Face Editing Using Part-Based Optimization of the Latent Space

We propose an approach for interactive 3D face editing based on deep generative models. Most of the current face modeling methods rely on linear methods and cannot express complex and non-linear deformations. In contrast to 3D morphable face models based on Principal Component Analysis (PCA), we introduce a novel architecture based on variational autoencoders. Our architecture has multiple encoders (one for each part of the face, such as the nose and mouth) which feed a single decoder. As a result, each sub-vector of the latent vector represents one part. We train our model with a novel loss function that further disentangles the space based on different parts of the face. The output of the network is a whole 3D face. Hence, unlike part-based PCA methods, our model learns to merge the parts intrinsically and does not require an additional merging process. To achieve interactive face modeling, we optimize for the latent variables given vertex positional constraints provided by a user. To avoid unwanted global changes elsewhere on the face, we only optimize the subset of the latent vector that corresponds to the part of the face being modified. Our editing optimization converges in less than a second. Our results show that the proposed approach supports a broader range of editing constraints and generates more realistic 3D faces.     

Comparison of the identification performance of conventional FEM updating and integrated DIC

Full‐field identification methods are increasingly used to adequately identify constitutive parameters to describe the mechanical behavior of materials. This paper investigates the more recently introduced one‐step method of integrated digital image correlation (IDIC) with respect to the most commonly used two‐step method of finite element model updating (FEMU), which uses a subset‐based DIC algorithm. To make the comparison as objective as possible, both methods are implemented in the most equivalent manner and use the same FE model. Various virtual test cases are studied to assess the performance of both methods when subjected to different error sources: (1) systematic errors, (2) poor initial guesses for the constitutive parameters, (3) image noise, (4) constitutive model errors, and (5) experimental errors. Results show that, despite the mathematical similarity of both methods, IDIC produces less erroneous and more reliable results than FEMU, particularly for more challenging test cases exhibiting small displacements, complex kinematics, misalignment of the specimen, and image noise. Copyright © 2015 John Wiley & Sons, Ltd.     

Specifying Real-Time Finite-State Systems in Linear Logic (Extended Abstract)

Abstract

Real-time finite-state systems may be specified in linear logic by means of linear implications between conjunctions of fixed finite length. In this setting, where time is treated as a dense linear ordering, safety properties may be expressed as certain provability problems. These provability problems are shown to be in pspace. They are solvable, with some guidance, by finite proof search in concurrent logic programming environments based on linear logic and acting as sort of model-checkers. One advantage of our approach is that either it provides unsafe runs or it actually establishes safety.     

Free-living inferential modeling of blood glucose level using only noninvasive inputs

The goal of this work is to present a causation modeling methodology with the ability to accurately infer blood glucose levels using a large set of highly correlated noninvasive input variables over an extended period of time. These models can provide insight to improve glucose monitoring, and glucose regulation through advanced model-based control technologies. The efficacy of this approach is demonstrated using real data from a type 2 diabetic (T2D) subject collected under free-living conditions over a period of 25 consecutive days. The model was identified and tested using eleven variables that included three food variables as well as several activity and stress variables. The model was trained using 20 days of data and validated using 5 days of data. This gave a fitted correlation coefficient of 0.70 and an average absolute error (AAE) (i.e., the average of the absolute values for the measured glucose concentration minus modeled glucose concentration) of 13.3 mg/dL for the validation data. This AAE result was significantly better than the subject’s personal glucose meter AAE of 15.3 mg/dL for replicated measurements.     

Nonlinear ceramics for tunable microwave devices

The research on materials and systems for tunable microwave devices has gained attraction within the last years. The radio frequency characterization and the component design of tunable microwave components based on dielectric ceramics especially barium-strontium-titanate (BST) are presented in this second part, whereas the basic material properties are discussed in detail in the first part. After a short introduction to the processing technology used for the fabrication of tunable components based on a BST thick film, the relations between microwave properties and material properties as well as the microstructure are presented in detail. The design process for tunable microwave components based on BST thick films is described. Especially the considerations related to micro- and macrostructure and their connection are highlighted. The paper closes with two different application examples: a reconfigurable array antenna for satellite communication and varactors for high power applications.     

Characterization of virus replication

New viruses spread faster than ever and current signature based detection do not protect against these unknown viruses. Behavior based detection is the currently preferred defense against unknown viruses. The drawback of behavior based detection is the ability only to detect specific classes of viruses or have successful detection under certain conditions plus false positives. This paper presents a characterization of virus replication which is the only virus characteristic guaranteed to be consistently present in all viruses. Two detection models based on virus replication are developed, one using operation sequence matching and the other using frequency measures. Regression analysis was generated for both models. A safe list is used to minimize false positives. In our testing using operation sequence matching, over 250 viruses were detected with 43 subsequences. There were minimal false negatives. The replication sequence of just one virus detected 130 viruses, 45% of all tested viruses. Our testing using frequency measures detected all test viruses with no false negatives. The paper shows that virus replication can be identified and used to detect known and unknown viruses.     

Formal Analysis of Multiparty Contract Signing

We analyze the multiparty contract-signing protocols of Garay and MacKenzie (GM) and of Baum and Waidner (BW). We use a finite-state tool, Mocha, which allows specification of protocol properties in a branching-time temporal logic with game semantics. While our analysis does not reveal any errors in the BW protocol, in the GM protocol we discover serious problems with fairness for four signers and an oversight regarding abuse-freeness for three signers. We propose a complete revision of the GM subprotocols in order to restore fairness.     

MERIS potential for ocean colour studies in the open ocean

The interest of space observations of ocean colour for determining variations in phytoplankton distribution and for deriving primary production (via models) has been largely demonstrated by the Coastal Zone Color Scanner (CZCS) which operated from 1978 to 1986. The capabilities of this pioneer sensor, however, were limited both in spectral resolution and radiometric accuracy. The next generation of ocean colour sensors will benefit from major improvements. The Medium Resolution Imaging Spectrometer (MERIS), planned by the European Space Agency (ESA) for the Envisat platform, has been designed to measure radiances in 15 visible and infrared channels. Three infrared channels will allow aerosol characterization, and therefore accurate atmospheric corrections, to be performed for each pixel. For the retrieval of marine parameters, nine channels between 410 and 705nm will be available (as opposed to only four with the CZCS). In coastal waters this should, in principle, allow a separate quantification of different substances (phytoplankton, mineral particles, yellow substance) to be performed. In open ocean waters optically dominated by phytoplankton and their associate detrital matter, the basic information (i.e. the concentration of phytoplanktonic pigments) will be retrieved with improved accuracy due to the increased radiometric performances of MERIS. The adoption of multi-wavelength algorithms could also lead to additional information concerning auxiliary pigments and taxonomic groups. Finally, MERIS will be one of the first sensors to allow measurements of Sun-induced chlorophyll a in vivo fluorescence, which could provide a complementary approach for the assessment of phytoplankton abundance. The development of these next-generation algorithms, however, requires a number of fundamental studies.