Refine
Year of publication
Document Type
- Conference Proceeding (926)
- Article (reviewed) (553)
- Article (unreviewed) (124)
- Part of a Book (65)
- Contribution to a Periodical (58)
- Book (29)
- Patent (29)
- Letter to Editor (28)
- Doctoral Thesis (19)
- Working Paper (19)
Conference Type
- Konferenzartikel (730)
- Konferenz-Abstract (134)
- Sonstiges (34)
- Konferenz-Poster (22)
- Konferenzband (8)
Language
- English (1857) (remove)
Is part of the Bibliography
- yes (1857) (remove)
Keywords
- RoboCup (32)
- Dünnschichtchromatographie (26)
- Gamification (17)
- Machine Learning (17)
- Export (16)
- Kommunikation (15)
- TRIZ (13)
- Plastizität (12)
- 3D printing (11)
- Deep Leaning (11)
Institute
- Fakultät Maschinenbau und Verfahrenstechnik (M+V) (562)
- Fakultät Elektrotechnik und Informationstechnik (E+I) (bis 03/2019) (486)
- Fakultät Elektrotechnik, Medizintechnik und Informatik (EMI) (ab 04/2019) (357)
- Fakultät Wirtschaft (W) (257)
- INES - Institut für nachhaltige Energiesysteme (165)
- Fakultät Medien und Informationswesen (M+I) (bis 21.04.2021) (146)
- ivESK - Institut für verlässliche Embedded Systems und Kommunikationselektronik (142)
- IMLA - Institute for Machine Learning and Analytics (72)
- ACI - Affective and Cognitive Institute (58)
- Fakultät Medien (M) (ab 22.04.2021) (51)
Open Access
- Open Access (801)
- Closed Access (625)
- Closed (241)
- Bronze (137)
- Gold (74)
- Diamond (62)
- Hybrid (45)
- Grün (12)
This paper presents the new Deep Reinforcement Learning (DRL) library RL-X and its application to the RoboCup Soccer Simulation 3D League and classic DRL benchmarks. RL-X provides a flexible and easy-to-extend codebase with self-contained single directory algorithms. Through the fast JAX-based implementations, RL-X can reach up to 4.5x speedups compared to well-known frameworks like Stable-Baselines3.
The use of artificial intelligence continues to impact a broad variety of domains, application areas, and people. However, interpretability, understandability, responsibility, accountability, and fairness of the algorithms' results - all crucial for increasing humans' trust into the systems - are still largely missing. The purpose of this seminar is to understand how these components factor into the holistic view of trust. Further, this seminar seeks to identify design guidelines and best practices for how to build interactive visualization systems to calibrate trust.
With the rising necessity of explainable artificial intelligence (XAI), we see an increase in task-dependent XAI methods on varying abstraction levels. XAI techniques on a global level explain model behavior and on a local level explain sample predictions. We propose a visual analytics workflow to support seamless transitions between global and local explanations, focusing on attributions and counterfactuals on time series classification. In particular, we adapt local XAI techniques (attributions) that are developed for traditional datasets (images, text) to analyze time series classification, a data type that is typically less intelligible to humans. To generate a global overview, we apply local attribution methods to the data, creating explanations for the whole dataset. These explanations are projected onto two dimensions, depicting model behavior trends, strategies, and decision boundaries. To further inspect the model decision-making as well as potential data errors, a what-if analysis facilitates hypothesis generation and verification on both the global and local levels. We constantly collected and incorporated expert user feedback, as well as insights based on their domain knowledge, resulting in a tailored analysis workflow and system that tightly integrates time series transformations into explanations. Lastly, we present three use cases, verifying that our technique enables users to (1)~explore data transformations and feature relevance, (2)~identify model behavior and decision boundaries, as well as, (3)~the reason for misclassifications.
There is an ongoing debate about the use and scope of Clayton M. Christensen´s idea of disruptive innovation, including the question of whether it is a management buzz phrase or a valuable theory. This discussion considers the general question of how innovation in the field of management theories and concepts finds its way to the different target groups. This conceptual paper combines the different concepts of the creation and dissemination of management trends in a basic framework based on a short review of models for the dissemination of management ideas. This framework allows an analysis of the character of new management ideas like disruptive innovation. By measuring the impact of the theory on the academic sphere using a bibliometric statistic of the number of academic publications on Google scholar and Scopus and a meta-analysis of research papers, we show the significant influence of disruptive innovation beyond pure management fads.
Variable refrigerant flow (VRF) and variable air volume (VAV) systems are considered among the best heating, ventilation, and air conditioning systems (HVAC) thanks to their ability to provide cooling and heating in different thermal zones of the same building. As well as their ability to recover the heat rejected from spaces requiring cooling and reuse it to heat another space. Nevertheless, at the same time, these systems are considered one of the most energy-consuming systems in the building. So, it is crucial to well size the system according to the building’s cooling and heating needs and the indoor temperature fluctuations. This study aims to compare these two energy systems by conducting an energy model simulation of a real building under a semi-arid climate for cooling and heating periods. The developed building energy model (BEM) was validated and calibrated using measured and simulated indoor air temperature and energy consumption data. The study aims to evaluate the effect of these HVAC systems on energy consumption and the indoor thermal comfort of the building. The numerical model was based on the Energy Plus simulation engine. The approach used in this paper has allowed us to reach significant quantitative energy saving along with a high level of indoor thermal comfort by using the VRF system compared to the VAV system. The findings prove that the VRF system provides 46.18% of the annual total heating energy savings and 6.14% of the annual cooling and ventilation energy savings compared to the VAV system.
Modern CNNs are learning the weights of vast numbers of convolutional operators. In this paper, we raise the fundamental question if this is actually necessary. We show that even in the extreme case of only randomly initializing and never updating spatial filters, certain CNN architectures can be trained to surpass the accuracy of standard training. By reinterpreting the notion of pointwise ($1\times 1$) convolutions as an operator to learn linear combinations (LC) of frozen (random) spatial filters, we are able to analyze these effects and propose a generic LC convolution block that allows tuning of the linear combination rate. Empirically, we show that this approach not only allows us to reach high test accuracies on CIFAR and ImageNet but also has favorable properties regarding model robustness, generalization, sparsity, and the total number of necessary weights. Additionally, we propose a novel weight sharing mechanism, which allows sharing of a single weight tensor between all spatial convolution layers to massively reduce the number of weights.
Learning programming fundamentals is considered as one of the most challenging and complex learning activities. Some authors have proposed visual programming language (VPL) approaches to address part of the inherent complexity [1]. A visual programming language lets users develop programs by combining program elements, like loops graphically rather than by specifying them textually. Visual expressions, spatial arrangements of text and graphic symbols are used either as syntax elements or secondary notation. VPLs are normally used for educational multimedia, video games, system development, and data warehousing/business analytics purposes. For example, Scratch, a platform of Massachusetts Institute of Technology, is designed for kids and after school programs.
Design of mobile software applications is considered as one of the most challenging application domains due to the build in sensors as part of a mobile device, like GPS, camera or Near Field Communication (NFC). Sensors enable creation of context-aware mobile applications in which applications can discover and take advantage of contextual information, such as user location, nearby people and objects, and the current user activity. As a consequence, context-aware mobile applications can sense clues about the situational environment making mobile devices more intelligent, adaptive, and personalized. Such context aware mobile applications seem to be motivating and attractive case studies, especially for programming beginners (“my own first app”).
In this work, we introduce a use-case centered approach as well as clear separation of user interface design and sensor-based program development. We provide an in-depth discussion of a new VPL based teaching method, a step by step development process to enable programming beginners the creation of context aware mobile applications. Finally, we argue that addressing challenges for programming beginners by our teaching approach could make programming teaching more motivating, with an additional impact on the final software quality and scalability.
The key contributions of our study are the following:
- An overview of existing attempts to use VPL approaches for mobile applications
- A use case centered teaching approach based on a clear separation of user interface design and sensor-based program development
- A teaching case study enabling beginners a step by step creation of context-aware mobile applications based on the MIT App Inventor (a platform of Massachusetts Institute of Technology)
- Open research challenges and perspectives for further development of our teaching approach
References:
[1] Idrees, M., Aslam, F. (2022). A Comprehensive Survey and Analysis of Diverse Visual Programming Languages, VFAST Transactions on Software Engineering, 2022, Volume 10, Number 2, pp 47-60.
During pyrolysis, biomass is carbonised in the absence of oxygen to produce biochar with heat and/or electricity as co-products making pyrolysis one of the promising negative emission technologies to reach climate goals worldwide. This paper presents a simplified representation of pyrolysis and analyses the impact of this technology on the energy system. Results show that the use of pyrolysis can allow getting zero emissions with lower costs by making changes in the unit commitment of the power plants, e.g. conventional power plants are used differently, as the emissions will be compensated by biochar. Additionally, the process of pyrolysis can enhance the flexibility of energy systems, as it shows a correlation between the electricity generated by pyrolysis and the hydrogen installation capacity, being hydrogen used less when pyrolysis appears. The results indicate that pyrolysis, which is available on the market, integrates well into the energy system with a promising potential to sequester carbon.
3D Bin Picking with an innovative powder filled gripper and a torque controlled collaborative robot
(2023)
A new and innovative powder filled gripper concept will be introduced to a process to pick parts out of a box without the use of a camera system which guides the robot to the part. The gripper is a combination of an inflatable skin, and a powder inside. In the unjammed condition, the powder is soft and can adjust to the geometry of the part which will be handled. By applying a vacuum to the inflatable skin, the powder gets jammed and transforms to a solid shaped form in which the gripper was brought before applying the vacuum. This physical principle is used to pick parts. The flexible skin of the gripper adjusts to all kinds of shapes, and therefore, can be used to realize 3D bin picking. With the help of a force controlled robot, the gripper can be pushed with a consistent force on flexible positions depending of the filling level of the box. A Kuka LBR iiwa with joint torque sensors in all of its seven axis’ was used to achieve a constant contact pressure. This is the basic criteria to achieve a robust picking process.
Socially assistive robots (SARs) are becoming more prevalent in everyday life, emphasizing the need to make them socially acceptable and aligned with users' expectations. Robots' appearance impacts users' behaviors and attitudes towards them. Therefore, product designers choose visual qualities to give the robot a character and to imply its functionality and personality. In this work, we sought to investigate the effect of cultural differences on Israeli and German designers' perceptions and preferences regarding the suitable visual qualities of SARs in four different contexts: a service robot for an assisted living/retirement residence facility, a medical assistant robot for a hospital environment, a COVID-19 officer robot, and a personal assistant robot for domestic use. Our results indicate that Israeli and German designers share similar perceptions of visual qualities and most of the robotics roles. However, we found differences in the perception of the COVID-19 officer robot's role and, by that, its most suitable visual design. This work indicates that context and culture play a role in users' perceptions and expectations; therefore, they should be taken into account when designing new SARs for diverse contexts.
Recent advances in spiked shoe design, characterized by increased longitudinal stiffness, thicker midsole foams, and reconfigured geometry are considered to improve sprint performance. However, so far there is no empirical data on the effects of advanced spikes technology on maximal sprinting speed (MSS) published yet. Consequently, we assessed MSS via ‘flying 30m’ sprints of 44 trained male (PR: 10.32 s - 12.08 s) and female (PR: 11.56 s - 14.18 s) athletes, wearing both traditional and advanced spikes in a randomized, repeated measures design. The results revealed a statistically significant increase in MSS by 1.21% on average when using advanced spikes technology. Notably, 87% of participants showed improved MSS with the use of advanced spikes. A cluster analysis unveiled that athletes with higher MSS may benefit to a greater extent. However, individual responses varied widely, suggesting the influence of multiple factors that need detailed exploration. Therefore, coaches and athletes are advised to interpret the promising performance enhancements cautiously and evaluate the appropriateness of the advanced spike technology for their athletes critically.
High-tech running shoes and spikes ("super-footwear") are currently being debated in sports. There is direct evidence that distance running super shoes improve running economy; however, it is not well established to which extent world-class performances are affected over the range of track and road running events.
This study examined publicly available performance datasets of annual best track and road performances for evidence of potential systematic performance effects following the introduction of super footwear. The analysis was based on the 100 best performances per year for men and women in outdoor events from 2010 to 2022, provided by the world governing body of athletics (World Athletics).
We found evidence of progressing improvements in track and road running performances after the introduction of super distance running shoes in 2016 and super spike technology in 2019. This evidence is more pronounced for distances longer than 1500 m in women and longer than 5000 m in men. Women seem to benefit more from super footwear in distance running events than men.
While the observational study design limits causal inference, this study provides a database on potential systematic performance effects following the introduction of super shoes/spikes in track and road running events in world-class athletes. Further research is needed to examine the underlying mechanisms and, in particular, potential sex differences in the performance effects of super footwear.
We revisit the quantitative analysis of the ultrafast magnetoacoustic experiment in a freestanding nickel thin film by Kim and Bigot [J.-W. Kim and J.-Y. Bigot, Phys. Rev. B 95, 144422 (2017)] by applying our recently proposed approach of magnetic and acoustic eigenmode decomposition. We show that the application of our modeling to the analysis of time-resolved reflectivity measurements allows for the determination of amplitudes and lifetimes of standing perpendicular acoustic phonon resonances with unprecedented accuracy. The acoustic damping is found to scale as ∝ω2 for frequencies up to 80 GHz, and the peak amplitudes reach 10−3. The experimentally measured magnetization dynamics for different orientations of an external magnetic field agrees well with numerical solutions of magnetoelastically driven magnon harmonic oscillators. Symmetry-based selection rules for magnon-phonon interactions predicted by our modeling approach allow for the unambiguous discrimination between spatially uniform and nonuniform modes, as confirmed by comparing the resonantly enhanced magnetoelastic dynamics simultaneously measured on opposite sides of the film. Moreover, the separation of timescales for (early) rising and (late) decreasing precession amplitudes provide access to magnetic (Gilbert) and acoustic damping parameters in a single measurement.
While most ultrafast time-resolved optical pump-probe experiments in magnetic materials reveal the spatially homogeneous magnetization dynamics of ferromagnetic resonance (FMR), here we explore the magneto-elastic generation of GHz-to-THz frequency spin waves (exchange magnons). Using analytical magnon oscillator equations, we apply time-domain and frequency-domain approaches to quantify the results of ultrafast time-resolved optical pump-probe experiments in free-standing ferromagnetic thin films. Simulations show excellent agreement with the experiment, provide acoustic and magnetic (Gilbert) damping constants and highlight the role of symmetry-based selection rules in phonon-magnon interactions. The analysis is extended to hybrid multilayer structures to explore the limits of resonant phonon-magnon interactions up to THz frequencies.
The technique of laser ultrasonics perfectly meets the need for noncontact, noninvasive, nondestructive mechanical probing of nanometer- to millimeter-size samples. However, this technique is limited to the excitation of low-amplitude strains, below the threshold for optical damage of the sample. In the context of strain engineering of materials, alternative optical techniques enabling the excitation of high-amplitude strains in a nondestructive optical regime are needed. We introduce here a nondestructive method for laser-shock wave generation based on additive superposition of multiple laser-excited strain waves. This technique enables strain generation up to mechanical failure of a sample at pump laser fluences below optical ablation or melting thresholds. We demonstrate the ability to generate nonlinear surface acoustic waves (SAWs) in Nb-SrTiO3 substrates, with associated strains in the percent range and pressures up to 3 GPa at 1 kHz repetition rate and close to 10 GPa for several hundred shocks. This study paves the way for the investigation of a host of high-strain SAW-induced phenomena, including phase transitions in conventional and quantum materials, plasticity and a myriad of material failure modes, chemistry and other effects in bulk samples, thin layers, and two-dimensional materials.
The utilisation of artificial intelligence (AI) is progressively emerging as a significant mechanism for innovation in human resource management (HRM). The capacity to facilitate the transformation of employee performance across numerous responsibilities. AI development, there remains a dearth of comprehensive exploration into the potential opportunities it presents for enhancing workplace performance among employees. To bridge this gap in knowledge, the present work carried out a survey with 300 participants, utilises a fuzzy set-theoretic method that is grounded on the conceptualisation of AI, KS, and HRM. The findings of our study indicate that the exclusive adoption of AI technologies does not adequately enhance HRM engagements. In contrast, the integration of AI and KS offers a more viable HRM approach for achieving optimal performance in a dynamic digital society. This approach has the potential to enhance employees’ proficiency in executing their responsibilities and cultivate a culture of creativity inside the firm.
Purpose
Although start-ups have gained increasing scholarly attention, we lack sufficient understanding of their entrepreneurial strategic posture (ESP) in emerging economies. The purpose of this study is to examine the processes of ESP of new technology venture start-ups (NTVs) in an emerging market context.
Design/methodology/approach
In line with grounded theory guidelines and the inductive research traditions, the authors adopted a qualitative approach involving 42 in-depth semi-structured interviews with Ghanaian NTV entrepreneurs to gain a comprehensive analysis at the micro-level on the entrepreneurs' strategic posturing. A systematic procedure for data analysis was adopted.
Findings
From the authors' analysis of Ghanaian NTVs, the authors derived a three-stage model to elucidate the nature and process of ESP Phase 1 spotting and exploiting market opportunities, Phase II identifying initial advantages and Phase III ascertaining and responding to change.
Originality/value
The study contributes to advancing research on ESP by explicating the process through which informal ties and networks are utilised by NTVs and NTVs' founders to overcome extreme resource constraints and information vacuums in contexts of institutional voids. The authors depart from past studies in demonstrating how such ties can be harnessed in spotting and exploiting market opportunities by NTVs. On this basis, the paper makes original contributions to ESP theory and practice.
Purpose
Although recent literature has examined diverse measures adopted by SMEs to navigate the COVID-19 turbulence, there is a shortage of evidence on how crisis-time strategy creation behaviour and digitalization activities increase (1) sales and (2) cash flow. Thus, predicated on a novel strategy creation perspective, this inquiry aims to investigate the crisis behaviour, sales and cash flow performance of 528 SMEs in Morocco.
Design/methodology/approach
Novel links between (1) aggregate wage cuts, (2) variable operating hours, (3) deferred payment to suppliers, (4) deferred payment to tax authorities and (5) sales performance are developed and tested. A further link between sales performance and cash flow is also examined and the analysis is conducted using a non-linear structural equation modelling technique.
Findings
While there is a significant association between strategy creation behaviours and sales performance, only variable operating hours have a positive effect. Also, sales performance increases cash flow and this relationship is substantially strengthened by e-commerce digitalization and innovation.
Originality/value
Theoretically, to the best of the authors’ knowledge, this is one of the first inquiries to espouse the strategy creation view to explain SMEs' crisis-time behaviour and digitalization. For practical purposes, to supplement Moroccan SMEs' propensity to seek tax deferrals, it is argued that debt and equity support measures are also needed to boost sales performance and cash flow.
Polyarticulated active prostheses constitute a promising solution for upper limb amputees. The bottleneck for their adoption though, is the lack of intuitive control. In this context, machine learning algorithms based on pattern recognition from electromyographic (EMG) signals represent a great opportunity for naturally operating prosthetic devices, but their performance is strongly affected by the selection of input features. In this study, we investigated different combinations of 13 EMG-derived features obtained from EMG signals of healthy individuals performing upper limb movements and tested their performance for movement classification using an Artificial Neural Network. We found that input data (i.e., the set of input features) can be reduced by more than 50% without any loss in accuracy, while diminishing the computing time required to train the classifier. Our results indicate that input features must be properly selected in order to optimize prosthetic control.
The main focus of this chapter is the theoretical and instrumental processes that underpin densitometric methods widely used in thin-layer chromatography (TLC). Densitometric methods include UV–vis, luminescence and fluorescence optical measurements as well as infrared and Raman spectroscopic measurements. The chapter is divided in two general parts: a theoretical part and a practical part. The systems for direct radioactivity measurements and the combination of TLC with mass spectrometry are also discussed. All these systems allow measuring an intensity distribution directly on a TLC plate. We call this “in situ detection” because no analyte is removed from the plate.
The main focus of this chapter is the theoretical and instrumental processes that underpin densitometric methods widely used in thin-layer chromatography (TLC). Densitometric methods include UV–vis, luminescence, and fluorescence optical measurements as well as infrared and Raman spectroscopic measurements. The chapter is divided in two general parts: a theoretical part and a practical part. The systems for direct radioactivity measurements and the combination of TLC with mass spectrometry are also discussed. All these systems allow measuring an intensity distribution directly on a TLC plate. We call this “in situ detection” because no analyte is removed from the plate.
Recently, photovoltaic (PV) with energy storage systems (ESS) have been widely adopted in buildings to overcome growing power demands and earn financial benefits. The overall energy cost can be optimized by combining a well-sized hybrid PV/ESS system with an efficient energy management system (EMS). Generally, EMS is implemented within the overall functions of the Building Automation System (BAS). However, due to its limited computing resources, BAS cannot handle complex algorithms that aim to optimize energy use in real-time under different operating conditions. Furthermore, islanding the building's local network to maximize the PV energy share represents a challenging task due to the potential technical risks. In this context, this article addresses an improved approach based on upgrading the BAS data analytics capability by means of an edge computing technology. The edge communicates with the BAS low-level controller using a serial communication protocol. Taking advantage of the high computing ability of the edge device, an optimization-based EMS of the PV/ESS hybrid system is implemented. Different testing scenarios have been carried out on a real prototype with different weather conditions, and the results show the implementation feasibility and technical performance of such advanced EMS for the management of building energy resources. It has also been proven to be feasible and advantageous to operate the local energy network in island mode while ensuring system safety. Additionally, an estimated energy saving improvement of 6.23 % has been achieved using optimization-based EMS compared to the classical rule-based EMS, with better ESS constraints fulfillment.
Analysing and predicting the advance rate of a tunnel boring machine (TBM) in hard rock is integral to tunnelling project planning and execution. It has been applied in the industry for several decades with varying success. Most prediction models are based on or designed for large-diameter TBMs, and much research has been conducted on related tunnelling projects. However, only a few models incorporate information from projects with an outer diameter smaller than 5 m and no penetration prediction model for pipe jacking machines exists to date. In contrast to large TBMs, small-diameter TBMs and their projects have been considered little in research. In general, they are characterised by distinctive features, including insufficient geotechnical information, sometimes rather short drive lengths, special machine designs and partially concurring lining methods like pipe jacking and segment lining. A database which covers most of the parameters mentioned above has been compiled to investigate the performance of small-diameter TBMs in hard rock. In order to provide sufficient geological and technical variance, this database contains 37 projects with 70 geotechnically homogeneous areas. Besides the technical parameters, important geotechnical data like lithological information, unconfined compressive strength, tensile strength and point load index is included and evaluated. The analysis shows that segment lining TBMs have considerably higher penetration rates in similar geological and technical settings mostly due to their design parameters. Different methodologies for predicting TBM penetration, including state-of-the-art models from the literature as well as newly derived regression and machine learning models, are discussed and deployed for backward modelling of the projects contained in the database. New ranges of application for small-diameter tunnelling in several industry-standard penetration models are presented, and new approaches for the penetration prediction of pipe jacking machines in hard rock are proposed.
Following the traditional paradigm of convolutional neural networks (CNNs), modern CNNs manage to keep pace with more recent, for example transformer-based, models by not only increasing model depth and width but also the kernel size. This results in large amounts of learnable model parameters that need to be handled during training. While following the convolutional paradigm with the according spatial inductive bias, we question the significance of \emph{learned} convolution filters. In fact, our findings demonstrate that many contemporary CNN architectures can achieve high test accuracies without ever updating randomly initialized (spatial) convolution filters. Instead, simple linear combinations (implemented through efficient 1×1 convolutions) suffice to effectively recombine even random filters into expressive network operators. Furthermore, these combinations of random filters can implicitly regularize the resulting operations, mitigating overfitting and enhancing overall performance and robustness. Conversely, retaining the ability to learn filter updates can impair network performance. Lastly, although we only observe relatively small gains from learning 3×3 convolutions, the learning gains increase proportionally with kernel size, owing to the non-idealities of the independent and identically distributed (\textit{i.i.d.}) nature of default initialization techniques.
We have developed a methodology for the systematic generation of a large image dataset of macerated wood references, which we used to generate image data for nine hardwood genera. This is the basis for a substantial approach to automate, for the first time, the identification of hardwood species in microscopic images of fibrous materials by deep learning. Our methodology includes a flexible pipeline for easy annotation of vessel elements. We compare the performance of different neural network architectures and hyperparameters. Our proposed method performs similarly well to human experts. In the future, this will improve controls on global wood fiber product flows to protect forests.
State-of-the-art models for pixel-wise prediction tasks such as image restoration, image segmentation, or disparity estimation, involve several stages of data resampling, in which the resolution of feature maps is first reduced to aggregate information and then sequentially increased to generate a high-resolution output. Several previous works have investigated the effect of artifacts that are invoked during downsampling and diverse cures have been proposed that facilitate to improve prediction stability and even robustness for image classification. However, equally relevant, artifacts that arise during upsampling have been less discussed. This is significantly relevant as upsampling and downsampling approaches face fundamentally different challenges. While during downsampling, aliases and artifacts can be reduced by blurring feature maps, the emergence of fine details is crucial during upsampling. Blurring is therefore not an option and dedicated operations need to be considered. In this work, we are the first to explore the relevance of context during upsampling by employing convolutional upsampling operations with increasing kernel size while keeping the encoder unchanged. We find that increased kernel sizes can in general improve the prediction stability in tasks such as image restoration or image segmentation, while a block that allows for a combination of small-size kernels for fine details and large-size kernels for artifact removal and increased context yields the best results.
Fix your downsampling ASAP! Be natively more robust via Aliasing and Spectral Artifact free Pooling
(2023)
Convolutional neural networks encode images through a sequence of convolutions, normalizations and non-linearities as well as downsampling operations into potentially strong semantic embeddings. Yet, previous work showed that even slight mistakes during sampling, leading to aliasing, can be directly attributed to the networks' lack in robustness. To address such issues and facilitate simpler and faster adversarial training, [12] recently proposed FLC pooling, a method for provably alias-free downsampling - in theory. In this work, we conduct a further analysis through the lens of signal processing and find that such current pooling methods, which address aliasing in the frequency domain, are still prone to spectral leakage artifacts. Hence, we propose aliasing and spectral artifact-free pooling, short ASAP. While only introducing a few modifications to FLC pooling, networks using ASAP as downsampling method exhibit higher native robustness against common corruptions, a property that FLC pooling was missing. ASAP also increases native robustness against adversarial attacks on high and low resolution data while maintaining similar clean accuracy or even outperforming the baseline.
Motivated by the recent trend towards the usage of larger receptive fields for more context-aware neural networks in vision applications, we aim to investigate how large these receptive fields really need to be. To facilitate such study, several challenges need to be addressed, most importantly: (i) We need to provide an effective way for models to learn large filters (potentially as large as the input data) without increasing their memory consumption during training or inference, (ii) the study of filter sizes has to be decoupled from other effects such as the network width or number of learnable parameters, and (iii) the employed convolution operation should be a plug-and-play module that can replace any conventional convolution in a Convolutional Neural Network (CNN) and allow for an efficient implementation in current frameworks. To facilitate such models, we propose to learn not spatial but frequency representations of filter weights as neural implicit functions, such that even infinitely large filters can be parameterized by only a few learnable weights. The resulting neural implicit frequency CNNs are the first models to achieve results on par with the state-of-the-art on large image classification benchmarks while executing convolutions solely in the frequency domain and can be employed within any CNN architecture. They allow us to provide an extensive analysis of the learned receptive fields. Interestingly, our analysis shows that, although the proposed networks could learn very large convolution kernels, the learned filters practically translate into well-localized and relatively small convolution kernels in the spatial domain.
Assessing the robustness of deep neural networks against out-of-distribution inputs is crucial, especially in safety-critical domains like autonomous driving, but also in safety systems where malicious actors can digitally alter inputs to circumvent safety guards. However, designing effective out-of-distribution tests that encompass all possible scenarios while preserving accurate label information is a challenging task. Existing methodologies often entail a compromise between variety and constraint levels for attacks and sometimes even both. In a first step towards a more holistic robustness evaluation of image classification models, we introduce an attack method based on image solarization that is conceptually straightforward yet avoids jeopardizing the global structure of natural images independent of the intensity. Through comprehensive evaluations of multiple ImageNet models, we demonstrate the attack's capacity to degrade accuracy significantly, provided it is not integrated into the training augmentations. Interestingly, even then, no full immunity to accuracy deterioration is achieved. In other settings, the attack can often be simplified into a black-box attack with model-independent parameters. Defenses against other corruptions do not consistently extend to be effective against our specific attack.
Project website: https://github.com/paulgavrikov/adversarial_solarization
Entity Matching (EM) defines the task of learning to group objects by transferring semantic concepts from example groups (=entities) to unseen data. Despite the general availability of image data in the context of many EM-problems, most currently available EM-algorithms solely rely on (textual) meta data. In this paper, we introduce the first publicly available large-scale dataset for "visual entity matching", based on a production level use case in the retail domain. Using scanned advertisement leaflets, collected over several years from different European retailers, we provide a total of ~786k manually annotated, high resolution product images containing ~18k different individual retail products which are grouped into ~3k entities. The annotation of these product entities is based on a price comparison task, where each entity forms an equivalence class of comparable products. Following on a first baseline evaluation, we show that the proposed "visual entity matching" constitutes a novel learning problem which can not sufficiently be solved using standard image based classification and retrieval algorithms. Instead, novel approaches which allow to transfer example based visual equivalent classes to new data are needed to address the proposed problem. The aim of this paper is to provide a benchmark for such algorithms.
Information about the dataset, evaluation code and download instructions are provided under https://www.retail-786k.org/.
For the treatment of bone defects, biodegradable, compressive biomaterials are needed as replacements that degrade as the bone regenerates. The problem with existing materials has either been their insufficient mechanical strength or the excessive differences in their elastic modulus, leading to stress shielding and eventual failure. In this study, the compressive strength of CPC ceramics (with a layer thickness of more than 12 layers) was compared with sintered β-TCP ceramics. It was assumed that as the number of layers increased, the mechanical strength of 3D-printed scaffolds would increase toward the value of sintered ceramics. In addition, the influence of the needle inner diameter on the mechanical strength was investigated. Circular scaffolds with 20, 25, 30, and 45 layers were 3D printed using a 3D bioplotter, solidified in a water-saturated atmosphere for 3 days, and then tested for compressive strength together with a β-TCP sintered ceramic using a Zwick universal testing machine. The 3D-printed scaffolds had a compressive strength of 41.56 ± 7.12 MPa, which was significantly higher than that of the sintered ceramic (24.16 ± 4.44 MPa). The 3D-printed scaffolds with round geometry reached or exceeded the upper limit of the compressive strength of cancellous bone toward substantia compacta. In addition, CPC scaffolds exhibited more bone-like compressibility than the comparable β-TCP sintered ceramic, demonstrating that the mechanical properties of CPC scaffolds are more similar to bone than sintered β-TCP ceramics.
Differentiation between human and non-human objects can increase efficiency of human-robot collaborative applications. This paper proposes to use convolutional neural networks for classifying objects in robotic applications. The body temperature of human beings is used to classify humans and to estimate the distance to the sensor. Using image classification with convolutional neural networks it is possible to detect humans in the surroundings of a robot up to five meters distance with low-cost and low-weight thermal cameras. Using transfer learning technique we trained the GoogLeNet and MobilenetV2. Results show accuracies of 99.48 % and 99.06 % respectively.
Detecting Images Generated by Deep Diffusion Models using their Local Intrinsic Dimensionality
(2023)
Diffusion models recently have been successfully applied for the visual synthesis of strikingly realistic appearing images. This raises strong concerns about their potential for malicious purposes. In this paper, we propose using the lightweight multi Local Intrinsic Dimensionality (multiLID), which has been originally developed in context of the detection of adversarial examples, for the automatic detection of synthetic images and the identification of the according generator networks. In contrast to many existing detection approaches, which often only work for GAN-generated images, the proposed method provides close to perfect detection results in many realistic use cases. Extensive experiments on known and newly created datasets demonstrate that the proposed multiLID approach exhibits superiority in diffusion detection and model identification.Since the empirical evaluations of recent publications on the detection of generated images are often mainly focused on the "LSUN-Bedroom" dataset, we further establish a comprehensive benchmark for the detection of diffusion-generated images, including samples from several diffusion models with different image sizes.The code for our experiments is provided at https://github.com/deepfake-study/deepfake-multiLID.
Following their success in visual recognition tasks, Vision Transformers(ViTs) are being increasingly employed for image restoration. As a few recent works claim that ViTs for image classification also have better robustness properties, we investigate whether the improved adversarial robustness of ViTs extends to image restoration. We consider the recently proposed Restormer model, as well as NAFNet and the "Baseline network" which are both simplified versions of a Restormer. We use Projected Gradient Descent (PGD) and CosPGD for our robustness evaluation. Our experiments are performed on real-world images from the GoPro dataset for image deblurring. Our analysis indicates that contrary to as advocated by ViTs in image classification works, these models are highly susceptible to adversarial attacks. We attempt to find an easy fix and improve their robustness through adversarial training. While this yields a significant increase in robustness for Restormer, results on other networks are less promising. Interestingly, we find that the design choices in NAFNet and Baselines, which were based on iid performance, and not on robust generalization, seem to be at odds with the model robustness.
The identification of vulnerabilities is an important element in the software development life cycle to ensure the security of software. While vulnerability identification based on the source code is a well studied field, the identification of vulnerabilities on basis of a binary executable without the corresponding source code is more challenging. Recent research [1] has shown how such detection can generally be enabled by deep learning methods, but appears to be very limited regarding the overall amount of detected vulnerabilities. We analyse to what extent we could cover the identification of a larger variety of vulnerabilities. Therefore, a supervised deep learning approach using recurrent neural networks for the application of vulnerability detection based on binary executables is used. The underlying basis is a dataset with 50,651 samples of vulnerable code in the form of a standardised LLVM Intermediate Representation. Te vectorised features of a Word2Vec model are used to train different variations of three basic architectures of recurrent neural networks (GRU, LSTM, SRNN). A binary classification was established for detecting the presence of an arbitrary vulnerability, and a multi-class model was trained for the identification of the exact vulnerability, which achieved an out-of-sample accuracy of 88% and 77%, respectively. Differences in the detection of different vulnerabilities were also observed, with non-vulnerable samples being detected with a particularly high precision of over 98%. Thus, our proposed technical approach and methodology enables an accurate detection of 23 (compared to 4 [1]) vulnerabilities.
The importance of machine learning (ML) has been increasing dramatically for years. From assistance systems to production optimisation to healthcare support, almost every area of daily life and industry is coming into contact with machine learning. Besides all the benefits ML brings, the lack of transparency and difficulty in creating traceability pose major risks. While solutions exist to make the training of machine learning models more transparent, traceability is still a major challenge. Ensuring the identity of a model is another challenge, as unnoticed modification of a model is also a danger when using ML. This paper proposes to create an ML Birth Certificate and ML Family Tree secured by blockchain technology. Important information about training and changes to the model through retraining can be stored in a blockchain and accessed by any user to create more security and traceability about an ML model.
Automation devices or automation stations (AS) take on the task of controlling, regulating, monitoring and, if necessary, optimising building systems and their system components (e.g. pumps, compressors, fans) based on recorded process variables. For this purpose, a wide range of control and regulation methods are used, starting with simple on/off controllers, through classic PID controllers, to higher-order controllers such as adaptive, model-predictive, knowledge-based or adaptive controllers.
Starting with a brief introduction to automation technology (Sect. 7.1), the chapter goes into the structure and functionality of the usual compact controllers using the application examples of solar thermal systems and heat pump systems (Sect. 7.2). Finally, the integration of system automation into a higher-level building automation system and into the building management system is described using specific application examples (Sect. 7.3).
This central book chapter now details the implementation of automation of solar domestic hot water systems, solar assisted building heating, rooms, solar cooling systems, heat pump heating systems, geothermal systems and thermally activated building component systems. Hydraulic and automation diagrams are used to explain how the automation of these systems works. A detailed insight into the engineering and technical interrelationships involved in the use of these systems, as well as the use of simulation tools, enables effective control and regulation. System characteristic curves and systematic procedures support the automation engineer in his tasks.
Renewable energy sources such as solar radiation, geothermal heat and ambient heat are available for energy conversion. With the help of special converters, these resources can be put to use. These include solar collectors, geothermal probes and chillers. They collect the energy and convert it to a temperature level high enough to be suitable for heat purposes. In the case of refrigeration machines, a distinction is made between electrically and thermally driven machines.
The use of renewable energy sources for heating and cooling in buildings today offers the best opportunities to avoid the use of fossil fuels and the associated climate-damaging emissions. However, unlike fossil fuels, renewable energy sources such as solar radiation are not available at the push of a button, but occur uncontrollably depending on weather conditions, the location of the building and the time of year. Their use is free of charge. However, complex converters and systems usually have to be installed in order to use them. These must be carefully planned and operated in order to avoid unnecessary costs and to generate the maximum possible yield. The regenerative energy systems are usually integrated into existing conventional systems. When designing the control and regulation equipment, it is crucial to design the automation of the systems in such a way that primarily renewable energy sources are used and the share of fossil energy sources is minimized.
This textbook helps use regenerative systems for heating and cooling effectively. Integration and automation schemes provide a quick overview. Practical examples clearly show standard solutions for the integration of regenerative energy sources. For the 2nd edition, improvements have been made to the text and illustrations, and references to standards have been updated. Control questions at the end of the main chapters serve to consolidate the understanding of the content.
Public educational institutions are increasingly confronted with a decline in the number of applicants, which is why competition between colleges and universities is also intensifying. For this reason, it is important to position oneself as an institution in order to be perceived by the various target groups and to differentiate oneself from the competition. In this context, the brand and thus its perception and impact play a decisive role, especially in view of the desired communication of the institution's own values and its self-image, the brand identity. To this end, emotions serve as an approach to creating positive stimulation and brand loyalty.
In this study, circular economy (CE) relevance in Germany will be discussed based on LinkedIn readily available data. LinkedIn company profiles located in Germany with ‘circular economy’ in their description or any other field were selected and used as a data source to analyze their CE relation. Overall, 514 German companies were analyzed in reference to the 15 German regions they belong. Most companies are located in the federal state of Berlin (126), followed by North Rhine-Westphalia (96) and Bavaria (77). In terms of the industry sector, they are self-classified to environmental services (64), management consulting (50), renewables & environment (33), research (31), and computer software (18) etc. Regarding their employees with LinkedIn profiles, 22,621 people are affiliated with these companies, ranging from one to 7,877. All examined companies have a total of 819,632 followers on LinkedIn, ranging from none to 88,167. An increase in CE-related companies was recorded in 13 of the 16 federal states of Germany over a one-year period. This work provides essential insights into the increasing relevance and trends of the circular economy in German enterprises and will help conduct further national studies with readily available data from LinkedIn.
Human interaction frequently includes decision-making processes during which interactants call on verbal and non-verbal resources to manage the flow of interaction. In 2017, Stevanovic et al. carried out pioneering work, analyzing the unfolding of moment-by-moment dynamics by investigating the behavioral matching during search and decision-making phases. By studying the similarities in the participant's body sway during a conversation task in Finnish, the authors showed higher behavioral matching during decision phases than during search phases. The purpose of this research was to investigate the whole-body sway and its coordination during joint search and decision-making phases as a replication of the study by Stevanovic et al. (2017) but based on a German population. Overall, 12 dyads participated in this study and were asked to decide on 8 adjectives, starting with a pre-defined letter, to describe a fictional character. During this joint-decision task (duration: 206.46 ± 116.08 s), body sway of both interactants was measured using a 3D motion capture system and center of mass (COM) accelerations were computed. Matching of body sway was calculated using a windowed cross correlation (WCC) of the COM accelerations. A total of 101 search and 101 decision phases were identified for the 12 dyads. Significant higher COM accelerations (5.4*10−3 vs. 3.7*10−3 mm/s2, p < 0.001) and WCC coefficients (0.47 vs. 0.45, p = 0.043) were found during decision-making phases than during search phases. The results suggest that body sway is one of the resources humans use to communicate the arrival at a joint decision. These findings contribute to a better understanding of interpersonal coordination from a human movement science perspective.