Raymond Veldhuis

Zahra Atashgahi, Tennison Liu, Mykola Pechenizkiy et al.

Sparse Neural Networks (SNNs) have emerged as powerful tools for efficient feature selection. Leveraging the dynamic sparse training (DST) algorithms within SNNs has demonstrated promising feature selection capabilities while drastically reducing computational overheads. Despite these advancements, several critical aspects remain insufficiently explored for feature selection. Questions persist regarding the choice of the DST algorithm for network training, the choice of metric for ranking features/neurons, and the comparative performance of these methods across diverse datasets when compared to dense networks. This paper addresses these gaps by presenting a comprehensive systematic analysis of feature selection with sparse neural networks. Moreover, we introduce a novel metric considering sparse neural network characteristics, which is designed to quantify feature importance within the context of SNNs. Our findings show that feature selection with SNNs trained with DST algorithms can achieve, on average, more than $50\%$ memory and $55\%$ FLOPs reduction compared to the dense networks, while outperforming them in terms of the quality of the selected features. Our code and the supplementary material are available on GitHub (\url{https://github.com/zahraatashgahi/Neuron-Attribution}).

Zahra Atashgahi, Xuhao Zhang, Neil Kichler et al.

Feature selection that selects an informative subset of variables from data not only enhances the model interpretability and performance but also alleviates the resource demands. Recently, there has been growing attention on feature selection using neural networks. However, existing methods usually suffer from high computational costs when applied to high-dimensional datasets. In this paper, inspired by evolution processes, we propose a novel resource-efficient supervised feature selection method using sparse neural networks, named \enquote{NeuroFS}. By gradually pruning the uninformative features from the input layer of a sparse neural network trained from scratch, NeuroFS derives an informative subset of features efficiently. By performing several experiments on $11$ low and high-dimensional real-world benchmarks of different types, we demonstrate that NeuroFS achieves the highest ranking-based score among the considered state-of-the-art supervised feature selection models. The code is available on GitHub.

Zahra Atashgahi, Decebal Constantin Mocanu, Raymond Veldhuis et al.

Change-point detection (CPD), which detects abrupt changes in the data distribution, is recognized as one of the most significant tasks in time series analysis. Despite the extensive literature on offline CPD, unsupervised online CPD still suffers from major challenges, including scalability, hyperparameter tuning, and learning constraints. To mitigate some of these challenges, in this paper, we propose a novel deep learning approach for unsupervised online CPD from multi-dimensional time series, named Adaptive LSTM-Autoencoder Change-Point Detection (ALACPD). ALACPD exploits an LSTM-autoencoder-based neural network to perform unsupervised online CPD. It continuously adapts to the incoming samples without keeping the previously received input, thus being memory-free. We perform an extensive evaluation on several real-world time series CPD benchmarks. We show that ALACPD, on average, ranks first among state-of-the-art CPD algorithms in terms of quality of the time series segmentation, and it is on par with the best performer in terms of the accuracy of the estimated change-points. The implementation of ALACPD is available online on Github\footnote{\url{https://github.com/zahraatashgahi/ALACPD}}.

Mahdi Ghafourian, Julian Fierrez, Luis Felipe Gomez et al.

The remarkable success of face recognition (FR) has endangered the privacy of internet users particularly in social media. Recently, researchers turned to use adversarial examples as a countermeasure. In this paper, we assess the effectiveness of using two widely known adversarial methods (BIM and ILLC) for de-identifying personal images. We discovered, unlike previous claims in the literature, that it is not easy to get a high protection success rate (suppressing identification rate) with imperceptible adversarial perturbation to the human visual system. Finally, we found out that the transferability of adversarial examples is highly affected by the training parameters of the network with which they are generated.

Marcel Grimmer, Christian Rathgeb, Raymond Veldhuis et al.

Accurate face recognition systems are increasingly important in sensitive applications like border control or migration management. Therefore, it becomes crucial to quantify the quality of facial images to ensure that low-quality images are not affecting recognition accuracy. In this context, the current draft of ISO/IEC 29794-5 introduces the concept of component quality to estimate how single factors of variation affect recognition outcomes. In this study, we propose a quality measure (NeutrEx) based on the accumulated distances of a 3D face reconstruction to a neutral expression anchor. Our evaluations demonstrate the superiority of our proposed method compared to baseline approaches obtained by training Support Vector Machines on face embeddings extracted from a pre-trained Convolutional Neural Network for facial expression classification. Furthermore, we highlight the explainable nature of our NeutrEx measures by computing per-vertex distances to unveil the most impactful face regions and allow operators to give actionable feedback to subjects.

Shunxin Wang, Raymond Veldhuis, Christoph Brune et al.

Frequency analysis is useful for understanding the mechanisms of representation learning in neural networks (NNs). Most research in this area focuses on the learning dynamics of NNs for regression tasks, while little for classification. This study empirically investigates the latter and expands the understanding of frequency shortcuts. First, we perform experiments on synthetic datasets, designed to have a bias in different frequency bands. Our results demonstrate that NNs tend to find simple solutions for classification, and what they learn first during training depends on the most distinctive frequency characteristics, which can be either low- or high-frequencies. Second, we confirm this phenomenon on natural images. We propose a metric to measure class-wise frequency characteristics and a method to identify frequency shortcuts. The results show that frequency shortcuts can be texture-based or shape-based, depending on what best simplifies the objective. Third, we validate the transferability of frequency shortcuts on out-of-distribution (OOD) test sets. Our results suggest that frequency shortcuts can be transferred across datasets and cannot be fully avoided by larger model capacity and data augmentation. We recommend that future research should focus on effective training schemes mitigating frequency shortcut learning.

Shunxin Wang, Christoph Brune, Raymond Veldhuis et al.

Neural networks are prone to learn easy solutions from superficial statistics in the data, namely shortcut learning, which impairs generalization and robustness of models. We propose a data augmentation strategy, named DFM-X, that leverages knowledge about frequency shortcuts, encoded in Dominant Frequencies Maps computed for image classification models. We randomly select X% training images of certain classes for augmentation, and process them by retaining the frequencies included in the DFMs of other classes. This strategy compels the models to leverage a broader range of frequencies for classification, rather than relying on specific frequency sets. Thus, the models learn more deep and task-related semantics compared to their counterpart trained with standard setups. Unlike other commonly used augmentation techniques which focus on increasing the visual variations of training data, our method targets exploiting the original data efficiently, by distilling prior knowledge about destructive learning behavior of models from data. Our experimental results demonstrate that DFM-X improves robustness against common corruptions and adversarial attacks. It can be seamlessly integrated with other augmentation techniques to further enhance the robustness of models.

Una M. Kelly, Meike Nauta, Lu Liu et al.

A morph is a combination of two separate facial images and contains identity information of two different people. When used in an identity document, both people can be authenticated by a biometric Face Recognition (FR) system. Morphs can be generated using either a landmark-based approach or approaches based on deep learning such as Generative Adversarial Networks (GAN). In a recent paper, we introduced a \emph{worst-case} upper bound on how challenging morphing attacks can be for an FR system. The closer morphs are to this upper bound, the bigger the challenge they pose to FR. We introduced an approach with which it was possible to generate morphs that approximate this upper bound for a known FR system (white box), but not for unknown (black box) FR systems. In this paper, we introduce a morph generation method that can approximate worst-case morphs even when the FR system is not known. A key contribution is that we include the goal of generating difficult morphs \emph{during} training. Our method is based on Adversarially Learned Inference (ALI) and uses concepts from Wasserstein GANs trained with Gradient Penalty, which were introduced to stabilise the training of GANs. We include these concepts to achieve similar improvement in training stability and call the resulting method Wasserstein ALI (WALI). We finetune WALI using loss functions designed specifically to improve the ability to manipulate identity information in facial images and show how it can generate morphs that are more challenging for FR systems than landmark- or GAN-based morphs. We also show how our findings can be used to improve MIPGAN, an existing StyleGAN-based morph generator.

Una M. Kelly, Luuk J. Spreeuwers, Raymond N. J. Veldhuis

Morphing is a challenge to face recognition (FR) for which several morphing attack detection solutions have been proposed. We argue that face recognition and differential morphing attack detection (D-MAD) in principle perform very similar tasks, which we support by comparing an FR system with two existing D-MAD approaches. We also show that currently used decision thresholds inherently lead to FR systems being vulnerable to morphing attacks and that this explains the tradeoff between performance on normal images and vulnerability to morphing attacks. We propose using FR systems that are already in place for morphing detection and introduce a new evaluation threshold that guarantees an upper limit to the vulnerability to morphing attacks - even of unknown types.

Ioana Mazilu, Shunxin Wang, Sven Dummer et al.

Though modern microscopes have an autofocusing system to ensure optimal focus, out-of-focus images can still occur when cells within the medium are not all in the same focal plane, affecting the image quality for medical diagnosis and analysis of diseases. We propose a method that can deblur images as well as synthesize defocus blur. We train autoencoders with implicit and explicit regularization techniques to enforce linearity relations among the representations of different blur levels in the latent space. This allows for the exploration of different blur levels of an object by linearly interpolating/extrapolating the latent representations of images taken at different focal planes. Compared to existing works, we use a simple architecture to synthesize images with flexible blur levels, leveraging the linear latent space. Our regularized autoencoders can effectively mimic blur and deblur, increasing data variety as a data augmentation technique and improving the quality of microscopic images, which would be beneficial for further processing and analysis.

Zahra Atashgahi, Mykola Pechenizkiy, Raymond Veldhuis et al.

Efficient time series forecasting has become critical for real-world applications, particularly with deep neural networks (DNNs). Efficiency in DNNs can be achieved through sparse connectivity and reducing the model size. However, finding the sparsity level automatically during training remains challenging due to the heterogeneity in the loss-sparsity tradeoffs across the datasets. In this paper, we propose \enquote{\textbf{P}runing with \textbf{A}daptive \textbf{S}parsity \textbf{L}evel} (\textbf{PALS}), to automatically seek a decent balance between loss and sparsity, all without the need for a predefined sparsity level. PALS draws inspiration from sparse training and during-training methods. It introduces the novel "expand" mechanism in training sparse neural networks, allowing the model to dynamically shrink, expand, or remain stable to find a proper sparsity level. In this paper, we focus on achieving efficiency in transformers known for their excellent time series forecasting performance but high computational cost. Nevertheless, PALS can be applied directly to any DNN. To this aim, we demonstrate its effectiveness also on the DLinear model. Experimental results on six benchmark datasets and five state-of-the-art (SOTA) transformer variants show that PALS substantially reduces model size while maintaining comparable performance to the dense model. More interestingly, PALS even outperforms the dense model, in \textcolor{blue}{12} and \textcolor{blue}{14} cases out of 30 cases in terms of MSE and MAE loss, respectively, while reducing \textcolor{blue}{65\%} parameter count and \textcolor{blue}{63\%} FLOPs on average. Our code and supplementary material are available on Github\footnote{\tiny \url{https://github.com/zahraatashgahi/PALS}}.

Shunxin Wang, Raymond Veldhuis, Christoph Brune et al.

The performance of computer vision models are susceptible to unexpected changes in input images caused by sensor errors or extreme imaging environments, known as common corruptions (e.g. noise, blur, illumination changes). These corruptions can significantly hinder the reliability of these models when deployed in real-world scenarios, yet they are often overlooked when testing model generalization and robustness. In this survey, we present a comprehensive overview of methods that improve the robustness of computer vision models against common corruptions. We categorize methods into three groups based on the model components and training methods they target: data augmentation, learning strategies, and network components. We release a unified benchmark framework (available at \url{https://github.com/nis-research/CorruptionBenchCV}) to compare robustness performance across several datasets, and we address the inconsistencies of evaluation practices in the literature. Our experimental analysis highlights the base corruption robustness of popular vision backbones, revealing that corruption robustness does not necessarily scale with model size and data size. Large models gain negligible robustness improvements, considering the increased computational requirements. To achieve generalizable and robust computer vision models, we foresee the need of developing new learning strategies that efficiently exploit limited data and mitigate unreliable learning behaviors.

Abhishek Dutta, Raymond Veldhuis, Luuk Spreeuwers

This paper proposes a data driven model to predict the performance of a face recognition system based on image quality features. We model the relationship between image quality features (e.g. pose, illumination, etc.) and recognition performance measures using a probability density function. To address the issue of limited nature of practical training data inherent in most data driven models, we have developed a Bayesian approach to model the distribution of recognition performance measures in small regions of the quality space. Since the model is based solely on image quality features, it can predict performance even before the actual recognition has taken place. We evaluate the performance predictive capabilities of the proposed model for six face recognition systems (two commercial and four open source) operating on three independent data sets: MultiPIE, FRGC and CAS-PEAL. Our results show that the proposed model can accurately predict performance using an accurate and unbiased Image Quality Assessor (IQA). Furthermore, our experiments highlight the impact of the unaccounted quality space -- the image quality features not considered by IQA -- in contributing to performance prediction errors.

Ahmad Hassanpour, Fatemeh Jamalbafrani, Bian Yang et al.

Face inpainting, the technique of restoring missing or damaged regions in facial images, is pivotal for applications like face recognition in occluded scenarios and image analysis with poor-quality captures. This process not only needs to produce realistic visuals but also preserve individual identity characteristics. The aim of this paper is to inpaint a face given periocular region (eyes-to-face) through a proposed new Generative Adversarial Network (GAN)-based model called Eyes-to-Face Network (E2F-Net). The proposed approach extracts identity and non-identity features from the periocular region using two dedicated encoders have been used. The extracted features are then mapped to the latent space of a pre-trained StyleGAN generator to benefit from its state-of-the-art performance and its rich, diverse and expressive latent space without any additional training. We further improve the StyleGAN output to find the optimal code in the latent space using a new optimization for GAN inversion technique. Our E2F-Net requires a minimum training process reducing the computational complexity as a secondary benefit. Through extensive experiments, we show that our method successfully reconstructs the whole face with high quality, surpassing current techniques, despite significantly less training and supervision efforts. We have generated seven eyes-to-face datasets based on well-known public face datasets for training and verifying our proposed methods. The code and datasets are publicly available.

Marcel Grimmer, Raymond N. J. Veldhuis, Christoph Busch

The recognition performance of biometric systems strongly depends on the quality of the compared biometric samples. Motivated by the goal of establishing a common understanding of face image quality and enabling system interoperability, the committee draft of ISO/IEC 29794-5 introduces expression neutrality as one of many component quality elements affecting recognition performance. In this study, we train classifiers to assess facial expression neutrality using seven datasets. We conduct extensive performance benchmarking to evaluate their classification and face recognition utility prediction abilities. Our experiments reveal significant differences in how each classifier distinguishes "neutral" from "non-neutral" expressions. While Random Forests and AdaBoost classifiers are most suitable for distinguishing neutral from non-neutral facial expressions with high accuracy, they underperform compared to Support Vector Machines in predicting face recognition utility.

Una M. Kelly, Raymond Veldhuis, Luuk Spreeuwers

Face Recognition (FR) systems have been shown to be vulnerable to morphing attacks. We examine exactly how challenging morphs can become. By showing a worst-case construction in the embedding space of an FR system and using a mapping from embedding space back to image space we generate images that show that this theoretical upper bound can be approximated if the FR system is known. The resulting morphs can also succesfully fool unseen FR systems and are useful for exploring and understanding the weaknesses of FR systems. Our method contributes to gaining more insight into the vulnerability of FR systems.

Amina Bassit, Florian Hahn, Joep Peeters et al.

Biometric verification has been widely deployed in current authentication solutions as it proves the physical presence of individuals. To protect the sensitive biometric data in such systems, several solutions have been developed that provide security against honest-but-curious (semi-honest) attackers. However, in practice attackers typically do not act honestly and multiple studies have shown drastic biometric information leakage in such honest-but-curious solutions when considering dishonest, malicious attackers. In this paper, we propose a provably secure biometric verification protocol to withstand malicious attackers and prevent biometric data from any sort of leakage. The proposed protocol is based on a homomorphically encrypted log likelihood-ratio-based (HELR) classifier that supports any biometric modality (e.g. face, fingerprint, dynamic signature, etc.) encoded as a fixed-length real-valued feature vector and performs an accurate and fast biometric recognition. Our protocol, that is secure against malicious adversaries, is designed from a protocol secure against semi-honest adversaries enhanced by zero-knowledge proofs. We evaluate both protocols for various security levels and record a sub-second speed (between $0.37$s and $0.88$s) for the protocol against semi-honest adversaries and between $0.95$s and $2.50$s for the protocol secure against malicious adversaries.

Zahra Atashgahi, Ghada Sokar, Tim van der Lee et al.

Major complications arise from the recent increase in the amount of high-dimensional data, including high computational costs and memory requirements. Feature selection, which identifies the most relevant and informative attributes of a dataset, has been introduced as a solution to this problem. Most of the existing feature selection methods are computationally inefficient; inefficient algorithms lead to high energy consumption, which is not desirable for devices with limited computational and energy resources. In this paper, a novel and flexible method for unsupervised feature selection is proposed. This method, named QuickSelection, introduces the strength of the neuron in sparse neural networks as a criterion to measure the feature importance. This criterion, blended with sparsely connected denoising autoencoders trained with the sparse evolutionary training procedure, derives the importance of all input features simultaneously. We implement QuickSelection in a purely sparse manner as opposed to the typical approach of using a binary mask over connections to simulate sparsity. It results in a considerable speed increase and memory reduction. When tested on several benchmark datasets, including five low-dimensional and three high-dimensional datasets, the proposed method is able to achieve the best trade-off of classification and clustering accuracy, running time, and maximum memory usage, among widely used approaches for feature selection. Besides, our proposed method requires the least amount of energy among the state-of-the-art autoencoder-based feature selection methods.

Dan Zeng, Raymond Veldhuis, Luuk Spreeuwers

The limited capacity to recognize faces under occlusions is a long-standing problem that presents a unique challenge for face recognition systems and even for humans. The problem regarding occlusion is less covered by research when compared to other challenges such as pose variation, different expressions, etc. Nevertheless, occluded face recognition is imperative to exploit the full potential of face recognition for real-world applications. In this paper, we restrict the scope to occluded face recognition. First, we explore what the occlusion problem is and what inherent difficulties can arise. As a part of this review, we introduce face detection under occlusion, a preliminary step in face recognition. Second, we present how existing face recognition methods cope with the occlusion problem and classify them into three categories, which are 1) occlusion robust feature extraction approaches, 2) occlusion aware face recognition approaches, and 3) occlusion recovery based face recognition approaches. Furthermore, we analyze the motivations, innovations, pros and cons, and the performance of representative approaches for comparison. Finally, future challenges and method trends of occluded face recognition are thoroughly discussed.

Kiran Raja, Matteo Ferrara, Annalisa Franco et al.

Morphing attacks have posed a severe threat to Face Recognition System (FRS). Despite the number of advancements reported in recent works, we note serious open issues such as independent benchmarking, generalizability challenges and considerations to age, gender, ethnicity that are inadequately addressed. Morphing Attack Detection (MAD) algorithms often are prone to generalization challenges as they are database dependent. The existing databases, mostly of semi-public nature, lack in diversity in terms of ethnicity, various morphing process and post-processing pipelines. Further, they do not reflect a realistic operational scenario for Automated Border Control (ABC) and do not provide a basis to test MAD on unseen data, in order to benchmark the robustness of algorithms. In this work, we present a new sequestered dataset for facilitating the advancements of MAD where the algorithms can be tested on unseen data in an effort to better generalize. The newly constructed dataset consists of facial images from 150 subjects from various ethnicities, age-groups and both genders. In order to challenge the existing MAD algorithms, the morphed images are with careful subject pre-selection created from the contributing images, and further post-processed to remove morphing artifacts. The images are also printed and scanned to remove all digital cues and to simulate a realistic challenge for MAD algorithms. Further, we present a new online evaluation platform to test algorithms on sequestered data. With the platform we can benchmark the morph detection performance and study the generalization ability. This work also presents a detailed analysis on various subsets of sequestered data and outlines open challenges for future directions in MAD research.

Sunyi Zheng, Ludo J. Cornelissen, Xiaonan Cui et al.

Objective: In clinical practice, small lung nodules can be easily overlooked by radiologists. The paper aims to provide an efficient and accurate detection system for small lung nodules while keeping good performance for large nodules. Methods: We propose a multi-planar detection system using convolutional neural networks. The 2-D convolutional neural network model, U-net++, was trained by axial, coronal, and sagittal slices for the candidate detection task. All possible nodule candidates from the three different planes are combined. For false positive reduction, we apply 3-D multi-scale dense convolutional neural networks to efficiently remove false positive candidates. We use the public LIDC-IDRI dataset which includes 888 CT scans with 1186 nodules annotated by four radiologists. Results: After ten-fold cross-validation, our proposed system achieves a sensitivity of 94.2% with 1.0 false positive/scan and a sensitivity of 96.0% with 2.0 false positives/scan. Although it is difficult to detect small nodules (i.e. < 6 mm), our designed CAD system reaches a sensitivity of 93.4% (95.0%) of these small nodules at an overall false positive rate of 1.0 (2.0) false positives/scan. At the nodule candidate detection stage, results show that a multi-planar method is capable to detect more nodules compared to using a single plane. Conclusion: Our approach achieves good performance not only for small nodules, but also for large lesions on this dataset. This demonstrates the effectiveness and efficiency of our developed CAD system for lung nodule detection. Significance: The proposed system could provide support for radiologists on early detection of lung cancer.

Joep Peeters, Andreas Peter, Raymond N. J. Veldhuis

As applications of biometric verification proliferate, users become more vulnerable to privacy infringement. Biometric data is very privacy sensitive as it may contain information as gender, ethnicity and health conditions which should not be shared with third parties during the verification process. Moreover, biometric data that has fallen into the wrong hands often leads to identity theft. Secure biometric verification schemes try to overcome such privacy threats. Unfortunately, existing secure solutions either introduce a heavy computational or communication overhead or have to accept a high loss in accuracy; both of which make them impractical in real-world settings. This paper presents a novel approach to secure biometric verification aiming at a practical trade-off between efficiency and accuracy, while guaranteeing full security against honest-but-curious adversaries. The system performs verification in the encrypted domain using elliptic curve based homomorphic ElGamal encryption for high efficiency. Classification is based on a log-likelihood ratio classifier which has proven to be very accurate. No private information is leaked during the verification process using a two-party secure protocol. Initial tests show highly accurate results that have been computed within milliseconds range.

Abhishek Dutta, Raymond Veldhuis, Luuk Spreeuwers

In Biometrics, facial uniqueness is commonly inferred from impostor similarity scores. In this paper, we show that such uniqueness measures are highly unstable in the presence of image quality variations like pose, noise and blur. We also experimentally demonstrate the instability of a recently introduced impostor-based uniqueness measure of [Klare and Jain 2013] when subject to poor quality facial images.