Predictive modeling of blood glucose levels of patients is a key enabler for automated management of diabetes. However, such models are typically executed on battery-operated and resource-constrained edge devices, which limits the deployment feasibility of complex machine learning methods. To alleviate this challenge, this paper proposes a neural architecture search formulation tailored to the problem of blood glucose forecasting. This approach exploits joint Bayesian optimization of task performance and computational cost, thereby producing highly optimized architectures. Experimental results show that the resulting models attain state-of-the-art performance in predicting occurrences of future hyperglycemia and hypoglycemia on the OhioT1DM dataset, with accuracies of 88.28% and 79.44%, respectively. Moreover, the proposed models achieve state-of-the-art performance in blood glucose level estimation for all dataset-prediction horizon combinations, while providing improvements up to 130× and 1000× for memory and processing costs, respectively.

Radar is a key sensing modality to obtain scene understanding in automotive driving applications. However, automotive radars that are based on the widely used frequency-modulated continuous-wave (FMCW) signal waveforms are vulnerable to signal spoofing attacks. These attacks can distort the resulting radar images at the victim radar receiver via a false positive by introducing artificial objects, or due to a false negative by concealing real objects. In this work, we propose a deep learning method to detect radar points compromised by spoofing attacks and to classify the type of attack. Leveraging mid-fusion of radar and camera sensors, as well as a multi-head architecture, our method simultaneously identifies the specific spatial locations where spoofing attacks distort the radar image, which enables the use of undistorted portions of the data for partial functionality. Evaluation on real data with simulated cases of spoofing attacks shows that our approach achieves mean classification F1scores of 0.93 for false positive and 0.82 for false negative attacks, demonstrating its effectiveness in enhancing radar image integrity.

Synthetic generation of radar range-azimuth (RA) maps remains challenging due to limited automotive radar datasets. While camera images are abundant across diverse driving environments, translating visual information directly to radar representations requires sophisticated multi-modal fusion techniques. In this paper, we introduce RAILS, a novel convolutional autoencoder architecture that synthesizes realistic RA maps by integrating RGB camera images, LiDAR depth information, and semantic scene segmentation. By leveraging a U-Net inspired architecture with convolutional block attention mechanisms, our approach transforms multi-modal inputs into accurate radar representations. We demonstrate the model's effectiveness across various driving scenarios using the RADIal dataset, showing superior performance in target localization and scene reconstruction compared to existing methods. Experimental results highlight the potential of using auxiliary depth and semantic information to address the scarcity of radar training data, offering a promising approach for enhancing machine learning-based radar perception in autonomous driving systems.

The scarcity of automotive radar datasets is a hindrance in the adoption of machine learning-based radar solutions for assisted and autonomous driving systems. In contrast, there is a larger availability of camera images under different driving environments. In this paper, we develop a radar point cloud data emulation methodology to estimate the radar point cloud probability density function (PDF) corresponding to co-centered camera images. Our method uses convolutional neural networks (CNNs) trained on camera images and available radar point cloud data to predict a set of Gaussian mixture model (GMM) parameters. The resulting parametric PDF model can then be used to emulate radar point clouds for a given scene, and also to produce a high-density radar point cloud dataset. Qualitative and quantitative evaluation of the proposed method on the RADIal dataset shows promising results in synthesizing realistic radar point clouds from the corresponding camera images.

The vast majority of cardiovascular diseases are avoidable or treatable by preventive measures and early de-tection. To efficiently detect early signs and risk factors, car-diovascular parameters can be monitored continuously with small sensor patches, which improve the comfort of patients. However, processing the sensor data is a challenging task with the demanding needs of robustness, reliability, performance and efficiency. The field of deep learning has tremendous potential to provide a way to analyze cardiovascular sensor data to detect anomalies which alleviates the workload of doctors for more effective data interpretation. In this work, we show the feasibility of applying deep learning for the classification of synchronized electrocardiogram and phonocardiogram recordings under very tight resource constraints. Our model employs an early fusion of data and uses convolutional layers to solve the problem of binary classification of anomalies. Our experiments show that our model matches the accuracy of the current state-of-the-art model on the “training-a” dataset of the Physionet Challenge 2016 database while being more than two orders of magnitude more efficient in memory footprint and compute cost. Further, we demonstrate the applicability of our model on edge devices, such as sensor patches, by estimating processor performance, power consumption, and silicon area.

Abstract Maritime vessel re-identification (re-ID) is a computer vision task of vessel identity matching across disjoint camera views. Prominent applications of vessel re-ID exist in the fields of surveillance and maritime traffic flow analysis. However, the field suffers from the absence of a large-scale dataset that enables training of deep learning models. In this study, we present a new dataset that includes 4614 images of 729 vessels along with 5-bin orientation and 8-class vessel-type annotations to promote further research. A second contribution of this study is the baseline re-ID analysis of our new dataset. Performances of 10 recent deep learning architectures are quantitatively compared to reveal the best practices. Lastly, we propose a novel multi-branch deep learning architecture, Maritime Vessel Re-ID network (MVR-net), to address the challenging problem of vessel re-ID. Evaluation of our approach on the new dataset yields 74.5% mAP and 77.9% Rank-1 score, providing a performance increase of 5.7% mAP and 5.0% Rank-1 over the best-performing baseline. MVR-net also outperforms the PRN (a pioneering vehicle re-ID network), by 2.9% and 4.3% higher mAP and Rank-1, respectively.

Person re-identification (re-ID) is a valuable tool for multi-camera tracking of persons. Up till now, research on person re-ID has mainly focused on the closed-set case, where a given query is assumed to always have a correct match in the gallery set, which does not hold for practical scenarios. In this study, we explore the open-set person re-ID problem with queries not always included in the gallery set. First, we convert the popular closed-set person re-ID datasets into the open-set scenario. Second, we compare the performances of six state-of-the-art closed-set person re-ID methods under open-set conditions. Third, we investigate the impact of a simple and fast statistics-driven gallery refinement approach on the open-set person re-ID performance. Extensive experimental evaluations show that, gallery refinement increases the performance of existing methods in the low false-accept rate (FAR) region, while simultaneously reducing the computational demands of retrieval. Results show an average detection and identification rate (DIR) increase of 7.91% and 3.31% on the DukeMTMC-reID and Market1501 datasets, respectively, for an FAR of 1%.