Advanced Techniques for Face Recognition under Challenging Environments

Automatically recognizing faces captured under uncontrolled environments has always been a challenging topic in the past decades. In this work, we investigate cohort score normalization that has been widely used in biometric verification as means to improve the robustness of face recognition under challenging environments. In particular, we introduce cohort score normalization into undersampled face recognition problem. Further, we develop an effective cohort normalization method specifically for the unconstrained face pair matching problem. Extensive experiments conducted on several well known face databases demonstrate the effectiveness of cohort normalization on these challenging scenarios. In addition, to give a proper understanding of cohort behavior, we study the impact of the number and quality of cohort samples on the normalization performance. The experimental results show that bigger cohort set size gives more stable and often better results to a point before the performance saturates. And cohort samples with different quality indeed produce different cohort normalization performance. Recognizing faces gone after alterations is another challenging problem for current face recognition algorithms. Face image alterations can be roughly classified into two categories: unintentional (e.g., geometrics transformations introduced by the acquisition devide) and intentional alterations (e.g., plastic surgery). We study the impact of these alterations on face recognition accuracy. Our results show that state-of-the-art algorithms are able to overcome limited digital alterations but are sensitive to more relevant modifications. Further, we develop two useful descriptors for detecting those alterations which can significantly affect the recognition performance. In the end, we propose to use the Structural Similarity (SSIM) quality map to detect and model variations due to plastic surgeries. Extensive experiments conducted on a plastic surgery face database demonstrate the potential of SSIM map for matching face images after surgeries.

Multimethodological study of molecular recognition phenomena

The study of the bio-recognition phenomena behind a biological process is nowadays considered a useful tool to deeply understand physiological mechanisms allowing the discovery of novel biological target and the development of new lead candidates. Moreover, understanding this kind of phenomena can be helpful in characterizing absorption, distribution, metabolism, elimination and toxicity properties of a new drug (ADMET parameters). Recent estimations show that about half of all drugs in development fail to make it to the market because of ADMET deficiencies; thus a rapid determination of ADMET parameters in early stages of drug discovery would save money and time, allowing to choose the better compound and to eliminate any losers. The monitoring of drug binding to plasma proteins is becoming essential in the field of drug discovery to characterize the drug distribution in human body. Human serum albumin (HSA) is the most abundant protein in plasma playing a fundamental role in the transport of drugs, metabolites and endogenous factors; so the study of the binding mechanism to HSA has become crucial to the early characterization of the pharmacokinetic profile of new potential leads. Furthermore, most of the distribution experiments carried out in vivo are performed on animals. Hence it is interesting to determine the binding of new compounds to albumins from different species to evaluate the reliability of extrapolating the distribution data obtained in animals to humans. It is clear how the characterization of interactions between proteins and drugs determines a growing need of methodologies to study any specific molecular event. A wide variety of biochemical techniques have been applied to this purpose. High-performance liquid affinity chromatography, circular dichroism and optical biosensor represent three techniques that can be able to elucidate the interaction of a new drug with its target and with others proteins that could interfere with ADMET parameters.

Eye-trauma: Event driven control software for gesture recognition in medical simulation

Il progetto Eye-Trauma si colloca all'interno dello sviluppo di un simulatore chirurgico per traumi alla zona oculare, sviluppato in collaborazione con Simulation Group in Boston, Harvard Medical School e Massachusetts General Hospital. Il simulatore presenta un busto in silicone fornito di moduli intercambiabili della zona oculare, per simulare diversi tipi di trauma. L'utilizzatore è chiamato ad eseguire la procedura medica di saturazione tramite degli strumenti chirurgici su cui sono installati dei sensori di forza e di apertura. I dati collezionati vengono utilizzati all'interno del software per il riconoscimento dei gesti e il controllo real-time della performance. L'algoritmo di gesture recognition, da me sviluppato, si basa sul concetto di macchine a stati; la transizione tra gli stati avviene in base agli eventi rilevati dal simulatore.

Template-assisted patterning of functional polymers

In this thesis, anodic aluminum oxide (AAO) membranes, which provide well-aligned uniform mesoscopic pores with adjustable pore parameters, were fabricated and successfully utilized as templates for the fabrication of functional organic nanowires, nanorods and the respective well-ordered arrays. The template-assisted patterning technique was successfully applied for the realization of different objectives:rnHigh-density and well-ordered arrays of hole-conducting nanorods composed of cross-linked triphenylamine (TPA) and tetraphenylbenzidine (TPD) derivatives on conductive substrates like ITO/glass have been successfully fabricated. By applying a freeze-drying technique to remove the aqueous medium after the wet-chemical etching of the template, aggregation and collapsing of the rods was prevented and macroscopic areas of perfectly freestanding nanorods were feasible. Based on the hole-conducting nanorod arrays and their subsequent embedding into an electron-conducting polymer matrix via spin-coating, a novel routine concept for the fabrication of well-ordered all-organic bulk heterojunction for organic photovoltaic applications was successfully demonstrated. The increased donor/acceptor interface of the fabricated devices resulted in a remarkable increase of the photoluminescence quenching compared to a planar bilayer morphology. Further, the fundamental working principle of the templating approach for the solution-based all-organic photovoltaic device was demonstrated for the first time.rnFurthermore, in order to broaden the applicability of patterned surfaces, which are feasible via the template-based patterning of functional materials, AAO with hierarchically branched pores were fabricated and utilized as templates. By pursuing the common templating process hierarchically polymeric replicas, which show remarkable similarities with interesting biostructures, like the surface of the lotus leaf and the feet of a gecko, were successfully prepared.rnIn contrast to the direct infiltration of organic functional materials, a novel route for the fabrication of functional nanowires via post-modification of reactive nanowires was established. Therefore, reactive nanowires based on cross-linked pentafluorophenylesters were fabricated by utilizing AAO templates. The post-modification with fluorescent dyes was demonstrated. Furthermore, reactive wires were converted into well-dispersed poly(N-isopropylacrylamide) (PNIPAM) hydrogels, which exhibit a thermal-responsive reversible phase transition. The reversible thermal-responsible swelling of the PNIPAM nanowires exhibited a more than 50 % extended length than in the collapsed PNIPAM state. rnLast but not least, the shape-anisotropic pores of AAO were utilized to uniformly align the mesogens of a nematic liquid crystalline elastomer. Liquid crystalline nanowires with a narrow size distribution and uniform orientation of the liquid crystalline material were fabricated. It was shown that during the transition from the nematic to the isotropic phase the rod’s length shortened by roughly 40 percent. As such these liquid crystalline elastomeric nanowires may find application, as wire-shaped nanoactuators in various fields of research, like lab-on-chip systems, micro fluidics and biomimetics.rn

IDACardOCR: App Android per la gestione di template di documenti di identità e l'acquisizione di dati anagrafici

Lo scopo di IDCardOCR è quello di permettere all'utente di acquisire i dati di una persona in formato strutturato eseguendo una semplice foto del suo documento di identità. I dati salvati saranno così a disposizione dell’utente in formato digitale. Permette, a differenze di altre applicazioni, la creazione di una maschera personalizzata dove è l’utente a decidere i dati da digitalizzare. Il riconoscimento ottico dei caratteri è implementato dal tool Tesseract-OCR Obiettivo della tesi è quello di approfondire lo studio dei sistemi di OCR e del loro funzionamento in ambiente Android, applicare tecniche di ritaglio e elaborazione delle immagini, approfondimento delle piattaforma di sviluppo. La tesi è cosi strutturata. Il primo capitolo contiene l’introduzione al problema, una panoramica sulle applicazioni esistenti e i motivi che ci hanno spinto ha scegliere la piattaforma di sviluppo Andriod. Il capitolo due tratta delle progettazione dell’applicazione attraverso diagrammi UML e introduce l’architettura del sistema. Il capitolo tre è incentrato sulla parte di sviluppo, infine nel quarto capitolo, si conclude la tesi con i risultati dei test effettuati e i possibili sviluppi futuri.

Deep learning for computer vision: a comparison between convolutional neural networks and hierarchical temporal memories on object recognition tasks

In recent years, Deep Learning techniques have shown to perform well on a large variety of problems both in Computer Vision and Natural Language Processing, reaching and often surpassing the state of the art on many tasks. The rise of deep learning is also revolutionizing the entire field of Machine Learning and Pattern Recognition pushing forward the concepts of automatic feature extraction and unsupervised learning in general. However, despite the strong success both in science and business, deep learning has its own limitations. It is often questioned if such techniques are only some kind of brute-force statistical approaches and if they can only work in the context of High Performance Computing with tons of data. Another important question is whether they are really biologically inspired, as claimed in certain cases, and if they can scale well in terms of "intelligence". The dissertation is focused on trying to answer these key questions in the context of Computer Vision and, in particular, Object Recognition, a task that has been heavily revolutionized by recent advances in the field. Practically speaking, these answers are based on an exhaustive comparison between two, very different, deep learning techniques on the aforementioned task: Convolutional Neural Network (CNN) and Hierarchical Temporal memory (HTM). They stand for two different approaches and points of view within the big hat of deep learning and are the best choices to understand and point out strengths and weaknesses of each of them. CNN is considered one of the most classic and powerful supervised methods used today in machine learning and pattern recognition, especially in object recognition. CNNs are well received and accepted by the scientific community and are already deployed in large corporation like Google and Facebook for solving face recognition and image auto-tagging problems. HTM, on the other hand, is known as a new emerging paradigm and a new meanly-unsupervised method, that is more biologically inspired. It tries to gain more insights from the computational neuroscience community in order to incorporate concepts like time, context and attention during the learning process which are typical of the human brain. In the end, the thesis is supposed to prove that in certain cases, with a lower quantity of data, HTM can outperform CNN.

Processing-dependent and -independent pathways for recognition of iodinated contrast media by specific human T cells

One to three percent of patients exposed to intravenously injected iodinated contrast media (CM) develop delayed hypersensitivity reactions. Positive patch test reactions, immunohistological findings, and CM-specific proliferation of T cells in vitro suggest a pathogenetic role for T cells. We have previously demonstrated that CM-specific T cell clones (TCCs) show a broad range of cross-reactivity to different CM. However, the mechanism of specific CM recognition by T cell receptors (TCRs) has not been analysed so far.

Quantitative 1-Step DNA Methylation Analysis with Native Genomic DNA as Template

DNA methylation analysis currently requires complex multistep procedures based on bisulfite conversion of unmethylated cytosines or on methylation-sensitive endonucleases. To facilitate DNA methylation analysis, we have developed a quantitative 1-step assay for DNA methylation analysis.

Is there a recognition memory deficit in Parkinson's disease? Evidence from estimates of recollection and familiarity

There is conflicting evidence whether Parkinson's disease (PD) is associated with impaired recognition memory and which of its underlying processes, namely recollection and familiarity, is more affected by the disease. The present study explored the contribution of recollection and familiarity to verbal recognition memory performance in 14 nondemented PD patients and a healthy control group with two different methods: (i) the word-frequency mirror effect, and (ii) Remember/Know judgments. Overall, recognition memory of patients was intact. The word-frequency mirror effect was observed both in patients and controls: Hit rates were higher and false alarm rates were lower for low-frequency compared to high-frequency words. However, Remember/Know judgments indicated normal recollection, but impaired familiarity. Our findings suggest that mild to moderate PD patients are selectively impaired at familiarity whereas recollection and overall recognition memory are intact.

Anatomically-Driven Soft-Tissue Simulation Strategy for Cranio-Maxillofacial Surgery Using Facial Muscle Template Model

We propose a computationally efficient and biomechanically relevant soft-tissue simulation method for cranio-maxillofacial (CMF) surgery. A template-based facial muscle reconstruction was introduced to minimize the efforts on preparing a patient-specific model. A transversely isotropic mass-tensor model (MTM) was adopted to realize the effect of directional property of facial muscles in reasonable computation time. Additionally, sliding contact around teeth and mucosa was considered for more realistic simulation. Retrospective validation study with postoperative scan of a real patient showed that there were considerable improvements in simulation accuracy by incorporating template-based facial muscle anatomy and sliding contact.

[Molecular analyses for (early) recognition of hematologic diseases - sense and sensibility for molecular analysis: the art of intelligent decision-making]

During the last 10 years several molecular markers have been established as useful tools among the armamentarium of a hematologist. As a consequence, the number of performed hematologic molecular analyses has immensely increased. Often, such tests replace or complement other laboratory methods. Molecular markers can be useful in many ways: they can serve for diagnostics, describe the prognostic profile, predict which types of drugs are indicated, and can be used for the therapeutic monitoring of the patient to indicate an adequate response or predict resistance or relapse of the disease. Many markers fulfill more than one of these aspects. Most important, however, is the right choice of analyses at the right time-points!

Atomistic simulations of 2D bicomponent self-assembly: from molecular recognition to self-healing

Supramolecular two-dimensional engineering epitomizes the design of complex molecular architectures through recognition events in multicomponent self-assembly. Despite being the subject of in-depth experimental studies, such articulated phenomena have not been yet elucidated in time and space with atomic precision. Here we use atomistic molecular dynamics to simulate the recognition of complementary hydrogen-bonding modules forming 2D porous networks on graphite. We describe the transition path from the melt to the crystalline hexagonal phase and show that self-assembly proceeds through a series of intermediate states featuring a plethora of polygonal types. Finally, we design a novel bicomponent system possessing kinetically improved self-healing ability in silico, thus demonstrating that a priori engineering of 2D self-assembly is possible.

Mapping of atrial tachycardias after catheter ablation for atrial fibrillation: use of bi-atrial activation patterns to facilitate recognition of origin

Mapping and ablation of atrial tachycardias (ATs) secondary to catheter ablation of atrial fibrillation (AF) is often challenging due to the complex atrial substrate, different AT mechanisms, and potential origin not only in the left atrium (LA) but also from the right atrium (RA) and the adjacent thoracic veins.