Abstract
This work presents the 2nd Cross-Spectrum Iris/Periocular Recognition Competition (Cross-Eyed2017). The main goal of the competition is to promote and evaluate advances in cross-spectrum iris and periocular recognition. This second edition registered an increase in the participation numbers ranging from academia to industry: five teams submitted twelve methods for the periocular task and five for the iris task. The benchmark dataset is an enlarged version of the dual-spectrum database containing both iris and periocular images synchronously captured from a distance and within a realistic indoor environment. The evaluation was performed on an undisclosed test-set. Methodology, tested algorithms, and obtained results are reported in this paper identifying the remaining challenges in path forward.

Abstract
The spread of biometric applications in mobile devices handled by untrained users opened the door to sources of noise in mobile iris recognition such as larger extent of rotation in the capture and more off-angle imagery not found so extensively in more constrained acquisition settings. As a result of the limitations of the methods in handling such large degrees of freedom there is often an increase in segmentation errors. In this work, a new near-infrared iris dataset captured with a mobile device is evaluated to analyse, in particular, the rotation observed in images and its impact on segmentation and biometric recognition accuracy. For this study a (manually annotated) ground truth segmentation was used which will be published in tandem with the paper. Similarly to most research challenges in biometrics and computer vision in general, deep learning techniques are proving to outperform classical methods in segmentation methods. The utilization of parameterized CNN-based iris segmentations in biometric recognition is a new but promising field. The results presented show how this CNN-based approach outperformed the segmentation traditional methods with respect to overall recognition accuracy for the dataset under investigation.

Abstract
This work presents a novel multimodal database comprising 3D face, 2D face, thermal face, visible iris, finger and hand veins, voice and anthropometrics. This dataset will constitute a valuable resource to the field with its number and variety of biometric traits. Acquired in the context of the EU PROTECT project, the dataset allows several combinations of biometric traits and envisages applications such as border control. Based upon the results of the unimodal data, a fusion scheme was applied to ascertain the recognition potential of combining these biometric traits in a multimodal approach. Due to the variability on the discriminative power of the traits, a leave the n-best out fusion technique was applied to obtain different recognition results. © 2018 Gesellschaft fuer Informatik.

Abstract
Presentation attack detection (PAD) methods are commonly evaluated using metrics based on the predicted labels. This is a limitation, especially for more elusive methods based on deep learning which can freely learn the most suitable features. Though often being more accurate, these models operate as complex black boxes which makes the inner processes that sustain their predictions still baffling. Interpretability tools are now being used to delve deeper into the operation of machine learning methods, especially artificial networks, to better understand how they reach their decisions. In this paper, we make a case for the integration of interpretability tools in the evaluation of PAD. A simple model for face PAD, based on convolutional neural networks, was implemented and evaluated using both traditional metrics (APCER, BPCER and EER) and interpretability tools (Grad-CAM), using data from the ROSE Youtu video collection. The results show that interpretability tools can capture more completely the intricate behavior of the implemented model, and enable the identification of certain properties that should be verified by a PAD method that is robust, coherent, meaningful, and can adequately generalize to unseen data and attacks. One can conclude that, with further efforts devoted towards higher objectivity in interpretability, this can be the key to obtain deeper and more thorough PAD performance evaluation setups.

Abstract
Fingerprint presentation attack detection (PAD) methods present a stunning performance in current literature. However, the fingerprint PAD generalisation problem is still an open challenge requiring the development of methods able to cope with sophisticated and unseen attacks as our eventual intruders become more capable. This work addresses this problem by applying a regularisation technique based on an adversarial training and representation learning specifically designed to to improve the PAD generalisation capacity of the model to an unseen attack. In the adopted approach, the model jointly learns the representation and the classifier from the data, while explicitly imposing invariance in the high-level representations regarding the type of attacks for a robust PAD. The application of the adversarial training methodology is evaluated in two different scenarios: i) a handcrafted feature extraction method combined with a Multilayer Perceptron (MLP); and ii) an end-to-end solution using a Convolutional Neural Network (CNN). The experimental results demonstrated that the adopted regularisation strategies equipped the neural networks with increased PAD robustness. The adversarial approach particularly improved the CNN models' capacity for attacks detection in the unseen-attack scenario, showing remarkable improved APCER error rates when compared to state-of-the-art methods in similar conditions.

Abstract
In recent years, deep neural networks have shown significant progress in computer vision due to their large generalization capacity; however, the overfitting problem ubiquitously threatens the learning process of these highly nonlinear architectures. Dropout is a recent solution to mitigate overfitting that has witnessed significant success in various classification applications. Recently, many efforts have been made to improve the Standard dropout using an unsupervised merit-based semantic selection of neurons in the latent space. However, these studies do not consider the task-relevant information quality and quantity and the diversity of the latent kernels. To solve the challenge of dropping less informative neurons in deep learning, we propose an efficient end-to-end dropout algorithm that selects the most informative neurons with the highest correlation with the target output considering the sparsity in its selection procedure. First, to promote activation diversity, we devise an approach to select the most diverse set of neurons by making use of determinantal point process (DPP) sampling. Furthermore, to incorporate task specificity into deep latent features, a mutual information (MI)-based merit function is developed. Leveraging the proposed MI with DPP sampling, we introduce the novel DPPMI dropout that adaptively adjusts the retention rate of neurons based on their contribution to the neural network task. Empirical studies on real-world classification benchmarks including, MNIST, SVHN, CIFAR10, CIFAR100, demonstrate the superiority of our proposed method over recent state-of-the-art dropout algorithms in the literature.