150 results
Filters
Settings
Now showing1 - 10 of 150
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Synthetic data generation procedures for domain-specific environments in manufacturingSynthetic data is being used lately for training deep neural networks in computer vision applications such as object detection, object segmentation and 6D object pose estimation. Domain randomization hereby plays an important role in reducing the simulation to reality gap. However, this generalization might not always be effective in specialized domains like manufacturing that involve complex assemblies. Individual parts are integrated in much larger assemblies making them indistinguishable from their counterparts. Moreover, individual parts are often partially occluded in the scene. These situations give rise to wrong detections. Target domain knowledge is vital in these cases and if conceived effectively while generating synthetic data, can show a considerable improvement in bridging the simulation to reality gap. This paper validates synthetic data generation procedures through practical experimentation ensuring that experiments are both comprehensive and reproducible. After combining domain randomization and domain adaptation procedures for parts and assemblies used in manufacturing the model performance improves by up to 15% than the state-of-the-art domain randomization techniques. Reducing the simulation to reality gap in this way can unlock the true potential of robot-assisted production using artificial intelligence.Publicationtype: Conference Paper37 17 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , An Intelligent Pipeline for Localization of Industrial Components in Robotic Manufacturing ApplicationsThe rising skill shortage problem in Europe threatens the economic slowdown in the manufacturing sector. Approaches based on artificial intelligence can play a crucial role in bridging the shortage gap if they can be integrated into robot-assisted production to simplify repetitive manual tasks. Localizing components in a production cell is a familiar problem of robot-assisted production. Robots are often taught trajectories manually, which requires expertise in robot programming. Some of the existing feature-based computer vision solutions can localize a component in 3D space. However, these solutions are not versatile enough to be integrated across different components and production cells. This paper proposes an AI-based solution in the form of a pipeline for the 6D localization of components that can be integrated into multiple industrial use cases. The pipeline encompasses flows for generating synthetic images of components from their CAD model, training deep neural networks to estimate component poses, and improving their accuracy for manufacturing applications. The performance of the pipeline has been validated for components in a production-related environment. The paper also demonstrates the versatility of the pipeline by deploying it for a robotic spray coating use case. Such AI skills can empower the skilled workforce on the shop floor so that they can focus on the overall manufacturing process.Publicationtype: Conference Paper42 41 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , FDIP : a Fast Diffraction Image Processing library for X-ray crystallography experiments(Multidisciplinary Digital Publishing Institute (MDPI), 2024-02-05); ; ; ; Serial crystallography (SX) is a cutting-edge technique in structural biology, involving the systematic collection of X-ray diffraction data from numerous randomly oriented microcrystals. To extract comprehensive three-dimensional information about the studied system, SX utilises thousands of measured diffraction patterns. As such, SX takes advantages of the properties of modern X-ray sources, including Free Electron Lasers (FELs) and third and fourth generation synchrotrons, as well as contemporary high-repetition-rate detectors. Efficient analysis of the extensive datasets generated during SX experiments demands fast and effective algorithms. The FDIP library offers meticulously optimised functions tailored for preprocessing data obtained in SX experiments. This encompasses tasks such as background subtraction, identification and masking of parasitic streaks, elimination of unwanted powder diffraction (e.g., from ice or salt crystals), and pinpointing useful Bragg peaks in each diffraction pattern. The library is equipped with a user-friendly graphical interface for facile parameter adjustment tailored to specific datasets. Compatible with popular SX processing software like OnDA, Cheetah, CrystFEL, and Merge3D, the FDIP library enhances the capabilities of these tools for streamlined and precise serial crystallography analyses.Publicationtype: Journal Article29 97 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , The effects of age on dyspnea and respiratory mechanical and neural responses to exercise in healthy men(2023-08); ; ; Rands, JoshuaThe respiratory muscle pressure generation and inspiratory and expiratory neuromuscular recruitment patterns in younger and older men were compared during exercise, alongside descriptors of dyspnea. Healthy younger (n = 8, 28 ± 5 years) and older (n = 8, 68 ± 4 years) men completed a maximal incremental cycling test. Esophageal, gastric (Pga) and transdiaphragmatic pressures, and electromyography (EMG) of the crural diaphragm were measured using a micro-transducer and EMG catheter. EMG of the parasternal intercostals, sternocleidomastoids, and rectus abdominis were measured using skin surface electrodes. After the exercise test, participants completed a questionnaire to evaluate descriptors of dyspnea. Pga at end-expiration, Pga expiratory tidal swings, and the gastric pressure–time product (PTPga) at absolute and relative minute ventilation were higher (p < 0.05) for older compared to younger men. There were no differences in EMG responses between older and younger men. Younger men were more likely to report shallow breathing (p = 0.005) than older men. Our findings showed younger and older men had similar respiratory neuromuscular activation patterns and reported different dyspnea descriptors, and that older men had greater expiratory muscle pressure generation during exercise. Greater expiratory muscle pressures in older men may be due to compensatory mechanisms designed to offset increasing airway resistance due to aging. These results may have implications for exercise-induced expiratory muscle fatigue in older men.Publicationtype: Journal Article23 35 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , X-ray screening identifies active site and allosteric inhibitors of SARS-CoV-2 main protease(American Association for the Advancement of Science (Assoc.), 2021-04-02); ; ; ; The coronavirus disease (COVID-19) caused by SARS-CoV-2 is creating tremendous human suffering. To date, no effective drug is available to directly treat the disease. In a search for a drug against COVID-19, we have performed a high-throughput X-ray crystallographic screen of two repurposing drug libraries against the SARS-CoV-2 main protease (Mpro), which is essential for viral replication. In contrast to commonly applied X-ray fragment screening experiments with molecules of low complexity, our screen tested alreadyapproved drugs and drugs in clinical trials. From the three-dimensional protein structures, we identified 37 compounds that bind to Mpro. In subsequent cell-based viral reduction assays, one peptidomimetic and six nonpeptidic compounds showed antiviral activity at nontoxic concentrations. We identified two allosteric binding sites representing attractive targets for drug development against SARS-CoV-2.Publicationtype: Journal ArticleScopus© Citations 214 186 279 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , On the use of multilayer Laue lenses with X-ray free electron lasers(SPIE, 2021); ; We report on the use of multilayer Laue lenses to focus the intense X-ray Free Electron Laser (XFEL) beam at the European XFEL to a spot size of a few tens of nanometers. We present the procedure to align and characterize these lenses and discuss challenges working with the pulse trains from this unique X-ray source.Publicationtype: Conference Paper18 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Segmented flow generator for serial crystallography at the European X-ray free electron laser(2020-09-09); ; ; ; Serial femtosecond crystallography (SFX) with X-ray free electron lasers (XFELs) allows structure determination of membrane proteins and time-resolved crystallography. Common liquid sample delivery continuously jets the protein crystal suspension into the path of the XFEL, wasting a vast amount of sample due to the pulsed nature of all current XFEL sources. The European XFEL (EuXFEL) delivers femtosecond (fs) X-ray pulses in trains spaced 100 ms apart whereas pulses within trains are currently separated by 889 ns. Therefore, continuous sample delivery via fast jets wastes >99% of sample. Here, we introduce a microfluidic device delivering crystal laden droplets segmented with an immiscible oil reducing sample waste and demonstrate droplet injection at the EuXFEL compatible with high pressure liquid delivery of an SFX experiment. While achieving ~60% reduction in sample waste, we determine the structure of the enzyme 3-deoxy-D-manno-octulosonate-8-phosphate synthase from microcrystals delivered in droplets revealing distinct structural features not previously reported.Publicationtype: Journal ArticleScopus© Citations 28 278 184 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Optimised preprocessing for automatic mouth gesture classificationMouth gestures are facial expressions in sign language, that do not refer to lip patterns of a spoken language. Research on this topic has been limited so far. The aim of this work is to automatically classify mouth gestures from video material by training a neural network. This could render time-consuming manual annotation unnecessary and help advance the field of automatic sign language translation. However, it is a challenging task due to the little data available as training material and the similarity of different mouth gesture classes. In this paper we focus on the preprocessing of the data, such as finding the area of the face important for mouth gesture recognition. Furthermore we analyse the duration of mouth gestures and determine the optimal length of video clips for classification. Our experiments show, that this can improve the classification results significantly and helps to reach a near human accuracy.Publicationtype: Conference Paper93 92 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Serial protein crystallography in an electron microscope(Nature Publishing Group, 2020-02-21); ; ; ; Serial X-ray crystallography at free-electron lasers allows to solve biomolecular structures from sub-micron-sized crystals. However, beam time at these facilities is scarce, and involved sample delivery techniques are required. On the other hand, rotation electron diffraction (MicroED) has shown great potential as an alternative means for protein nanocrystallography. Here, we present a method for serial electron diffraction of protein nanocrystals combining the benefits of both approaches. In a scanning transmission electron microscope, crystals randomly dispersed on a sample grid are automatically mapped, and a diffraction pattern at fixed orientation is recorded from each at a high acquisition rate. Dose fractionation ensures minimal radiation damage effects. We demonstrate the method by solving the structure of granulovirus occlusion bodies and lysozyme to resolutions of 1.55 Å and 1.80 Å, respectively. Our method promises to provide rapid structure determination for many classes of materials with minimal sample consumption, using readily available instrumentation.Publicationtype: Journal ArticleScopus© Citations 68 234 188 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Time-resolved serial femtosecond crystallography at the European XFEL(2020-01-01); ; ; ; The European XFEL (EuXFEL) is a 3.4-km long X-ray source, which produces femtosecond, ultrabrilliant and spatially coherent X-ray pulses at megahertz (MHz) repetition rates. This X-ray source has been designed to enable the observation of ultrafast processes with near-atomic spatial resolution. Time-resolved crystallographic investigations on biological macromolecules belong to an important class of experiments that explore fundamental and functional structural displacements in these molecules. Due to the unusual MHz X-ray pulse structure at the EuXFEL, these experiments are challenging. Here, we demonstrate how a biological reaction can be followed on ultrafast timescales at the EuXFEL. We investigate the picosecond time range in the photocycle of photoactive yellow protein (PYP) with MHz X-ray pulse rates. We show that difference electron density maps of excellent quality can be obtained. The results connect the previously explored femtosecond PYP dynamics to timescales accessible at synchrotrons. This opens the door to a wide range of time-resolved studies at the EuXFEL.Publicationtype: Journal Article189