Publikationen

Laser beam butt welding is often the technique of choice for a wide range of industrial tasks. To achieve high quality welds, manufacturers often rely on heavy and expensive clamping systems to limit the sheet movement during the welding process, which can affect quality. Jiggless welding offers a cost-effective and highly flexible alternative to common clamping systems. In laser butt welding, the process-induced joint gap has to be monitored in order to counteract the effect by means of an active position control of the sheet metal. Various studies have shown that sheet metal displacement can be detected using inductive probes, allowing the prediction of weld quality by ML-based data analysis. The probes are dependent on the sheet metal geometry and are limited in their applicability to complex geometric structures. Camera systems such as long-wave infrared (LWIR) cameras can instead be mounted directly behind the laser to overcome a geometry dependent limitation of the jiggles system. In this study we will propose a deep learning approach that utilizes LWIR camera recordings to predict the remaining welding process to enable an early detection of weld interruptions. Our approach reaches 93.33% accuracy for time-wise prediction of the point of failure during the weld.

Future production technologies will comprise a multitude of systems whose core functionality is closely related to machine-learned models. Such systems require reliable components to ensure the safety of workers and their trust in the systems. The evaluation of the functional reliability and resilience of systems based on machine-learned models is generally challenging. For this purpose, appropriate test data must be available, which also includes abnormal cases. These abnormal cases can be unexpected usage scenarios, erroneous inputs, accidents during operation or even the failure of certain subcomponents. In this work, approaches to the model-based generation of an arbitrary abundance of data representing such abnormal cases are explored. Such computer-based generation requires domain-specific approaches, especially with respect to the nature and distribution of the data, protocols used, or domain-specific communication structures. In previous work, we found that different use cases impose different requirements on synthetic data, and the requirements in turn imply different generation methods [1]. Based on this, various use cases are identified and different methods for computer-based generation of realistic data, as well as for the quality assessment of such data, are explored. Ultimately we explore the use of Federated Learning (FL) to address data privacy and security challenges in Industrial Control Systems. FL enables local model training while keeping sensitive information decentralized and private to their owners. In detail, we investigate whether FL can benefit clients with limited knowledge by leveraging collaboratively trained models that aggregate client-specific knowledge distributions. We found that in such scenarios federated training results in a significant increase in classification accuracy by 31.3% compared to isolated local training. Furthermore, as we introduce Differential Privacy, the resulting model achieves on par accuracy of 99.62% to an idealized case where data is independent and identically distributed across clients.

Plant phenology plays a vital role in assessing climate change. To monitor this, individual plants are traditionally visited and observed by trained volunteers organized in national or international networks - in Germany, for example, by the German Weather Service, DWD. However, their number of observers is continuously decreasing. In this study, we explore the feasibility of using opportunistically captured plant observations, collected via the plant identification app Flora Incognita to determine the onset of flowering and, based on that, create interpolation maps comparable to those of the DWD. Therefore, the opportunistic observations of 17 species collected in 2020 and 2021 were assigned to “Flora Incognita stations” based on location and altitude in order to mimic the network of stations forming the data basis for the interpolation conducted by the DWD. From the distribution of observations, the percentile representing onset of flowering date was calculated using a parametric bootstrapping approach and then interpolated following the same process as applied by the DWD. Our results show that for frequently observed, herbaceous and conspicuous species, the patterns of onset of flowering were similar and comparable between both data sources. We argue that a prominent flowering stage is crucial for accurately determining the onset of flowering from opportunistic plant observations, and we discuss additional factors, such as species distribution, location bias and societal events contributing to the differences among species and phenology data. In conclusion, our study demonstrates that the phenological monitoring of certain species can benefit from incorporating opportunistic plant observations. Furthermore, we highlight the potential to expand the taxonomic range of monitored species for phenological stage assessment through opportunistic plant observation data.

Mining change patterns can give unique understanding on the evolution of dynamically changing systems like social relation graphs, weblinks, hardware descriptions and models. A more recent focus is source code change pattern mining that may qualitatively justify expected or uncover unexpected patterns. These patterns then offer a basis, e.g., for program language evolution or auto-completion support. We present a change pattern mining method that greatly expands the limits of input data and pattern complexity, over existing methods. We propose scalability solutions on conceptual and algorithmic level, thereby evolving the state-of-the-art sampling-based frequent subgraph mining method FS3, resulting in 75% reduction in memory consumption and a speedup of 6500 for a large scale dataset. Patterns can have 100,000 s of occurrences for which manual review is impossible and may lead to misinterpretation. We propose the novel content track approach for interactively exploring pattern contents in context, based on marginal distributions. We evaluate our approach by mining 1,000 open source projects contributing a total of 558 million changes and 2 billion contextual connections among them, thereby, demonstrating its scalability. A manual interpretation of 19 patterns shows sensible mined patterns allowing to deduct implications for language design and demonstrating the soundness of the approach.

Background: The Progress Test Medizin (PTM) is a 200-question formative test that is administered to approximately 11,000 students at medical universities (Germany, Austria, Switzerland) each term. Students receive feedback on their knowledge (development) mostly in comparison to their own cohort. In this study, we use the data of the PTM to find groups with similar response patterns. Methods: We performed k-means clustering with a dataset of 5,444 students, selected cluster number k = 5, and answers as features. Subsequently, the data was passed to XGBoost with the cluster assignment as target enabling the identification of cluster-relevant questions for each cluster with SHAP. Clusters were examined by total scores, response patterns, and confidence level. Relevant questions were evaluated for difficulty index, discriminatory index, and competence levels. Results: Three of the five clusters can be seen as “performance” clusters: cluster 0 (n = 761) consisted predominantly of students close to graduation. Relevant questions tend to be difficult, but students answered confidently and correctly. Students in cluster 1 (n = 1,357) were advanced, cluster 3 (n = 1,453) consisted mainly of beginners. Relevant questions for these clusters were rather easy. The number of guessed answers increased. There were two “drop-out” clusters: students in cluster 2 (n = 384) dropped out of the test about halfway through after initially performing well; cluster 4 (n = 1,489) included students from the first semesters as well as “non-serious” students both with mostly incorrect guesses or no answers. Conclusion: Clusters placed performance in the context of participating universities. Relevant questions served as good cluster separators and further supported our “performance” cluster groupings.

Analyzing time series data like EEG or MEG is challenging due to noisy, high-dimensional, and patient-specific signals. Deep learning methods have been demonstrated to be superior in analyzing time series data compared to shallow learning methods which utilize handcrafted and often subjective features. Especially, recurrent deep neural networks (RNN) are considered suitable to analyze such continuous data. However, previous studies show that they are computationally expensive and difficult to train. In contrast, feed-forward networks (FFN) have previously mostly been considered in combination with hand-crafted and problem-specific feature extractions, such as short time Fourier and discrete wavelet transform. A sought-after are easily applicable methods that efficiently analyze raw data to remove the need for problem-specific adaptations. In this work, we systematically compare RNN and FFN topologies as well as advanced architectural concepts on multiple datasets with the same data preprocessing pipeline. We examine the behavior of those approaches to provide an update and guideline for researchers who deal with automated analysis of EEG time series data. To ensure that the results are meaningful, it is important to compare the presented approaches while keeping the same experimental setup, which to our knowledge was never done before. This paper is a first step toward a fairer comparison of different methodologies with EEG time series data. Our results indicate that a recurrent LSTM architecture with attention performs best on less complex tasks, while the temporal convolutional network (TCN) outperforms all the recurrent architectures on the most complex dataset yielding a 8.61% accuracy improvement. In general, we found the attention mechanism to substantially improve classification results of RNNs. Toward a light-weight and online learning-ready approach, we found extreme learning machines (ELM) to yield comparable results for the less complex tasks.

Prediction of multi-dimensional time-series data, which may represent such diverse phenomena as climate changes or financial markets, remains a challenging task in view of inherent nonlinearities and non-periodic behavior In contrast to other recurrent neural networks, echo state networks (ESNs) are attractive for (online) learning due to lower requirements w.r.t.training data and computational power. However, the randomly-generated reservoir renders the choice of suitable hyper-parameters as an open research topic. We systematically derive and exemplarily demonstrate design guidelines for the hyper-parameter optimization of ESNs. For the evaluation, we focus on the prediction of chaotic time series, an especially challenging problem in machine learning. Our findings demonstrate the power of a hyper-parameter-tuned ESN when auto-regressively predicting time series over several hundred steps. We found that ESNs’ performance improved by 85.1%-99.8% over an already wisely chosen default parameter initialization. In addition, the fluctuation range is considerably reduced such that significantly worse performance becomes very unlikely across random reservoir seeds. Moreover, we report individual findings per hyper-parameter partly contradicting common knowledge to further, help researchers when training new models.

Transformer networks such as CodeBERT already achieve outstanding results for code clone detection in benchmark datasets, so one could assume that this task has already been solved. However, code clone detection is not a trivial task. Semantic code clones, in particular, are challenging to detect. We show that the generalizability of CodeBERT decreases by evaluating two different subsets of Java code clones from BigCloneBench. We observe a significant drop in F1 score when we evaluate different code snippets and functionality IDs than those used for model building.

The prediction of soil moisture for automated irrigation applications is a major challenge, as it is affected by various environmental parameters. The Application of Convolutional Neural Networks (CNN), to this end, has shown remarkable results for soil moisture prediction. These models, however, typically need large datasets, which are scarce in the agriculture field. To this end, this paper presents a Deep Convolutional Generative Adversarial Network (DCGAN) that can learn good data representations and generate highly realistic samples. Traditionally, Generative Adversarial Networks (GANs) have been used for generating data for segmentation and classification tasks or used in conjunction with CNNs or Multi Layer Perceptrons (MLPs) for regression tasks. In this paper, we propose a novel approach in which GANs are used to generate conjointly training images for plants as well as realistic regression values for their corresponding moisture levels without the use of any additional network. The generated images and regression vector targets, together with the training data, are then used to train a CNN which is then evaluated with actual test data from the dataset. We observe an improvement of error rate by 33 percent which shows the validity of our approach.

In this paper, a deep learning approach is presented for direction of arrival estimation using automotive-grade ultrasonic sensors which are used for driving assistance systems such as automatic parking. A study and implementation of the state of the art deterministic direction of arrival estimation algorithms is used as a benchmark for the performance of the proposed approach. Analysis of the performance of the proposed algorithms against the existing algorithms is carried out over simulation data as well as data from a measurement campaign done using automotive-grade ultrasonic sensors. Both sets of results clearly show the superiority of the proposed approach under realistic conditions such as noise from the environment as well as eventual errors in measurements. It is demonstrated as well how the proposed approach can overcome some of the known limitations of the existing algorithms such as precision dilution of triangulation and aliasing.

This work presents the application of droplet-based microfluidics for the cultivation of microspores from Brassica napus using the doubled haploid technology. Under stress conditions (e.g. heat shock) or by chemical induction a certain fraction of the microspores can be reprogrammed and androgenesis can be induced. This process is an important approach for plant breeding because desired plant properties can be anchored in the germline on a genetic level. However, the reprogramming rate of the microspores is generally very low, increasing it by specific stimulation is, therefore, both a necessary and challenging task. In order to accelerate the optimisation and development process, the application of droplet-based microfluidics can be a promising tool. Here, we used a tube-based microfluidic system for the generation and cultivation of microspores inside nL-droplets. Different factors like cell density, tube material and heat shock conditions were investigated to improve the yield of vital plant organoids. Evaluation and analysis of the stimuli response were done on an image base aided by an artificial intelligence cell detection algorithm. Droplet-based microfluidics allowed us to apply large concentration programs in small test volumes and to screen the best conditions for reprogramming cells by the histone deacetylase inhibitor trichostatin A and for enhancing the yield of vital microspores in droplets. An enhanced reprogramming rate was found under the heat shock conditions at 32 &ring;C for about 3 to 6 days. In addition, the comparative experiment with MTP showed that droplet cultivation with lower cell density (<10 cells per droplet) or adding media after 3 or 6 days significantly positively affects the microspore growth and embryo rate inside 120 nL droplets. Finally, the developed embryos could be removed from the droplets and further grown into mature plants. Overall, we demonstrated that the droplet-based tube system is suitable for implementation in an automated, miniaturized system to achieve the induction of embryogenic development in haploid microspore stem cells of Brassica napus.

Spiking Neural Networks are a recent and new neural network design approach that promises tremendous improvements in power efficiency, computation efficiency, and processing latency. They do so by using asynchronous spike-based data flow, event-based signal generation, processing, and modifying the neuron model to resemble biological neurons closely. While some initial works have shown significant initial evidence of applicability to common deep learning tasks, their applications in complex real-world tasks have been relatively low. In this work, we first illustrate the applicability of spiking neural networks to a complex deep learning task, namely LiDAR based 3D object detection for automated driving. Secondly, we make a step-by-step demonstration of simulating spiking behavior using a pre-trained Convolutional Neural Network. We closely model essential aspects of spiking neural networks in simulation and achieve equivalent run-time and accuracy on a GPU. We expect significant improvements in power efficiency when the model is implemented on neuromorphic hardware.

Environmental monitoring involves the quantification of microscopic cells and particles such as algae, plant cells, pollen, or fungal spores. Traditional methods using conventional microscopy require expert knowledge, are time-intensive and not well-suited for automated high throughput. Multispectral imaging flow cytometry (MIFC) allows measurement of up to 5000 particles per second from a fluid suspension and can simultaneously capture up to 12 images of every single particle for brightfield and different spectral ranges, with up to 60x magnification. The high throughput of MIFC has high potential for increasing the amount and accuracy of environmental monitoring, such as for plant-pollinator interactions, fossil samples, air, water or food quality that currently rely on manual microscopic methods. Automated recognition of particles and cells is also possible, when MIFC is combined with deep-learning computational techniques. Furthermore, various fluorescence dyes can be used to stain specific parts of the cell to highlight physiological and chemical features including: vitality of pollen or algae, allergen content of individual pollen, surface chemical composition (carbohydrate coating) of cells, DNA- or enzyme-activity staining. Here, we outline the great potential for MIFC in environmental research for a variety of research fields and focal organisms. In addition, we provide best practice recommendations.

A 360&ring; perception of scene geometry is essential for automated driving, notably for parking and urban driving scenarios. Typically, it is achieved using surround-view fisheye cameras, focusing on the near-field area around the vehicle. The majority of current depth estimation approaches focus on employing just a single camera, which cannot be straightforwardly generalized to multiple cameras. The depth estimation model must be tested on a variety of cameras equipped to millions of cars with varying camera geometries. Even within a single car, intrinsics vary due to manufacturing tolerances. Deep learning models are sensitive to these changes, and it is practically infeasible to train and test on each camera variant. As a result, we present novel camera-geometry adaptive multi-scale convolutions which utilize the camera parameters as a conditional input, enabling the model to generalize to previously unseen fisheye cameras. Additionally, we improve the distance estimation by pairwise and patchwise vector-based self-attention encoder networks. We evaluate our approach on the Fisheye WoodScape surround-view dataset, significantly improving over previous approaches. We also show a generalization of our approach across different camera viewing angles and perform extensive experiments to support our contributions. To enable comparison with other approaches, we evaluate the front camera data on the KITTI dataset (pinhole camera images) and achieve state-of-the-art performance among self-supervised monocular methods. An overview video with qualitative results is provided at https://youtu.be/bmX0UcU9wtA. Baseline code and dataset will be made public.

Laser beam butt welding of thin sheets of high-alloy steel can be really challenging due to the formation of joint gaps, affecting weld seam quality. Industrial approaches rely on massive clamping systems to limit joint gap formation. However, those systems have to be adapted for each individually component geometry, making them very cost-intensive and leading to a limited flexibility. In contrast, jigless welding can be a high flexible alternative to substitute conventionally used clamping systems. Based on the collaboration of different actuators, motions systems or robots, the approach allows an almost free workpiece positioning. As a result, jigless welding gives the possibility for influencing the formation of the joint gap by realizing an active position control. However, the realization of an active position control requires an early and reliable error prediction to counteract the formation of joint gaps during laser beam welding. This paper proposes different approaches to predict the formation of joint gaps and gap induced weld discontinuities in terms of lack of fusion based on optical and tactile sensor data. Our approach achieves 97.4 % accuracy for video-based weld discontinuity detection and a mean absolute error of 0.02 mm to predict the formation of joint gaps based on tactile length measurements by using inductive probes.

A combined convolutional autoencoder-recurrent neural network machine learning model is presented to directly analyze and forecast the dynamics and low-order statistics of the local convective heat flux field in a two-dimensional turbulent Rayleigh-Bénard convection flow at Prandtl number Pr=7 and Rayleigh number Ra=10^7. Two recurrent neural networks are applied for the temporal advancement of turbulent heat transfer data in the reduced latent data space, an echo state network, and a recurrent gated unit. Thereby, our work exploits the modular combination of three different machine learning algorithms to build a fully data-driven and reduced model for the dynamics of the turbulent heat transfer in a complex thermally driven flow. The convolutional autoencoder with 12 hidden layers is able to reduce the dimensionality of the turbulence data to about 0.2% of their original size. Our results indicate a fairly good accuracy in the first- and second-order statistics of the convective heat flux. The algorithm is also able to reproduce the intermittent plume-mixing dynamics at the upper edges of the thermal boundary layers with some deviations. The same holds for the probability density function of the local convective heat flux with differences in the far tails. Furthermore, we demonstrate the noise resilience of the framework. This suggests that the present model might be applicable as a reduced dynamical model that delivers transport fluxes and their variations to coarse grids of larger-scale computational models, such as global circulation models for atmosphere and ocean.

Relocalization inside pre-built maps provides a big benefit in the course of today’s autonomous driving tasks where the map can be considered as an additional sensor for refining the estimated current pose of the vehicle. Due to potentially large drifts in the initial pose guess as well as maps containing unfiltered dynamic and temporal static objects (e.g. parking cars), traditional methods like ICP tend to fail and show high computation times. We propose a novel and fast relocalization method for accurate pose estimation inside a pre-built map based on 3D point clouds. The method is robust against inaccurate initialization caused by low performance GPS systems and tolerates the presence of unfiltered objects by specifically learning to extract significant features from current scans and adjacent map sections. More specifically, we introduce a novel distance-based matching loss enabling us to simultaneously extract important information from raw point clouds and aggregating inner- and inter-cloud context by utilizing self- and cross-attention inside a Graph Neural Network. We evaluate StickyLocalization’s (SL) performance through an extensive series of experiments using two benchmark datasets in terms of Relocalization on NuScenes and Loop Closing using KITTI’s Odometry dataset. We found that SL outperforms state-of-the art point cloud registration and relocalization methods in terms of transformation errors and runtime.

Poaceae represent one of the largest plant families in the world. Many species are of great economic importance as food and forage plants while others represent important weeds in agriculture. Although a large number of studies currently address the question of how plants can be best recognized on images, there is a lack of studies evaluating specific approaches for uniform species groups considered difficult to identify because they lack obvious visual characteristics. Poaceae represent an example of such a species group, especially when they are non-flowering. Here we present the results from an experiment to automatically identify Poaceae species based on images depicting six well-defined perspectives. One perspective shows the inflorescence while the others show vegetative parts of the plant such as the collar region with the ligule, adaxial and abaxial side of the leaf and culm nodes. For each species we collected 80 observations, each representing a series of six images taken with a smartphone camera. We extract feature representations from the images using five different convolutional neural networks (CNN) trained on objects from different domains and classify them using four state-of-the art classification algorithms. We combine these perspectives via score level fusion. In order to evaluate the potential of identifying non-flowering Poaceae we separately compared perspective combinations either comprising inflorescences or not. We find that for a fusion of all six perspectives, using the best combination of feature extraction CNN and classifier, an accuracy of 96.1% can be achieved. Without the inflorescence, the overall accuracy is still as high as 90.3%. In all but one case the perspective conveying the most information about the species (excluding inflorescence) is the ligule in frontal view. Our results show that even species considered very difficult to identify can achieve high accuracies in automatic identification as long as images depicting suitable perspectives are available. We suggest that our approach could be transferred to other difficult-to-distinguish species groups in order to identify the most relevant perspectives.

Due to the continuous digitalization of our society, distributed and web-based applications become omnipresent and making them more secure gains paramount relevance. Deep learning (DL) and its representation learning approach are increasingly been proposed for program code analysis potentially providing a powerful means in making software systems less vulnerable. This systematic literature review (SLR) is aiming for a thorough analysis and comparison of 32 primary studies on DL-based vulnerability analysis of program code. We found a rich variety of proposed analysis approaches, code embeddings and network topologies. We discuss these techniques and alternatives in detail. By compiling commonalities and differences in the approaches, we identify the current state of research in this area and discuss future directions. We also provide an overview of publicly available datasets in order to foster a stronger benchmarking of approaches. This SLR provides an overview and starting point for researchers interested in deep vulnerability analysis on program code.

Die Bildung der Augen führte zum Urknall der Evolution. Die Dynamik änderte sich von einem primitiven Organismus, der auf den Kontakt mit der Nahrung wartete, zu einem Organismus, der durch visuelle Sensoren gesucht wurde. Das menschliche Auge ist eine der raffiniertesten Entwicklungen der Evolution, aber es hat immer noch Mängel. Der Mensch hat über Millionen von Jahren einen biologischen Wahrnehmungsalgorithmus entwickelt, der in der Lage ist, Autos zu fahren, Maschinen zu bedienen, Flugzeuge zu steuern und Schiffe zu navigieren. Die Automatisierung dieser Fähigkeiten für Computer ist entscheidend für verschiedene Anwendungen, darunter selbstfahrende Autos, Augmented Realität und architektonische Vermessung. Die visuelle Nahfeldwahrnehmung im Kontext von selbstfahrenden Autos kann die Umgebung in einem Bereich von 0-10 Metern und 360&ring; Abdeckung um das Fahrzeug herum wahrnehmen. Sie ist eine entscheidende Entscheidungskomponente bei der Entwicklung eines sichereren automatisierten Fahrens. Jüngste Fortschritte im Bereich Computer Vision und Deep Learning in Verbindung mit hochwertigen Sensoren wie Kameras und LiDARs haben ausgereifte Lösungen für die visuelle Wahrnehmung hervorgebracht. Bisher stand die Fernfeldwahrnehmung im Vordergrund. Ein weiteres wichtiges Problem ist die begrenzte Rechenleistung, die für die Entwicklung von Echtzeit-Anwendungen zur Verfügung steht. Aufgrund dieses Engpasses kommt es häufig zu einem Kompromiss zwischen Leistung und Laufzeiteffizienz. Wir konzentrieren uns auf die folgenden Themen, um diese anzugehen: 1) Entwicklung von Nahfeld-Wahrnehmungsalgorithmen mit hoher Leistung und geringer Rechenkomplexität für verschiedene visuelle Wahrnehmungsaufgaben wie geometrische und semantische Aufgaben unter Verwendung von faltbaren neuronalen Netzen. 2) Verwendung von Multi-Task-Learning zur Überwindung von Rechenengpässen durch die gemeinsame Nutzung von initialen Faltungsschichten zwischen den Aufgaben und die Entwicklung von Optimierungsstrategien, die die Aufgaben ausbalancieren.

In recent years smartphones considerably changed our communication and are used on a daily (or even every minute) basis especially by students without any difficulties. Fluid flows also belong to our daily experiences. However, the education of the basic principles of fluid mechanics is sometimes cumbersome due to its non-linear nature. This problem may be tackled in practical sessions applying flow visualization techniques in wind or water tunnels and directly learn from own observations. Nowadays, often optical methods like particle imaging velocimetry (PIV) or particle tracking velocimetry (PTV) are used for these purposes. A typical PIV/PTV setup consists of a (double)pulse laser, a scientific camera and a synchronization device. The costs for this equipment can easily add up to more than 100,000 euros and the installations and set up of the systems requires experiences and is complex. For these reasons Universities often only offer practical courses for a small amount of students and the students may not be allowed to use and set up the systems by their own as the equipment is also needed for scientific research. Due to the COVID-19 pandemic it is also often not allowed to share equipment or even to work in larger groups during practical sessions.

FisheyeDistanceNet [1] proposed a self-supervised monocular depth estimation method for fisheye cameras with a large field of view (> 180&ring;). To achieve scale-invariant depth estimation, FisheyeDistanceNet supervises depth map predictions over multiple scales during training. To overcome this bottleneck, we incorporate self-attention layers and robust loss function [2] to FisheyeDistanceNet. A general adaptive robust loss function helps obtain sharp depth maps without a need to train over multiple scales and allows us to learn hyperparameters in loss function to aid in better optimization in terms of convergence speed and accuracy. We also ablate the importance of Instance Normalization over Batch Normalization in the network architecture. Finally, we generalize the network to be invariant to camera views by training multiple perspectives using front, rear, and side cameras. Proposed algorithm improvements, FisheyeDistanceNet++, result in 30% relative improvement in RMSE while reducing the training time by 25% on the WoodScape dataset. We also obtain state-of-the-art results on the KITTI dataset, in comparison to other self-supervised monocular methods.

3D object detection based on LiDAR point clouds is a crucial module in autonomous driving particularly for long range sensing. Most of the research is focused on achieving higher accuracy and these models are not optimized for deployment on embedded systems from the perspective of latency and power efficiency. For high speed driving scenarios, latency is a crucial parameter as it provides more time to react to dangerous situations. Typically a voxel or point-cloud based 3D convolution approach is utilized for this module. Firstly, they are inefficient on embedded platforms as they are not suitable for efficient parallelization. Secondly, they have a variable runtime due to level of sparsity of the scene which is against the determinism needed in a safety system. In this work, we aim to develop a very low latency algorithm with fixed runtime. We propose a novel semantic segmentation architecture as a single unified model for object center detection using key points, box predictions and orientation prediction using binned classification in a simpler Bird's Eye View (BEV) 2D representation. The proposed architecture can be trivially extended to include semantic segmentation classes like road without any additional computation. The proposed model has a latency of 4 ms on the embedded Nvidia Xavier platform. The model is 5X faster than other top accuracy models with a minimal accuracy degradation of 2% in Average Precision at IoU = 0.5 on KITTI dataset.

Deep learning algorithms classify plant species with high accuracy, and smartphone applications leverage this technology to enable users to identify plant species in the field. The question we address here is whether such crowd-sourced data contain substantial macroecological information. In particular, we aim to understand if we can detect known environmental gradients shaping plant co-occurrences. In this study we analysed 1 million data points collected through the use of the mobile app Flora Incognita between 2018 and 2019 in Germany and compared them with Florkart, containing plant occurrence data collected by more than 5000 floristic experts over a 70-year period. The direct comparison of the two data sets reveals that the crowd-sourced data particularly undersample areas of low population density. However, using nonlinear dimensionality reduction we were able to uncover macroecological patterns in both data sets that correspond well to each other. Mean annual temperature, temperature seasonality and wind dynamics as well as soil water content and soil texture represent the most important gradients shaping species composition in both data collections. Our analysis describes one way of how automated species identification could soon enable near real-time monitoring of macroecological patterns and their changes, but also discusses biases that must be carefully considered before crowd-sourced biodiversity data can effectively guide conservation measures.

Assisting and automating software engineering tasks is a state-of-the-art way to support stakeholders of development projects. A common assistance function of IDEs is the auto-completion of source code. Assistance functions, such as auto-completion, are almost entirely missing in modeling tools though auto-completion in general gains continuously more importance in software development. We analyze a user’s performed editing operations in order to anticipate modeling activities and to recommend appropriate auto-completions for them. Editing operations are captured as events and modeling activities are defined as complex event patterns, facilitating the matching by complex-event-processing. The approach provides adapted auto-completions reactively upon each editing operation of the user. We implemented the RapMOD prototype as add-in for the modeling tool Sparx Enterprise Architect™ . A controlled user experiment with 37 participants performing modeling tasks demonstrated the approach's potential to reduce modeling effort significantly. Users having auto-completions available for a modeling scenario performed the task 27 percent faster, needed to perform 56 percent less actions, and perceived the task 29 percent less difficult.

Defocus particle tracking (DPT) has gained increasing importance for its use to determine particle trajectories in all three dimensions with a single-camera system, as typical for a standard microscope, the workhorse of todays ongoing biomedical revolution. DPT methods derive the depth coordinates of particle images from the different defocusing patterns that they show when observed in a volume much larger than the respective depth of field. Therefore it has become common for state-of-the-art methods to apply image recognition techniques. Two of the most commonly and widely used DPT approaches are the application of (astigmatism) particle image model functions (MF methods) and the normalized cross-correlations between measured particle images and reference templates (CC methods). Though still young in the field, the use of neural networks (NN methods) is expected to play a significant role in future and more complex defocus tracking applications. To assess the different strengths of such defocus tracking approaches, we present in this work a general and objective assessment of their performances when applied to synthetic and experimental images of different degrees of astigmatism, noise levels, and particle image overlapping. We show that MF methods work very well in low-concentration cases, while CC methods are more robust and provide better performance in cases of larger particle concentration and thus stronger particle image overlap. The tested NN methods generally showed the lowest performance, however, in comparison to the MF and CC methods, they are yet in an early stage and have still great potential to develop within the field of DPT.

Surround View fisheye cameras are commonly deployed in automated driving for 360&ring; near-field sensing around the vehicle. This work presents a multi-task visual perception network on unrectified fisheye images to enable the vehicle to sense its surrounding environment. It consists of six primary tasks necessary for an autonomous driving system: depth estimation, visual odometry, semantic segmentation, motion segmentation, object detection, and lens soiling detection. We demonstrate that the jointly trained model performs better than the respective single task versions. Our multi-task model has a shared encoder providing a significant computational advantage and has synergized decoders where tasks support each other. We propose a novel camera geometry based adaptation mechanism to encode the fisheye distortion model both at training and inference. This was crucial to enable training on the WoodScape dataset, comprised of data from different parts of the world collected by 12 different cameras mounted on three different cars with different intrinsics and viewpoints. Given that bounding boxes is not a good representation for distorted fisheye images, we also extend object detection to use a polygon with non-uniformly sampled vertices. We additionally evaluate our model on standard automotive datasets, namely KITTI and Cityscapes. We obtain the state-of-the-art results on KITTI for depth estimation and pose estimation tasks and competitive performance on the other tasks. We perform extensive ablation studies on various architecture choices and task weighting methodologies. A short video at https://youtu.be/xbSjZ5OfPes provides qualitative results.

Pollen identification and quantification are crucial but challenging tasks in addressing a variety of evolutionary and ecological questions (pollination, paleobotany), but also for other fields of research (e.g. allergology, honey analysis or forensics). Researchers are exploring alternative methods to automate these tasks but, for several reasons, manual microscopy is still the gold standard. In this study, we present a new method for pollen analysis using multispectral imaging flow cytometry in combination with deep learning. We demonstrate that our method allows fast measurement while delivering high accuracy pollen identification. A dataset of 426 876 images depicting pollen from 35 plant species was used to train a convolutional neural network classifier. We found the best-performing classifier to yield a species-averaged accuracy of 96%. Even species that are difficult to differentiate using microscopy could be clearly separated. Our approach also allows a detailed determination of morphological pollen traits, such as size, symmetry or structure. Our phylogenetic analyses suggest phylogenetic conservatism in some of these traits. Given a comprehensive pollen reference database, we provide a powerful tool to be used in any pollen study with a need for rapid and accurate species identification, pollen grain quantification and trait extraction of recent pollen.

Many applications in chemistry, biology and medicine use microfluidic devices to separate, detect and analyze samples on a miniaturized size-level. Fluid flows evolving in channels of only several tens to hundreds of micrometers in size are often of a 3D nature, affecting the tailored transport of cells and particles. To analyze flow phenomena and local distributions of particles within those channels, astigmatic particle tracking velocimetry (APTV) has become a valuable tool, on condition that basic requirements like low optical aberrations and particles with a very narrow size distribution are fulfilled. Making use of the progress made in the field of machine vision, deep neural networks may help to overcome these limiting requirements, opening new fields of applications for APTV and allowing them to be used by nonexpert users. To qualify the use of a cascaded deep convolutional neural network (CNN) for particle detection and position regression, a detailed investigation was carried out starting from artificial particle images with known ground truth to real flow measurements inside a microchannel, using particles with uni- and bimodal size distributions. In the case of monodisperse particles, the mean absolute error and standard deviation of particle depth-position of less than and about 1 [my]m were determined, employing the deep neural network and the classical evaluation method based on the minimum Euclidean distance approach. While these values apply to all particle size distributions using the neural network, they continuously increase towards the margins of the measurement volume of about one order of magnitude for the classical method, if nonmonodisperse particles are used. Nevertheless, limiting the depth of measurement volume in between the two focal points of APTV, reliable flow measurements with low uncertainty are also possible with the classical evaluation method and polydisperse tracer particles. The results of the flow measurements presented herein confirm this finding. The source code of the deep neural network used here is available on https://github.com/SECSY-Group/DNN-APTV.

Requirements traceability has long been recognized as an important quality of a well-engineered system. Among stakeholders, traceability is often unpopular due to the unclear benefits. In fact, little evidence exists regarding the expected traceability benefits. There is a need for empirical work that studies the effect of traceability. In this paper, we focus on the four main requirements implementation supporting activities that utilize traceability. For each activity, we propose generalized traceability completeness measures. In a defined process, we selected 24 medium to large-scale open-source projects. For each software project, we quantified the degree to which a studied development activity was enabled by existing traceability with the proposed measures. We analyzed that data in a multi-level Poisson regression analysis. We found that the degree of traceability completeness for three of the studied activities significantly affects software quality, which we quantified as defect rate. Our results provide for the first time empirical evidence that more complete traceability decreases the expected defect rate in the developed software. The strong impact of traceability completeness on the defect rate suggests that traceability is of great practical value for any kind of software development project, even if traceability is not mandated by a standard or regulation

Die Nachvollziehbarkeit von Anforderungen ist wichtiges Qualitätsmerkmal der Softwareentwicklung. Für eine Vielzahl von Softwareentwicklungsaktivitäten ist die Nachvollziehbarkeit von Anforderungen eine notwenige Voraussetzung. Dazu gehören unter anderem die Analyse funktionaler Sicherheit, die Einflussanalyse, die Analyse des Abdeckungsgrades oder die Compliance. Für die Entwicklung sicherheitskritischer Softwaresysteme ist dieses Qualitätsmerkmal von besonderer Bedeutung. Daher wird dieses von entsprechenden Richtlinien zur Entwicklung sicherheitskritischer Software explizit vorgeschrieben. Obwohl die Relevanz der Nachvollziehbarkeit in Softwareprojekten allgemein bekannt ist, findet nur in wenigen Fällen eine systematische Planung zur Erreichung dieses Qualitätsmerkmals Anwendung. Häufig wird Nachvollziehbarkeit erst nachträglich umgesetzt. Daraus resultieren oft unvollständige Implementierungen der Nachvollziehbarkeit, die trotzdem als Grundlage für schwerwiegende Entscheidungen herangezogen werden. Aus diesem Grunde sollten die entsprechenden Implementierungen einer eingehenden Prüfung unterzogen werden, besonders im Rahmen der Entwicklung sicherheitskritischer Systeme. Dazu ist jedoch eine Vielzahl von Herausforderungen zu meistern. Zum einen hängt die Nachvollziehbarkeit von den projektspezifischen Zielen ab. Bei sicherheitskritischen Systemen müssen oft Vorgaben aus Richtlinien erfüllt werden. Auch die Nutzung der Nachvollziehbarkeit ist sehr stark von den jeweiligen Zielen abhängig. In dieser Arbeit wird ein Ansatz zur systematischen Prüfung von Softwareprojekten im Hinblick auf deren Nachvollziehbarkeit der Anforderungen vorgeschlagen. Eine notwendige Voraussetzung für den Prüfansatz ist die präzise Planung und Definition der Nachvollziehbarkeit von Anforderungen in einem Softwareprojekt. Daher wird im Rahmen dieser Arbeit ein entsprechender Planungsansatz präsentiert. Weiterhin wird ein analytisches Modell zur systematischen Prüfung der Nachvollziehbarkeit in Softwareprojekten präsentiert. Dieses Modell umfasst eine vollständige Klassifikation möglicher Fehlertypen. Außerdem werden Kriterien zur systematischen Erkennung dieser Fehler vorgeschlagen. Die Ergebnisse einer Expertenbefragung bestätigen die Vollständigkeit des analytischen Prüfmodells. Zudem wurde der vorgeschlagene Ansatz zur systematischen Prüfung der Nachvollziehbarkeit von Anforderungen in zwei Studien evaluiert. Dabei konnte der Nutzen des Ansatzes für die Entwicklung von sicherheitskritischer und nicht sicherheitskritischer Software nachgewiesen werden.