ii) we study face re-identification on various public face datasets, including real surveillance and low-resolution subsets of large-scale datasets, presenting a baseline result for several deep learning-based approaches, and improve them by introducing a generative adversarial network pre-training approach and fully convolutional architecture; and (iii) we explore the low-resolution face identification by employing a state-of-the-art supervised discriminative learning approach. The evaluations are conducted on challenging portions of the SCface and UCCSface datasets.

This paper addresses the problem of automated recognition of faces and facial attributes by proposing a new general approach called Accumulative Local Sparse Representation (ALSR). In the learning stage, we build a general dictionary of patches that are extracted from face images in a dense manner on a grid. In the testing stage, patches of the query image are sparsely represented using a local dictionary. This dictionary contains similar atoms of the general dictionary that are spatially in the same neighborhood. If the sparsity concentration index of the query patch is high enough, we build a descriptor by using a sum-pooling operator that evaluates the contribution provided by the atoms of each class. The classification is performed by maximizing the sum of the descriptors of all selected patches. ALSR can learn a model for each recognition task dealing with more variability in ambient lighting, pose, expression, occlusion, face size, etc. Experiments on three popular face databases (LFW for faces, AR for gender and Oulu-CASIA for expressions), show that ALSR outperforms representative methods in the literature, when a huge number of training images is not available.

Person re-identification (ReID) is the task of automatically matching persons across surveillance cameras with location or time differences. Nearly all proposed ReID approaches exploit body features. Even if successfully captured in the scene, faces are often assumed to be unhelpful to the ReID process[3]. As cameras and surveillance systems improve, ‘Facial ReID’ approaches deserve attention. The following contributions are made in this work: 1) We describe a high-quality dataset for person re-identification featuring faces. This dataset was collected from a real surveillance network in a municipal rapid transit system, and includes the same people appearing in multiple sites at multiple times wearing different attire. 2) We employ new DNN architectures and patch matching techniques to handle face misalignment in quality regimes where landmarking fails. We further boost the performance by adopting the fully convolutional structure and spatial pyramid pooling (SPP).

To ensure safety in the construction of important metallic components for roadworthiness, it is necessary to check every component thoroughly using non-destructive testing. In last decades, X-ray testing has been adopted as the principal non-destructive testing method to identify defects within a component which are undetectable to the naked eye. Nowadays, modern computer vision techniques, such as deep learning and sparse representations, are opening new avenues in automatic object recognition in optical images. These techniques have been broadly used in object and texture recognition by the computer vision community with promising results in optical images. However, a comprehensive evaluation in X-ray testing is required. In this paper, we release a new dataset containing around 47.500 cropped X-ray images of 32 32 pixels with defects and no-defects in automotive components. Using this dataset, we evaluate and compare 24 computer vision techniques including deep learning, sparse representations, local descriptors and texture features, among others. We show in our experiments that the best performance was achieved by a simple LBP descriptor with a SVM-linear classifier obtaining 97% precision and 94% recall. We believe that the methodology presented could be used in similar projects that have to deal with automated detection of defects.

Unconstrained face recognition is still an open problem as the state-of-the-art algorithms have not yet reached high recognition performance in real-world environments. This paper addresses this problem by proposing a new approach called sparse fingerprint classification algorithm (SFCA). In the training phase, for each enrolled subject, a grid of patches is extracted from each subject's face images in order to construct representative dictionaries. In the testing phase, a grid is extracted from the query image and every patch is transformed into a binary sparse representation using the dictionary, creating a fingerprint of the face. The binary coefficients vote for their corresponding classes and the maximum-vote class decides the identity of the query image. Experiments were carried out on seven widely-used face databases. The results demonstrate that when the size of the data set is small or medium (e.g., the number of subjects is not greater than one hundred), SFCA is able to deal with a larger degree of variability in ambient lighting, pose, expression, occlusion, face size, and distance from the camera than other current state-of-the-art algorithms.

X-ray screening systems have been used to safeguard environments in which access control is of paramount importance. Security checkpoints have been placed at the entrances to many public places to detect prohibited items, such as handguns and explosives. Generally, human operators are in charge of these tasks as automated recognition in baggage inspection is still far from perfect. Research and development on X-ray testing is, however, exploring new approaches based on computer vision that can be used to aid human operators. This paper attempts to make a contribution to the field of object recognition in X-ray testing by evaluating different computer vision strategies that have been proposed in the last years. We tested ten approaches. They are based on bag of words, sparse representations, deep learning, and classic pattern recognition schemes among others. For each method, we: 1) present a brief explanation; 2) show experimental results on the same database; and 3) provide concluding remarks discussing pros and cons of each method. In order to make fair comparisons, we define a common experimental protocol based on training, validation, and testing data (selected from the public GDXray database). The effectiveness of each method was tested in the recognition of three different threat objects: 1) handguns; 2) shuriken (ninja stars); and 3) razor blades. In our experiments, the highest recognition rate was achieved by methods based on visual vocabularies and deep features with more than 95% of accuracy. We strongly believe that it is possible to design an automated aid for the human inspection task using these computer vision algorithms.

This work presents an approach to category-based action recognition in video using sparse coding techniques. The proposed approach includes two main contributions: i) A new method to handle intra-class variations by decomposing each video into a reduced set of representative atomic action acts or key-sequences, and ii) A new video descriptor, ITRA: Inter-Temporal Relational Act Descriptor, that exploits the power of comparative reasoning to capture relative similarity relations among key-sequences. In terms of the method to obtain key-sequences, we introduce a loss function that, for each video, leads to the identification of a sparse set of representative key-frames capturing both, relevant particularities arising in the input video, as well as relevant generalities arising in the complete class collection. In terms of the method to obtain the ITRA descriptor, we introduce a novel scheme to quantify relative intra and inter-class similarities among local temporal patterns arising in the videos. The resulting ITRA descriptor demonstrates to be highly effective to discriminate among action categories. As a result, the proposed approach reaches remarkable action recognition performance on several popular benchmark datasets, outperforming alternative state-of the-art techniques by a large margin.

Baggage inspection using X-ray screening is a priority task that reduces the risk of crime and terrorist attacks. Manual detection of threat items is tedious because very few bags actually contain threat items and the process requires a high degree of concentration. An automated solution would be a welcome development in this field. We propose a methodology for automatic detection of threat objects using single X-ray images. Our approach is an adaptation of a methodology originally created for recognizing objects in photographs based on implicit shape models. Our detection method uses a visual vocabulary and an occurrence structure generated from a training dataset that contains representative X-ray images of the threat object to be detected. Our method can be applied to single views of grayscale X-ray images obtained using a single energy acquisition system. We tested the effectiveness of our method for the detection of three different threat objects: 1) razor blades; 2) shuriken (ninja stars); and 3) handguns. The testing dataset for each threat object consisted of 200 X-ray images of bags. The true positive and false positive rates (TPR and FPR) are: (0.99 and 0.02) for razor blades, (0.97 and 0.06) for shuriken, and (0.89 and 0.18) for handguns. If other representative training datasets were utilized, we believe that our methodology could aid in the detection of other kinds of threat objects.

This work describes our experience developing a system to access density and flow of people in large indoor spaces using a network of RGB cameras. The proposed system is based on a set of overlapped and calibrated cameras. This facilitates the use of geometric constraints that help to reduce visual ambiguities. These constraints are combined with classifiers based on visual appearance to produce an efficient and robust method to detect and track humans. In this work, we argue that flow and density of people are low level measurements that need to be complemented with suitable analytic tools to bridge semantic gaps and become useful information for a target application. Consequently, we also propose a set of analytic tools that help a human user to effectively take advantage of the measurements provided by the system. Finally, we report results that demonstrate the relevance of the proposed ideas.

This paper presents the results of applying Haralik's textures in the upland landform classification process. In environmental and landuse projects is important to know what kind of landforms are present in a geographic area. The objective of this work is to improve the performance of such classifications, adding information to the commonly used data features in such problems. The texture information was extracted using the method of Haralick, parameterized with moving windows of geo-referenced maps in raster format. Those maps contains information of the terrain morphology such as elevation, slopes, etc. Several tests were performed using different classifiers and cross-validation over a dataset with a total of 203401 samples. It was shown that Haralick's textures features are useful for the problem, because its performance is higher (about 97 %) in comparison with the achieved when using only morphological information of the terrain.

Histograms of Local Binary Patterns (LBPs) and variations thereof are a popular local visual descriptor for face recognition. So far, most variations of LBP are designed by hand or are learned with non-supervised methods. In this work we propose a simple method to learn discriminative LBPs in a supervised manner. The method represents an LBP-like descriptor as a set of pixel comparisons within a neighborhood and heuristically seeks for a set of pixel comparisons so as to maximize a Fisher separability criterion for the resulting histograms. Tests on standard face recognition datasets show that this method can create compact yet discriminative descriptors.

Presents a method for inspecting aluminum castings automatically from a sequence of radioscopic images taken at different positions of the casting. The classic image-processing methods for flaw detection of aluminum castings use a bank of filters to generate an error-free reference image. This reference image is compared with the real radioscopic image, and flaws are detected at the pixels where the difference between them is considerable. However, the configuration of each filter depends strongly on the size and shape of the structure of the casting under inspection. A two-step technique is proposed to detect flaws automatically and that uses a single filter. First, the method identifies potential defects in each image of the sequence, and second, it matches and tracks them from image to image. The key idea of the paper is to consider as false alarms those potential defects which cannot be tracked in the sequence. The robustness and reliability of the method have been verified on both real data in which synthetic flaws have been added and real radioscopic image sequences recorded from cast aluminum wheels with known defects. Using this method, the real defects can be detected with high certainty. This approach achieves good discrimination from false alarms.