Explore the vast range of titles available.

The success of deep learning in computer vision is based on the availability of large annotated datasets. To lower the need for hand labeled images, virtually rendered 3D worlds have recently gained popularity. Unfortunately, creating realistic 3D content is challenging on its own and requires significant human effort. In this work, we propose an alternative paradigm which combines real and synthetic data for learning semantic instance segmentation and object detection models. Exploiting the fact that not all aspects of the scene are equally important for this task, we propose to augment real-world imagery with virtual objects of the target category. Capturing real-world images at large scale is easy and cheap, and directly provides real background appearances without the need for creating complex 3D models of the environment. We present an efficient procedure to augment these images with virtual objects. In contrast to modeling complete 3D environments, our data augmentation approach requires only a few user interactions in combination with 3D models of the target object category. Leveraging our approach, we introduce a novel dataset of augmented urban driving scenes with 360 degree images that are used as environment maps to create realistic lighting and reflections on rendered objects. We analyze the significance of realistic object placement by comparing manual placement by humans to automatic methods based on semantic scene analysis. This allows us to create composite images which exhibit both realistic background appearance as well as a large number of complex object arrangements. Through an extensive set of experiments, we conclude the right set of parameters to produce augmented data which can maximally enhance the performance of instance segmentation models. Further, we demonstrate the utility of the proposed approach on training standard deep models for semantic instance segmentation and object detection of cars in outdoor driving scenarios. We test the models trained on our augmented data on the KITTI 2015 dataset, which we have annotated with pixel-accurate ground truth, and on the Cityscapes dataset. Our experiments demonstrate that the models trained on augmented imagery generalize better than those trained on fully synthetic data or models trained on limited amounts of annotated real data.

Multi-object tracking (MOT) has been notoriously difficult to evaluate. Previous metrics overemphasize the importance of either detection or association. To address this, we present a novel MOT evaluation metric, higher order tracking accuracy (HOTA), which explicitly balances the effect of performing accurate detection, association and localization into a single unified metric for comparing trackers. HOTA decomposes into a family of sub-metrics which are able to evaluate each of five basic error types separately, which enables clear analysis of tracking performance. We evaluate the effectiveness of HOTA on the MOTChallenge benchmark, and show that it is able to capture important aspects of MOT performance not previously taken into account by established metrics. Furthermore, we show HOTA scores better align with human visual evaluation of tracking performance.

Jawbone differs from other bones in many aspects, including its developmental origin and the occurrence of jawbone-specific diseases like MRONJ (medication-related osteonecrosis of the jaw). Although there is a strong need, adequate in vitro models of this unique environment are sparse to date. While previous approaches are reliant e.g. on scaffolds or spheroid culture, 3D bioprinting enables free-form fabrication of complex living tissue structures. In the present work, production of human jawbone models was realised via projection-based stereolithography. Constructs were bioprinted containing primary jawbone-derived osteoblasts and vasculature-like channel structures optionally harbouring primary endothelial cells. After 28 days of cultivation in growth medium or osteogenic medium, expression of cell type-specific markers was confirmed on both the RNA and protein level, while prints maintained their overall structure. Survival of endothelial cells in the printed channels, co-cultured with osteoblasts in medium without supplementation of endothelial growth factors, was demonstrated. Constructs showed not only mineralisation, being one of the characteristics of osteoblasts, but also hinted at differentiation to an osteocyte phenotype. These results indicate the successful biofabrication of an in vitro model of the human jawbone, which presents key features of this special bone entity and hence appears promising for application in jawbone-specific research.

Barrier organ models need a scaffold structure to create a two compartment culture. Technical filter membranes used most often as scaffolds may impact cell behaviour and present a barrier themselves, ultimately limiting transferability of test results. In this work we present an alternative for technical filter membrane systems: a 3D bioprinted biological membrane in 24 well format. The biological membrane, based on extracellular matrix (ECM), is highly permeable and presents a natural 3D environment for cell culture. Inspired by the human placenta we established a coculture of a trophoblast-derived cell line (BeWo b30), together with primary placental fibroblasts within the biological membrane (simulating villous stroma) and primary human placental endothelial cells—representing three cellular components of the human placental villus. All cell types maintained their cell type specific marker expression after two weeks of coculture on the biological membrane. In permeability assays the trophoblast layer developed a barrier on the biological membrane, which was even more pronounced when cocultured with fibroblasts. In this work we present a filter membrane free scaffold, we characterize its properties and assess its suitability for cell culture and barrier models. Further we show a novel placenta inspired model in a complex bioprinted coculture. In the absence of an artificial filter membrane, we demonstrate barrier architecture and functionality.

We address the problem of 3D shape completion from sparse and noisy point clouds, a fundamental problem in computer vision and robotics. Recent approaches are either data-driven or learning-based: Data-driven approaches rely on a shape model whose parameters are optimized to fit the observations; Learning-based approaches, in contrast, avoid the expensive optimization step by learning to directly predict complete shapes from incomplete observations in a fully-supervised setting. However, full supervision is often not available in practice. In this work, we propose a weakly-supervised learning-based approach to 3D shape completion which neither requires slow optimization nor direct supervision. While we also learn a shape prior on synthetic data, we amortize, i.e., learn, maximum likelihood fitting using deep neural networks resulting in efficient shape completion without sacrificing accuracy. On synthetic benchmarks based on ShapeNet (Chang et al. Shapenet: an information-rich 3d model repository, 2015. arXiv:1512.03012) and ModelNet (Wu et al., in: Proceedings of IEEE conference on computer vision and pattern recognition (CVPR), 2015) as well as on real robotics data from KITTI (Geiger et al., in: Proceedings of IEEE conference on computer vision and pattern recognition (CVPR), 2012) and Kinect (Yang et al., 3d object dense reconstruction from a single depth view, 2018. arXiv:1802.00411), we demonstrate that the proposed amortized maximum likelihood approach is able to compete with the fully supervised baseline of Dai et al. (in: Proceedings of IEEE conference on computer vision and pattern recognition (CVPR), 2017) and outperforms the data-driven approach of Engelmann et al. (in: Proceedings of the German conference on pattern recognition (GCPR), 2016), while requiring less supervision and being significantly faster.

The Production Routing Problem under demand uncertainty is an integrated problem containing production, inventory, and distribution decisions. At the planning level, the aim is to meet retailers demand, when only the demand distribution is known in advance, while minimizing the corresponding costs. In this study, a two-stage formulation is presented in which the routing can be adjusted at short notice. In the first stage, only production decisions are made, while delivery and inventory quantities and retailer visit schedules are determined in the second stage. To handle a large number of scenarios, two solution methods based on Sample Average Approximation are introduced. Furthermore, the impact of the routing quality is explored by applying a simple heuristic and an effective metaheuristic on the routing part. It is shown that, on average, the simple heuristic within an adjustable Sample Average Approximation approach provides better objective function values than the metaheuristic within a non-adjustable approach. Also all solution approaches outperform an expected value based approach in terms of runtime and objective function value.

The Production Routing Problem under demand uncertainty is an integrated problem containing production, inventory, and distribution decisions. At the planning level, the aim is to meet retailers demand, when only the demand distribution is known in advance, while minimizing the corresponding costs. In this study, a two-stage formulation is presented in which the routing can be adjusted at short notice. In the first stage, only production decisions are made, while delivery and inventory quantities and retailer visit schedules are determined in the second stage. To handle a large number of scenarios, two solution methods based on Sample Average Approximation are introduced. Furthermore, the impact of the routing quality is explored by applying a simple heuristic and an effective metaheuristic on the routing part. It is shown that, on average, the simple heuristic within an adjustable Sample Average Approximation approach provides better objective function values than the metaheuristic within a non-adjustable approach. Also all solution approaches outperform an expected value based approach in terms of runtime and objective function value.

Safety workwear often requires antistatic protection to prevent the build-up of static electricity and sparks, which can be extremely dangerous in a working environment. In order to make synthetic antistatic fibers, electrically conducting materials such as carbon black are added to the fiber-forming polymer. This leads to unwanted dark colors in the respective melt-spun fibers. To attenuate the undesired dark color, we looked into various possibilities including the embedding of the conductive element inside a dull side-by-side bicomponent fiber. The bicomponent approach, with an antistatic compound as a minor element, also helped in preventing the severe loss of tenacity often caused by a high additive loading. We could melt-spin a bicomponent fiber with a specific resistance as low as 0.1 Ωm and apply it in a fabric that fulfills the requirements regarding the antistatic properties, luminance and flame retardancy of safety workwear.

Digital human models (DHMs) are virtual representations of human beings. They are used to conduct, among other things, ergonomic assessments in factory layout planning. DHM software tools are challenging in their use and thus require a high amount of training for engineers. In this paper, we present a virtual reality (VR) application that enables engineers to work with DHMs easily. Since VR systems with head-mounted displays (HMDs) are less expensive than CAVE systems, HMDs can be integrated more extensively into the product development process. Our application provides a reality-based interface and allows users to conduct an assembly task in VR and thus to manipulate the virtual scene with their real hands. These manipulations are used as input for the DHM to simulate, on that basis, human ergonomics. Therefore, we introduce a software and hardware architecture, the VATS (virtual action tracking system). This paper furthermore presents the results of a user study in which the VATS was compared to the existing WIMP (Windows, Icons, Menus and Pointer) interface. The results show that the VATS system enables users to conduct tasks in a significantly faster way.

The Neurotransmitter:Sodium Symporters (NSSs) represent an important class of proteins mediating sodium-dependent uptake of neurotransmitters from the extracellular space. The substrate binding stoichiometry of the bacterial NSS protein, LeuT, and thus the principal transport mechanism, has been heavily debated. Here we used solid state NMR to specifically characterize the bound leucine ligand and probe the number of binding sites in LeuT. We were able to produce high-quality NMR spectra of substrate bound to microcrystalline LeuT samples and identify one set of sodium-dependent substrate-specific chemical shifts. Furthermore, our data show that the binding site mutants F253A and L400S, which probe the major S1 binding site and the proposed S2 binding site, respectively, retain sodium-dependent substrate binding in the S1 site similar to the wild-type protein. We conclude that under our experimental conditions there is only one detectable leucine molecule bound to LeuT. All living cells need amino acids – the building blocks of proteins – in order to survive, yet few cells can make all the amino acids that they need. Instead, transporter proteins in cell membranes must take these molecules from the outside of the cell and release them to the inside. Some cells, including those in the brain, also release amino acids and molecules derived from them into the spaces outside of the cell to send signals to other nearby cells. Again, transporter proteins must move these signaling molecules back inside cells, to stop the signaling and to allow the molecules to be recycled. Importantly, problems with these uptake mechanisms have been linked to disorders such as depression, epilepsy and Parkinson’s disease. One family of transporters involved in the uptake of amino acids are the “Neurotransmitter:Sodium Symporters”. Though these proteins are involved in processes that are fundamental to life, it remains unclear exactly how they work. Specifically, it has been heavily debated whether this family of transporters require one or two amino acid molecules to bind at the same time in order to help transport them across the membrane. Now Erlendsson, Gotfryd et al. have analyzed a bacterial protein in the Neurotransmitter:Sodium Symporter family. This transporter takes up an amino acid called leucine into cells, and is commonly used as a model to understand this family of transporter proteins more generally. Using a technique called solid state nuclear magnetic resonance, Erlendsson, Gotfryd et al. could detect a single molecule of leucine bound to each transporter, but not a second one. This technique could also pinpoint that the leucine was located at the transporter’s central binding site. Leucine was never found at the proposed secondary binding site. Together these findings suggest that only one molecule of leucine binds to the transporter at any one time, and that it binds to the transporter’s central binding site. Erlendsson, Gotfryd et al. have shown now how solid state nuclear magnetic resonance can be used to explore in detail how Neurotransmitter:Sodium Symporters move molecules across cell membranes. The next challenge is to use the same experimental setup to characterize other Neurotransmitter:Sodium Symporters. Doing so could potentially lay the groundwork for designing more specific and improved drugs to treat disorders like depression and Parkinson’s disease.