Oliver Bringmann

Jörg Gamerdinger, Sven Teufel, Simon Roller et al.

The growing number of road users has significantly increased the risk of accidents in recent years. Vulnerable Road Users (VRUs) are particularly at risk, especially in urban environments where they are often occluded by parked vehicles or buildings. Autonomous Driving (AD) and Collective Perception (CP) are promising solutions to mitigate these risks. In particular, infrastructure-assisted CP, where sensor units are mounted on infrastructure elements such as traffic lights or lamp posts, can help overcome perceptual limitations by providing enhanced points of view, which significantly reduces occlusions. To encourage decision makers to adopt this technology, comprehensive studies and datasets demonstrating safety improvements for VRUs are essential. In this paper, we propose a framework for evaluating the safety improvement by infrastructure-based CP specifically targeted at VRUs including a dataset with safety-critical EuroNCAP scenarios (CarlaNCAP) with 11k frames. Using this dataset, we conduct an in-depth simulation study and demonstrate that infrastructure-assisted CP can significantly reduce accident rates in safety-critical scenarios, achieving up to 100% accident avoidance compared to a vehicle equipped with sensors with only 33%. Code is available at https://github.com/ekut-es/carla_ncap

Sven Teufel, Jörg Gamerdinger, Georg Volk et al.

Comprehensive perception of the environment is crucial for the safe operation of autonomous vehicles. However, the perception capabilities of autonomous vehicles are limited due to occlusions, limited sensor ranges, or environmental influences. Collective Perception (CP) aims to mitigate these problems by enabling the exchange of information between vehicles. A major challenge in CP is the fusion of the exchanged information. Due to the enormous bandwidth requirement of early fusion approaches and the interchangeability issues of intermediate fusion approaches, only the late fusion of shared detections is practical. Current late fusion approaches neglect valuable information for local detection, this is why we propose a novel fusion method to fuse the detections of cooperative vehicles within the local LiDAR-based detection pipeline. Therefore, we present Collective PV-RCNN (CPV-RCNN), which extends the PV-RCNN++ framework to fuse collective detections. Code is available at https://github.com/ekut-es

Sven Teufel, Jörg Gamerdinger, Georg Volk et al.

The safe operation of automated vehicles depends on their ability to perceive the environment comprehensively. However, occlusion, sensor range, and environmental factors limit their perception capabilities. To overcome these limitations, collective perception enables vehicles to exchange information. However, fusing this exchanged information is a challenging task. Early fusion approaches require large amounts of bandwidth, while intermediate fusion approaches face interchangeability issues. Late fusion of shared detections is currently the only feasible approach. However, it often results in inferior performance due to information loss. To address this issue, we propose MR3D-Net, a dynamic multi-resolution 3D sparse voxel grid fusion backbone architecture for LiDAR-based collective perception. We show that sparse voxel grids at varying resolutions provide a meaningful and compact environment representation that can adapt to the communication bandwidth. MR3D-Net achieves state-of-the-art performance on the OPV2V 3D object detection benchmark while reducing the required bandwidth by up to 94% compared to early fusion. Code is available at https://github.com/ekut-es/MR3D-Net

Federico Nicolas Peccia, Svetlana Pavlitska, Tobias Fleck et al.

The growing concerns regarding energy consumption and privacy have prompted the development of AI solutions deployable on the edge, circumventing the substantial CO2 emissions associated with cloud servers and mitigating risks related to sharing sensitive data. But deploying Convolutional Neural Networks (CNNs) on non-off-the-shelf edge devices remains a complex and labor-intensive task. In this paper, we present and end-to-end workflow for deployment of CNNs on Field Programmable Gate Arrays (FPGAs) using the Gemmini accelerator, which we modified for efficient implementation on FPGAs. We describe how we leverage the use of open source software on each optimization step of the deployment process, the customizations we added to them and its impact on the final system's performance. We were able to achieve real-time performance by deploying a YOLOv7 model on a Xilinx ZCU102 FPGA with an energy efficiency of 36.5 GOP/s/W. Our FPGA-based solution demonstrates superior power efficiency compared with other embedded hardware devices, and even outperforms other FPGA reference implementations. Finally, we present how this kind of solution can be integrated into a wider system, by testing our proposed platform in a traffic monitoring scenario.

Dennis Rieber, Moritz Reiber, Oliver Bringmann et al.

The process of optimizing the latency of DNN operators with ML models and hardware-in-the-loop, called auto-tuning, has established itself as a pervasive method for the deployment of neural networks. From a search space of loop-optimizations, the candidate providing the best performance has to be selected. Performance of individual configurations is evaluated through hardware measurements. The combinatorial explosion of possible configurations, together with the cost of hardware evaluation makes exhaustive explorations of the search space infeasible in practice. Machine Learning methods, like random forests or reinforcement learning are used to aid in the selection of candidates for hardware evaluation. For general purpose hardware like x86 and GPGPU architectures impressive performance gains can be achieved, compared to hand-optimized libraries like cuDNN. The method is also useful in the space of hardware accelerators with less wide-spread adoption, where a high-performance library is not always available. However, hardware accelerators are often less flexible with respect to their programming which leads to operator configurations not executable on the hardware target. This work evaluates how these invalid configurations affect the auto-tuning process and its underlying performance prediction model for the VTA hardware. From these results, a validity-driven initialization method for AutoTVM is developed, only requiring 41.6% of the necessary hardware measurements to find the best solution, while improving search robustness.

Georg Volk, Jörg Gamerdinger, Alexander von Bernuth et al.

Complete perception of the environment and its correct interpretation is crucial for autonomous vehicles. Object perception is the main component of automotive surround sensing. Various metrics already exist for the evaluation of object perception. However, objects can be of different importance depending on their velocity, orientation, distance, size, or the potential damage that could be caused by a collision due to a missed detection. Thus, these additional parameters have to be considered for safety evaluation. We propose a new safety metric that incorporates all these parameters and returns a single easily interpretable safety assessment score for object perception. This new metric is evaluated with both real world and virtual data sets and compared to state of the art metrics.

Jörg Gamerdinger, Sven Teufel, Stephan Amann et al.

Ensuring safety is the primary objective of automated driving, which necessitates a comprehensive and accurate perception of the environment. While numerous performance evaluation metrics exist for assessing perception capabilities, incorporating safety-specific metrics is essential to reliably evaluate object detection systems. A key component for safety evaluation is the ability to distinguish between relevant and non-relevant objects - a challenge addressed by criticality or relevance metrics. This paper presents the first in-depth analysis of criticality metrics for safety evaluation of object detection systems. Through a comprehensive review of existing literature, we identify and assess a range of applicable metrics. Their effectiveness is empirically validated using the DeepAccident dataset, which features a variety of safety-critical scenarios. To enhance evaluation accuracy, we propose two novel application strategies: bidirectional criticality rating and multi-metric aggregation. Our approach demonstrates up to a 100% improvement in terms of criticality classification accuracy, highlighting its potential to significantly advance the safety evaluation of object detection systems in automated vehicles.

Jörg Gamerdinger, Sven Teufel, Patrick Schulz et al.

Collective perception has received considerable attention as a promising approach to overcome occlusions and limited sensing ranges of vehicle-local perception in autonomous driving. In order to develop and test novel collective perception technologies, appropriate datasets are required. These datasets must include not only different environmental conditions, as they strongly influence the perception capabilities, but also a wide range of scenarios with different road users as well as realistic sensor models. Therefore, we propose the Synthetic COllective PErception (SCOPE) dataset. SCOPE is the first synthetic multi-modal dataset that incorporates realistic camera and LiDAR models as well as parameterized and physically accurate weather simulations for both sensor types. The dataset contains 17,600 frames from over 40 diverse scenarios with up to 24 collaborative agents, infrastructure sensors, and passive traffic, including cyclists and pedestrians. In addition, recordings from two novel digital-twin maps from Karlsruhe and Tübingen are included. The dataset is available at https://ekut-es.github.io/scope

Jörg Gamerdinger, Sven Teufel, Stephan Amann et al.

Comprehensive perception of the vehicle's environment and correct interpretation of the environment are crucial for the safe operation of autonomous vehicles. The perception of surrounding objects is the main component for further tasks such as trajectory planning. However, safe trajectory planning requires not only object detection, but also the detection of drivable areas and lane corridors. While first approaches consider an advanced safety evaluation of object detection, the evaluation of lane detection still lacks sufficient safety metrics. Similar to the safety metrics for object detection, additional factors such as the semantics of the scene with road type and road width, the detection range as well as the potential causes of missing detections, incorporated by vehicle speed, should be considered for the evaluation of lane detection. Therefore, we propose the Lane Safety Metric (LSM), which takes these factors into account and allows to evaluate the safety of lane detection systems by determining an easily interpretable safety score. We evaluate our offline safety metric on various virtual scenarios using different lane detection approaches and compare it with state-of-the-art performance metrics.

Julia Werner, Christoph Gerum, Moritz Reiber et al.

This paper presents a method to efficiently classify the gastroenterologic section of images derived from Video Capsule Endoscopy (VCE) studies by exploring the combination of a Convolutional Neural Network (CNN) for classification with the time-series analysis properties of a Hidden Markov Model (HMM). It is demonstrated that successive time-series analysis identifies and corrects errors in the CNN output. Our approach achieves an accuracy of $98.04\%$ on the Rhode Island (RI) Gastroenterology dataset. This allows for precise localization within the gastrointestinal (GI) tract while requiring only approximately 1M parameters and thus, provides a method suitable for low power devices

Jörg Gamerdinger, Benedict Wetzel, Patrick Schulz et al.

Lane detection for autonomous driving in snow-covered environments remains a major challenge due to the frequent absence or occlusion of lane markings. In this paper, we present a novel, robust and realtime capable approach that bypasses the reliance on traditional lane markings by detecting roadside features,specifically vertical roadside posts called delineators, as indirect lane indicators. Our method first perceives these posts, then fits a smooth lane trajectory using a parameterized Bezier curve model, leveraging spatial consistency and road geometry. To support training and evaluation in these challenging scenarios, we introduce SnowyLane, a new synthetic dataset containing 80,000 annotated frames capture winter driving conditions, with varying snow coverage, and lighting conditions. Compared to state-of-the-art lane detection systems, our approach demonstrates significantly improved robustness in adverse weather, particularly in cases with heavy snow occlusion. This work establishes a strong foundation for reliable lane detection in winter scenarios and contributes a valuable resource for future research in all-weather autonomous driving. The dataset is available at https://ekut-es.github.io/snowy-lane

Julia Werner, Julius Oexle, Oliver Bause et al.

The classification performance of deep neural networks relies strongly on access to large, accurately annotated datasets. In medical imaging, however, obtaining such datasets is particularly challenging since annotations must be provided by specialized physicians, which severely limits the pool of annotators. Furthermore, class boundaries can often be ambiguous or difficult to define which further complicates machine learning-based classification. In this paper, we want to address this problem and introduce a framework for mislabel detection in medical datasets. This is validated on the two largest, publicly available datasets for Video Capsule Endoscopy, an important imaging procedure for examining the gastrointestinal tract based on a video stream of lowresolution images. In addition, potentially mislabeled samples identified by our pipeline were reviewed and re-annotated by three experienced gastroenterologists. Our results show that the proposed framework successfully detects incorrectly labeled data and results in an improved anomaly detection performance after cleaning the datasets compared to current baselines.

Federico Nicolas Peccia, Frederik Haxel, Oliver Bringmann

RISC-V provides a flexible and scalable platform for applications ranging from embedded devices to high-performance computing clusters. Particularly, its RISC-V Vector Extension (RVV) becomes of interest for the acceleration of AI workloads. But writing software that efficiently utilizes the vector units of RISC-V CPUs without expert knowledge requires the programmer to rely on the autovectorization features of compilers or hand-crafted libraries like muRISCV-NN. Smarter approaches, like autotuning frameworks, have been missing the integration with the RISC-V RVV extension, thus heavily limiting the efficient deployment of complex AI workloads. In this paper, we present a workflow based on the TVM compiler to efficiently map AI workloads onto RISC-V vector units. Instead of relying on hand-crafted libraries, we integrated the RVV extension into TVM's MetaSchedule framework, a probabilistic program framework for tensor operation tuning. We implemented different RISC-V SoCs on an FPGA and tuned a wide range of AI workloads on them. We found that our proposal shows a mean improvement of 46% in execution latency when compared against the autovectorization feature of GCC, and 29% against muRISCV-NN. Moreover, the binary resulting from our proposal has a smaller code memory footprint, making it more suitable for embedded devices. Finally, we also evaluated our solution on a commercially available RISC-V SoC implementing the RVV 1.0 Vector Extension and found our solution is able to find mappings that are 35% faster on average than the ones proposed by LLVM. We open-sourced our proposal for the community to expand it to target other RISC-V extensions.

Jörg Gamerdinger, Sven Teufel, Stephan Amann et al.

Extensive evaluation of perception systems is crucial for ensuring the safety of intelligent vehicles in complex driving scenarios. Conventional performance metrics such as precision, recall and the F1-score assess the overall detection accuracy, but they do not consider the safety-relevant aspects of perception. Consequently, perception systems that achieve high scores in these metrics may still cause misdetections that could lead to severe accidents. Therefore, it is important to evaluate not only the overall performance of perception systems, but also their safety. We therefore introduce a novel safety metric for jointly evaluating the most critical perception tasks, object and lane detection. Our proposed framework integrates a new, lightweight object safety metric that quantifies the potential risk associated with object detection errors, as well as an lane safety metric including the interdependence between both tasks that can occur in safety evaluation. The resulting combined safety score provides a unified, interpretable measure of perception safety performance. Using the DeepAccident dataset, we demonstrate that our approach identifies safety critical perception errors that conventional performance metrics fail to capture. Our findings emphasize the importance of safety-centric evaluation methods for perception systems in autonomous driving.

Konstantin Lübeck, Alexander Louis-Ferdinand Jung, Felix Wedlich et al.

Implementing Deep Neural Networks (DNNs) on resource-constrained edge devices is a challenging task that requires tailored hardware accelerator architectures and a clear understanding of their performance characteristics when executing the intended AI workload. To facilitate this, we present an automated generation approach for fast performance models to accurately estimate the latency of a DNN mapped onto systematically modeled and concisely described accelerator architectures. Using our accelerator architecture description method, we modeled representative DNN accelerators such as Gemmini, UltraTrail, Plasticine-derived, and a parameterizable systolic array. Together with DNN mappings for those modeled architectures, we perform a combined DNN/hardware dependency graph analysis, which enables us, in the best case, to evaluate only 154 loop kernel iterations to estimate the performance for 4.19 billion instructions achieving a significant speedup. We outperform regression and analytical models in terms of mean absolute percentage error (MAPE) compared to simulation results, while being several magnitudes faster than an RTL simulation.

Julia Werner, Christoph Gerum, Jorg Nick et al.

Capsule endoscopy is a method to capture images of the gastrointestinal tract and screen for diseases which might remain hidden if investigated with standard endoscopes. Due to the limited size of a video capsule, embedding AI models directly into the capsule demands careful consideration of the model size and thus complicates anomaly detection in this field. Furthermore, the scarcity of available data in this domain poses an ongoing challenge to achieving effective anomaly detection. Thus, this work introduces an ensemble strategy to address this challenge in anomaly detection tasks in video capsule endoscopies, requiring only a small number of individual neural networks during both the training and inference phases. Ensemble learning combines the predictions of multiple independently trained neural networks. This has shown to be highly effective in enhancing both the accuracy and robustness of machine learning models. However, this comes at the cost of higher memory usage and increased computational effort, which quickly becomes prohibitive in many real-world applications. Instead of applying the same training algorithm to each individual network, we propose using various loss functions, drawn from the anomaly detection field, to train each network. The methods are validated on the two largest publicly available datasets for video capsule endoscopy images, the Galar and the Kvasir-Capsule dataset. We achieve an AUC score of 76.86% on the Kvasir-Capsule and an AUC score of 76.98% on the Galar dataset. Our approach outperforms current baselines with significantly fewer parameters across all models, which is a crucial step towards incorporating artificial intelligence into capsule endoscopies.

Sven Teufel, Jörg Gamerdinger, Oliver Bringmann

Collective Perception (CP) has emerged as a promising approach to overcome the limitations of individual perception in the context of autonomous driving. Various approaches have been proposed to realize collective perception; however, the Sensor2Sensor domain gap that arises from the utilization of different sensor systems in Connected and Automated Vehicles (CAVs) remains mostly unaddressed. This is primarily due to the paucity of datasets containing heterogeneous sensor setups among the CAVs. The recently released SCOPE datasets address this issue by providing data from three different LiDAR sensors for each CAV. This study is the first to address the Sensor2Sensor domain gap in vehicle-to-vehicle (V2V) collective perception. First, we present our sensor-domain robust architecture S2S-Net. Then an in-depth analysis of the Sensor2Sensor domain adaptation capabilities of state-of-the-art CP methods and S2S-Net is conducted on the SCOPE dataset. This study shows that, all evaluated state-of-the-art mehtods for collective perception highly suffer from the Sensor2Sensor domain gap, while S2S-Net demonstrates the capability to maintain very high performance in unseen sensor domains and outperforms the evaluated state-of-the-art methods by up to 44 percentage points.

Oliver Bause, Paul Palomero Bernardo, Oliver Bringmann

As machine learning applications continue to evolve, the demand for efficient hardware accelerators, specifically tailored for deep neural networks (DNNs), becomes increasingly vital. In this paper, we propose a configurable memory hierarchy framework tailored for per layer adaptive memory access patterns of DNNs. The hierarchy requests data on-demand from the off-chip memory to provide it to the accelerator's compute units. The objective is to strike an optimized balance between minimizing the required memory capacity and maintaining high accelerator performance. The framework is characterized by its configurability, allowing the creation of a tailored memory hierarchy with up to five levels. Furthermore, the framework incorporates an optional shift register as final level to increase the flexibility of the memory management process. A comprehensive loop-nest analysis of DNN layers shows that the framework can efficiently execute the access patterns of most loop unrolls. Synthesis results and a case study of the DNN accelerator UltraTrail indicate a possible reduction in chip area of up to 62.2% as smaller memory modules can be used. At the same time, the performance loss can be minimized to 2.4%.

Pascal Zwick, Kevin Roesch, Marvin Klemp et al.

Anonymization plays a key role in protecting sensible information of individuals in real world datasets. Self-driving cars for example need high resolution facial features to track people and their viewing direction to predict future behaviour and react accordingly. In order to protect people's privacy whilst keeping important features in the dataset, it is important to replace the full body of a person with a highly detailed anonymized one. In contrast to doing face anonymization, full body replacement decreases the ability of recognizing people by their hairstyle or clothes. In this paper, we propose a workflow for full body person anonymization utilizing Stable Diffusion as a generative backend. Text-to-image diffusion models, like Stable Diffusion, OpenAI's DALL-E or Midjourney, have become very popular in recent time, being able to create photorealistic images from a single text prompt. We show that our method outperforms state-of-the art anonymization pipelines with respect to image quality, resolution, Inception Score (IS) and Frechet Inception Distance (FID). Additionally, our method is invariant with respect to the image generator and thus able to be used with the latest models available.

Mika Markus Müller, Alexander Richard Manfred Borst, Konstantin Lübeck et al.

Artificial Intelligence (AI) has witnessed remarkable growth, particularly through the proliferation of Deep Neural Networks (DNNs). These powerful models drive technological advancements across various domains. However, to harness their potential in real-world applications, specialized hardware accelerators are essential. This demand has sparked a market for parameterizable AI hardware accelerators offered by different vendors. Manufacturers of AI-integrated products face a critical challenge: selecting an accelerator that aligns with their product's performance requirements. The decision involves choosing the right hardware and configuring a suitable set of parameters. However, comparing different accelerator design alternatives remains a complex task. Often, engineers rely on data sheets, spreadsheet calculations, or slow black-box simulators, which only offer a coarse understanding of the performance characteristics. The Abstract Computer Architecture Description Language (ACADL) is a concise formalization of computer architecture block diagrams, which helps to communicate computer architecture on different abstraction levels and allows for inferring performance characteristics. In this paper, we demonstrate how to use the ACADL to model AI hardware accelerators, use their ACADL description to map DNNs onto them, and explain the timing simulation semantics to gather performance results.

Julia Werner, Oliver Bause, Julius Oexle et al.

Video capsule endoscopy has become increasingly important for investigating the small intestine within the gastrointestinal tract. However, a persistent challenge remains the short battery lifetime of such compact sensor edge devices. Integrating artificial intelligence can help overcome this limitation by enabling intelligent real-time decision-making, thereby reducing the energy consumption and prolonging the battery life. However, this remains challenging due to data sparsity and the limited resources of the device restricting the overall model size. In this work, we introduce a multi-task neural network that combines the functionalities of precise self-localization within the gastrointestinal tract with the ability to detect anomalies in the small intestine within a single model. Throughout the development process, we consistently restricted the total number of parameters to ensure the feasibility to deploy such model in a small capsule. We report the first multi-task results using the recently published Galar dataset, integrating established multi-task methods and Viterbi decoding for subsequent time-series analysis. This outperforms current single-task models and represents a significant advance in AI-based approaches in this field. Our model achieves an accuracy of 93.63% on the localization task and an accuracy of 87.48% on the anomaly detection task. The approach requires only 1 million parameters while surpassing the current baselines.

Oliver Bause, Julia Werner, Paul Palomero Bernardo et al.

For many real-world applications involving low-power sensor edge devices deep neural networks used for image classification might not be suitable. This is due to their typically large model size and require- ment of operations often exceeding the capabilities of such resource lim- ited devices. Furthermore, camera sensors usually capture images with a Bayer color filter applied, which are subsequently converted to RGB images that are commonly used for neural network training. However, on resource-constrained devices, such conversions demands their share of energy and optimally should be skipped if possible. This work ad- dresses the need for hardware-suitable AI targeting sensor edge devices by means of the Video Capsule Endoscopy, an important medical proce- dure for the investigation of the small intestine, which is strongly limited by its battery lifetime. Accurate organ classification is performed with a final accuracy of 93.06% evaluated directly on Bayer images involv- ing a CNN with only 63,000 parameters and time-series analysis in the form of Viterbi decoding. Finally, the process of capturing images with a camera and raw image processing is demonstrated with a customized PULPissimo System-on-Chip with a RISC-V core and an ultra-low power hardware accelerator providing an energy-efficient AI-based image clas- sification approach requiring just 5.31 μJ per image. As a result, it is possible to save an average of 89.9% of energy before entering the small intestine compared to classic video capsules.

Sven Teufel, Dominique Mayer, Jörg Gamerdinger et al.

While automated vehicles hold the potential to significantly reduce traffic accidents, their perception systems remain vulnerable to sensor degradation caused by adverse weather and environmental occlusions. Collective perception, which enables vehicles to share information, offers a promising approach to overcoming these limitations. However, to this date collective perception in adverse weather is mostly unstudied. Therefore, we conduct the first study of LiDAR-based collective perception under diverse weather conditions and present a novel multi-task architecture for LiDAR-based collective perception under adverse weather. Adverse weather conditions can not only degrade perception capabilities, but also negatively affect bandwidth requirements and latency due to the introduced noise that is also transmitted and processed. Denoising prior to communication can effectively mitigate these issues. Therefore, we propose DenoiseCP-Net, a novel multi-task architecture for LiDAR-based collective perception under adverse weather conditions. DenoiseCP-Net integrates voxel-level noise filtering and object detection into a unified sparse convolution backbone, eliminating redundant computations associated with two-stage pipelines. This design not only reduces inference latency and computational cost but also minimizes communication overhead by removing non-informative noise. We extended the well-known OPV2V dataset by simulating rain, snow, and fog using our realistic weather simulation models. We demonstrate that DenoiseCP-Net achieves near-perfect denoising accuracy in adverse weather, reduces the bandwidth requirements by up to 23.6% while maintaining the same detection accuracy and reducing the inference latency for cooperative vehicles.

Christian Reichenbächer, Philipp Rank, Jochen Hipp et al.

This paper presents the first application of Gaussian Mixture Copula Models to the statistical modeling of driving scenarios for the safety validation of automated driving systems. Knowledge of the joint probability distribution of scenario parameters is essential for scenario-based safety assessment, where risk quantification depends on the likelihood of concrete parameter combinations. Gaussian Mixture Copula Models bring together the multimodal expressivity of Gaussian Mixture Models and the flexibility of copulas, enabling separate modeling of marginal distributions and dependencies. We benchmark Gaussian Mixture Copula Models against previously proposed approaches - Gaussian Mixture Models and Gaussian Copula Models - using real-world driving data drawn from scenarios defined in United Nations Regulation No. 157. Our evaluation across approximately 18 million scenario instances demonstrates that Gaussian Mixture Copula Models consistently surpass Gaussian Copula Models and perform better than, or at least comparably to, Gaussian Mixture Models, as measured by both log-likelihood and Sinkhorn distance. These results are promising for the adoption of Gaussian Mixture Copula Models as a statistical foundation for future scenario-based validation frameworks.

Mika Markus Müller, Konstantin Lübeck, Alexander Louis-Ferdinand Jung et al.

Artificial Intelligence (AI) algorithms, such as Deep Neural Networks (DNNs), have become an important tool for a wide range of applications, from computer vision to natural language processing. However, the computational complexity of DNN inference poses a significant challenge, particularly for processing on resource-constrained edge devices. One promising approach to address this challenge is the exploitation of sparsity in DNN operator weights. In this work, we present FlexiSAGA, an architecturally configurable and dataflow-flexible AI hardware accelerator for the sparse and dense processing of general matrix multiplications (GEMMs). FlexiSAGA supports seven different sparse and dense dataflows, enabling efficient processing of resource intensive DNN operators. Additionally, we propose a DNN pruning method specifically tailored towards the FlexiSAGA architecture, allowing for near-optimal processing of dense and sparse convolution and fully-connected operators, facilitating a DNN/HW co-design flow. Our results show a whole DNN sparse-over-dense inference speedup ranging from 1.41 up to 4.28, outperforming commercial and literature-reported accelerator platforms.

Pascal Zwick, Nils Friederich, Maximilian Beichter et al.

Enhancing the efficiency of high-quality image generation using Diffusion Models (DMs) is a significant challenge due to the iterative nature of the process. Flow Matching (FM) is emerging as a powerful generative modeling paradigm based on a simulation-free training objective instead of a score-based one used in DMs. Typical FM approaches rely on a Gaussian distribution prior, which induces curved, conditional probability paths between the prior and target data distribution. These curved paths pose a challenge for the Ordinary Differential Equation (ODE) solver, requiring a large number of inference calls to the flow prediction network. To address this issue, we present Learned Distribution-guided Flow Matching (LeDiFlow), a novel scalable method for training FM-based image generation models using a better-suited prior distribution learned via a regression-based auxiliary model. By initializing the ODE solver with a prior closer to the target data distribution, LeDiFlow enables the learning of more computationally tractable probability paths. These paths directly translate to fewer solver steps needed for high-quality image generation at inference time. Our method utilizes a State-Of-The-Art (SOTA) transformer architecture combined with latent space sampling and can be trained on a consumer workstation. We empirically demonstrate that LeDiFlow remarkably outperforms the respective FM baselines. For instance, when operating directly on pixels, our model accelerates inference by up to 3.75x compared to the corresponding pixel-space baseline. Simultaneously, our latent FM model enhances image quality on average by 1.32x in CLIP Maximum Mean Discrepancy (CMMD) metric against its respective baseline.

Jörg Gamerdinger, Sven Teufel, Oliver Bringmann

Accurate lane detection is essential for automated driving, enabling safe and reliable vehicle navigation in a variety of road scenarios. Numerous datasets have been introduced to support the development and evaluation of lane detection algorithms, each differing in terms of the amount of data, sensor types, annotation granularity, environmental conditions, and scenario diversity. This paper provides a comprehensive review of over 30 publicly available lane detection datasets, systematically analysing their characteristics, advantages and limitations. We classify these datasets based on key factors such as sensor resolution, annotation types and diversity of road and weather conditions. By identifying existing challenges and research gaps, we highlight opportunities for future dataset improvements that can further drive innovation in robust lane detection. This survey serves as a resource for researchers seeking appropriate datasets for lane detection, and contributes to the broader goal of advancing autonomous driving.

Evgenia Rusak, Patrik Reizinger, Attila Juhos et al.

Prior theory work on Contrastive Learning via the InfoNCE loss showed that, under certain assumptions, the learned representations recover the ground-truth latent factors. We argue that these theories overlook crucial aspects of how CL is deployed in practice. Specifically, they either assume equal variance across all latents or that certain latents are kept invariant. However, in practice, positive pairs are often generated using augmentations such as strong cropping to just a few pixels. Hence, a more realistic assumption is that all latent factors change with a continuum of variability across all factors. We introduce AnInfoNCE, a generalization of InfoNCE that can provably uncover the latent factors in this anisotropic setting, broadly generalizing previous identifiability results in CL. We validate our identifiability results in controlled experiments and show that AnInfoNCE increases the recovery of previously collapsed information in CIFAR10 and ImageNet, albeit at the cost of downstream accuracy. Finally, we discuss the remaining mismatches between theoretical assumptions and practical implementations.

Julia Werner, Bhavya Kohli, Paul Palomero Bernardo et al.

Epilepsy is the most common, chronic, neurological disease worldwide and is typically accompanied by reoccurring seizures. Neuro implants can be used for effective treatment by suppressing an upcoming seizure upon detection. Due to the restricted size and limited battery lifetime of those medical devices, the employed approach also needs to be limited in size and have low energy requirements. We present an energy-efficient seizure detection approach involving a TC-ResNet and time-series analysis which is suitable for low-power edge devices. The presented approach allows for accurate seizure detection without preceding feature extraction while considering the stringent hardware requirements of neural implants. The approach is validated using the CHB-MIT Scalp EEG Database with a 32-bit floating point model and a hardware suitable 4-bit fixed point model. The presented method achieves an accuracy of 95.28%, a sensitivity of 92.34% and an AUC score of 0.9384 on this dataset with 4-bit fixed point representation. Furthermore, the power consumption of the model is measured with the low-power AI accelerator UltraTrail, which only requires 495 nW on average. Due to this low-power consumption this classification approach is suitable for real-time seizure detection on low-power wearable devices such as neural implants.

Alexander Louis-Ferdinand Jung, Jannik Steinmetz, Jonathan Gietz et al.

Statistical models are widely used to estimate the performance of commercial off-the-shelf (COTS) AI hardware accelerators. However, training of statistical performance models often requires vast amounts of data, leading to a significant time investment and can be difficult in case of limited hardware availability. To alleviate this problem, we propose a novel performance modeling methodology that significantly reduces the number of training samples while maintaining good accuracy. Our approach leverages knowledge of the target hardware architecture and initial parameter sweeps to identify a set of Performance Representatives (PR) for deep neural network (DNN) layers. These PRs are then used for benchmarking, building a statistical performance model, and making estimations. This targeted approach drastically reduces the number of training samples needed, opposed to random sampling, to achieve a better estimation accuracy. We achieve a Mean Absolute Percentage Error (MAPE) of as low as 0.02% for single-layer estimations and 0.68% for whole DNN estimations with less than 10000 training samples. The results demonstrate the superiority of our method for single-layer estimations compared to models trained with randomly sampled datasets of the same size.

Anwesh Mohanty, Adrian Frischknecht, Christoph Gerum et al.

Keyword spotting (KWS) is becoming a ubiquitous need with the advancement in artificial intelligence and smart devices. Recent work in this field have focused on several different architectures to achieve good results on datasets with low to moderate noise. However, the performance of these models deteriorates under high noise conditions as shown by our experiments. In our paper, we present an extensive comparison between state-of-the-art KWS networks under various noisy conditions. We also suggest adaptive batch normalization as a technique to improve the performance of the networks when the noise files are unknown during the training phase. The results of such high noise characterization enable future work in developing models that perform better in the aforementioned conditions.

Sebastian Reiter, Marc Zeller, Kai Hoefig et al.

The growing complexity of safety-relevant systems causes an increasing effort for safety assurance. The reduction of development costs and time-to-market, while guaranteeing safe operation, is therefore a major challenge. In order to enable efficient safety assessment of complex architectures, we present an approach, which combines deductive safety analyses, in form of Component Fault Trees (CFTs), with an Error Effect Simulation (EES) for sanity checks. The combination reduces the drawbacks of both analyses, such as the subjective failure propagation assumptions in the CFTs or the determination of relevant fault scenarios for the EES. Both CFTs and the EES provide a modular, reusable and compositional safety analysis and are applicable throughout the whole design process. They support continuous model refinement and the reuse of conducted safety analysis and simulation models. Hence, safety goal violations can be identified in early design stages and the reuse of conducted safety analyses reduces the overhead for safety assessment.

Evgenia Rusak, Steffen Schneider, George Pachitariu et al.

We demonstrate that self-learning techniques like entropy minimization and pseudo-labeling are simple and effective at improving performance of a deployed computer vision model under systematic domain shifts. We conduct a wide range of large-scale experiments and show consistent improvements irrespective of the model architecture, the pre-training technique or the type of distribution shift. At the same time, self-learning is simple to use in practice because it does not require knowledge or access to the original training data or scheme, is robust to hyperparameter choices, is straight-forward to implement and requires only a few adaptation epochs. This makes self-learning techniques highly attractive for any practitioner who applies machine learning algorithms in the real world. We present state-of-the-art adaptation results on CIFAR10-C (8.5% error), ImageNet-C (22.0% mCE), ImageNet-R (17.4% error) and ImageNet-A (14.8% error), theoretically study the dynamics of self-supervised adaptation methods and propose a new classification dataset (ImageNet-D) which is challenging even with adaptation.

Steffen Schneider, Evgenia Rusak, Luisa Eck et al.

Today's state-of-the-art machine vision models are vulnerable to image corruptions like blurring or compression artefacts, limiting their performance in many real-world applications. We here argue that popular benchmarks to measure model robustness against common corruptions (like ImageNet-C) underestimate model robustness in many (but not all) application scenarios. The key insight is that in many scenarios, multiple unlabeled examples of the corruptions are available and can be used for unsupervised online adaptation. Replacing the activation statistics estimated by batch normalization on the training set with the statistics of the corrupted images consistently improves the robustness across 25 different popular computer vision models. Using the corrected statistics, ResNet-50 reaches 62.2% mCE on ImageNet-C compared to 76.7% without adaptation. With the more robust DeepAugment+AugMix model, we improve the state of the art achieved by a ResNet50 model up to date from 53.6% mCE to 45.4% mCE. Even adapting to a single sample improves robustness for the ResNet-50 and AugMix models, and 32 samples are sufficient to improve the current state of the art for a ResNet-50 architecture. We argue that results with adapted statistics should be included whenever reporting scores in corruption benchmarks and other out-of-distribution generalization settings.

Evgenia Rusak, Lukas Schott, Roland S. Zimmermann et al.

The human visual system is remarkably robust against a wide range of naturally occurring variations and corruptions like rain or snow. In contrast, the performance of modern image recognition models strongly degrades when evaluated on previously unseen corruptions. Here, we demonstrate that a simple but properly tuned training with additive Gaussian and Speckle noise generalizes surprisingly well to unseen corruptions, easily reaching the previous state of the art on the corruption benchmark ImageNet-C (with ResNet50) and on MNIST-C. We build on top of these strong baseline results and show that an adversarial training of the recognition model against uncorrelated worst-case noise distributions leads to an additional increase in performance. This regularization can be combined with previously proposed defense methods for further improvement.

Claudio Michaelis, Benjamin Mitzkus, Robert Geirhos et al.

The ability to detect objects regardless of image distortions or weather conditions is crucial for real-world applications of deep learning like autonomous driving. We here provide an easy-to-use benchmark to assess how object detection models perform when image quality degrades. The three resulting benchmark datasets, termed Pascal-C, Coco-C and Cityscapes-C, contain a large variety of image corruptions. We show that a range of standard object detection models suffer a severe performance loss on corrupted images (down to 30--60\% of the original performance). However, a simple data augmentation trick---stylizing the training images---leads to a substantial increase in robustness across corruption type, severity and dataset. We envision our comprehensive benchmark to track future progress towards building robust object detection models. Benchmark, code and data are publicly available.