Artificial Intelligence in Sensors, 2nd Volume

Journal Name	Impact Factor	CiteScore	Launched Year	First Decision (median)	APC
Applied Sciences applsci	2.5	5.3	2011	18.4 Days	CHF 2400
Drones drones	4.4	5.6	2017	19.2 Days	CHF 2600
Electronics electronics	2.6	5.3	2012	16.4 Days	CHF 2400
Remote Sensing remotesensing	4.2	8.3	2009	23.9 Days	CHF 2700
Sensors sensors	3.4	7.3	2001	18.6 Days	CHF 2600
Technologies technologies	4.2	6.7	2013	21.1 Days	CHF 1600

24 pages, 4336 KiB

Open AccessArticle

Use of a Residual Neural Network to Demonstrate Feasibility of Ship Detection Based on Synthetic Aperture Radar Raw Data

by Giorgio Cascelli, Cataldo Guaragnella, Raffaele Nutricato, Khalid Tijani, Alberto Morea, Nicolò Ricciardi and Davide Oscar Nitti

Technologies 2023, 11(6), 178; https://doi.org/10.3390/technologies11060178 - 11 Dec 2023

Cited by 6 | Viewed by 2311

Abstract

Synthetic Aperture Radar (SAR) is a well-established 2D imaging technique employed as a consolidated practice in several oil spill monitoring services. In this scenario, onboard detection undoubtedly represents an interesting solution to reduce the latency of these services, also enabling transmission to the [...] Read more.

Synthetic Aperture Radar (SAR) is a well-established 2D imaging technique employed as a consolidated practice in several oil spill monitoring services. In this scenario, onboard detection undoubtedly represents an interesting solution to reduce the latency of these services, also enabling transmission to the ground segment of alert signals with a notable reduction in the required downlink bandwidth. However, the reduced computational capabilities available onboard require alternative approaches with respect to the standard processing flows. In this work, we propose a feasibility study of oil spill detection applied directly to raw data, which is a solution not sufficiently addressed in the literature that has the advantage of not requiring the execution of the focusing step. The study is concentrated only on the accuracy of detection, while computational cost analysis is not within the scope of this work. More specifically, we propose a complete framework based on the use of a Residual Neural Network (ResNet), including a simple and automatic simulation method for generating the training data set. The final tests with ERS real data demonstrate the feasibility of the proposed approach showing that the trained ResNet correctly detects ships with a Signal-to-Clutter Ratio (SCR) > 10.3 dB. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

36 pages, 4604 KiB

Open AccessArticle

Use of Domain Labels during Pre-Training for Domain-Independent WiFi-CSI Gesture Recognition

by Bram van Berlo, Richard Verhoeven and Nirvana Meratnia

Sensors 2023, 23(22), 9233; https://doi.org/10.3390/s23229233 - 16 Nov 2023

Cited by 2 | Viewed by 1650

Abstract

To minimize dependency on the availability of data labels, some WiFi-CSI based-gesture recognition solutions utilize an unsupervised representation learning phase prior to fine-tuning downstream task classifiers. In this case, however, the overall performance of the solution is negatively affected by domain factors present [...] Read more.

To minimize dependency on the availability of data labels, some WiFi-CSI based-gesture recognition solutions utilize an unsupervised representation learning phase prior to fine-tuning downstream task classifiers. In this case, however, the overall performance of the solution is negatively affected by domain factors present in the WiFi-CSI data used by the pre-training models. To reduce this negative effect, we propose an integration of the adversarial domain classifier in the pre-training phase. We consider this as an effective step towards automatic domain discovery during pre-training. We also experiment with multi-class and label versions of domain classification to improve situations, in which integrating a multi-class and single label-based domain classifier during pre-training fails to reduce the negative impact domain factors have on overall solution performance. For our extensive random and leave-out domain factor cross-validation experiments, we utilise (i) an end-to-end and unsupervised representation learning baseline, (ii) integration of both single- and multi-label domain classification, and (iii) so-called domain-aware versions of the aformentioned unsupervised representation learning baseline in (i) with two different datasets, i.e., Widar3 and SignFi. We also consider an input sample type that generalizes, in terms of overall solution performance, to both aforementioned datasets. Experiment results with the Widar3 dataset indicate that multi-label domain classification reduces domain shift in position (1.2% mean metric improvement and 0.5% variance increase) and orientation (0.4% mean metric improvement and 1.0% variance decrease) in domain factor leave-out cross-validation experiments. The results also indicate that domain shift reduction, when considering single- or multi-label domain classification during pre-training, is negatively impacted when a large proportion of negative view combinations contain views that originate from different domains within a substantial amount of mini-batches considered during pre-training. This is caused by the view contrastive loss repelling the aforementioned negative view combinations, eventually causing more domain shift in the intermediate feature space of the overall solution. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

15 pages, 1987 KiB

Open AccessArticle

RST: Rough Set Transformer for Point Cloud Learning

by Xinwei Sun and Kai Zeng

Sensors 2023, 23(22), 9042; https://doi.org/10.3390/s23229042 - 8 Nov 2023

Viewed by 1615

Abstract

Point cloud data generated by LiDAR sensors play a critical role in 3D sensing systems, with applications encompassing object classification, part segmentation, and point cloud recognition. Leveraging the global learning capacity of dot product attention, transformers have recently exhibited outstanding performance in point [...] Read more.

Point cloud data generated by LiDAR sensors play a critical role in 3D sensing systems, with applications encompassing object classification, part segmentation, and point cloud recognition. Leveraging the global learning capacity of dot product attention, transformers have recently exhibited outstanding performance in point cloud learning tasks. Nevertheless, existing transformer models inadequately address the challenges posed by uncertainty features in point clouds, which can introduce errors in the dot product attention mechanism. In response to this, our study introduces a novel global guidance approach to tolerate uncertainty and provide a more reliable guidance. We redefine the granulation and lower-approximation operators based on neighborhood rough set theory. Furthermore, we introduce a rough set-based attention mechanism tailored for point cloud data and present the rough set transformer (RST) network. Our approach utilizes granulation concepts derived from token clusters, enabling us to explore relationships between concepts from an approximation perspective, rather than relying on specific dot product functions. Empirically, our work represents the pioneering fusion of rough set theory and transformer networks for point cloud learning. Our experimental results, including point cloud classification and segmentation tasks, demonstrate the superior performance of our method. Our method establishes concepts based on granulation generated from clusters of tokens. Subsequently, relationships between concepts can be explored from an approximation perspective, instead of relying on specific dot product or addition functions. Empirically, our work represents the pioneering fusion of rough set theory and transformer networks for point cloud learning. Our experimental results, including point cloud classification and segmentation tasks, demonstrate the superior performance of our method. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

20 pages, 7247 KiB

Open AccessArticle

Machine Learning Model for Leak Detection Using Water Pipeline Vibration Sensor

by Suan Lee and Byeonghak Kim

Sensors 2023, 23(21), 8935; https://doi.org/10.3390/s23218935 - 2 Nov 2023

Cited by 15 | Viewed by 9691

Abstract

Water leakage from aging water and wastewater pipes is a persistent problem, necessitating the improvement of existing leak detection and response methods. In this study, we conducted an analysis of essential features based on data collected from leak detection sensors installed at water [...] Read more.

Water leakage from aging water and wastewater pipes is a persistent problem, necessitating the improvement of existing leak detection and response methods. In this study, we conducted an analysis of essential features based on data collected from leak detection sensors installed at water meter boxes and water outlets of pipelines. The water pipeline data collected through the vibration sensor were preprocessed by converting it into a tabular form by frequency band and applied to various machine learning models. The characteristics of each model were analyzed, and XGBoost was selected as the most suitable leak detection model with a high accuracy of 99.79%. These systems can effectively reduce leak detection and response time, minimize water waste, and minimize economic losses. Additionally, this technology can be applied to various fields that utilize water pipes, making it widely applicable. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

18 pages, 4426 KiB

Open AccessArticle

Fault Diagnosis of Vibration Sensors Based on Triage Loss Function-Improved XGBoost

by Chao Fan, Cheng Li, Yanfeng Peng, Yiping Shen, Guanghui Cao and Sai Li

Electronics 2023, 12(21), 4442; https://doi.org/10.3390/electronics12214442 - 29 Oct 2023

Cited by 2 | Viewed by 1704

Abstract

Vibration sensors are prone to bias, drift, and other failures. To avoid misjudgments in state monitoring systems and potential safety accidents caused by vibration sensor failures, it is significant to diagnose the faults of vibration sensors. Existing methods for vibration sensor fault diagnosis [...] Read more.

Vibration sensors are prone to bias, drift, and other failures. To avoid misjudgments in state monitoring systems and potential safety accidents caused by vibration sensor failures, it is significant to diagnose the faults of vibration sensors. Existing methods for vibration sensor fault diagnosis are primarily based on Deep Learning, but Extreme Gradient Boosting stands out due to its excellent interpretability, and compared to other ensemble learning algorithms, it boasts superior accuracy and efficiency. Therefore, a vibration sensor fault diagnosis method based on Extreme Gradient Boosting is proposed to diagnose seven common types of faults in vibration sensors. To prevent the model from being overwhelmed by simple negative cases during training, a new loss function named Triage Loss is designed to improve the classifier’s performance. The vibration sensor fault diagnosis has confirmed the efficacy and practicality of the suggested approach. The experimental results indicate that the training of the model done using Triage Loss outperforms the training model done using the default loss function, with a maximum improvement of 5.4% accuracy, 5.45% in the F1-score, and 9.87% in the mean Average Precision under different fault rates. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

29 pages, 4849 KiB

Open AccessReview

Energy Prediction for Energy-Harvesting Wireless Sensor: A Systematic Mapping Study

by Zhenbo Yuan, Yongqi Ge, Jiayuan Wei, Shuhua Yuan, Rui Liu and Xian Mo

Electronics 2023, 12(20), 4304; https://doi.org/10.3390/electronics12204304 - 18 Oct 2023

Cited by 2 | Viewed by 2843

Abstract

Energy prediction plays a significant role in energy-harvesting wireless sensors (EHWS), as it helps wireless sensors regulate their duty cycles, achieve energy neutrality, and extend their lifespan. To explore and analyze advanced technologies and methods regarding energy prediction for EHWS, this study identifies [...] Read more.

Energy prediction plays a significant role in energy-harvesting wireless sensors (EHWS), as it helps wireless sensors regulate their duty cycles, achieve energy neutrality, and extend their lifespan. To explore and analyze advanced technologies and methods regarding energy prediction for EHWS, this study identifies future research directions and addresses the challenges faced based on the current research status, assisting with future literature research. This scholarly inquiry delineates future research prospects and addresses prevailing challenges within the context of the extant research landscape, thereby facilitating prospective scholarly endeavors. This study employed the systematic mapping study (SMS) approach to screen and further investigate the relevant literature. After searching and screening for papers from the ACM, IEEE Xplore, and Web of Science (WOS) databases from January 2007 to December 2022, 98 papers met the requirements of this study. Subsequently, the SMS was conducted for five research questions. The results showed that the solution proposal type category had the largest proportion among all research types, accounting for 58% of the total number, indicating that the research focusing on this field is placed on improving the existing methods or proposing new ones. Additionally, based on the SMS analysis, this study provides a systematic review of the technical utilization and improvement approaches, as well as the strengths and limitations of the selected prediction methods. Furthermore, by considering the current research landscape, this paper identifies the existing challenges and suggests future research directions, thereby offering valuable insights to researchers for making informed decisions regarding their chosen paths. The significance of this study lies in its contribution to driving advancements in the field of energy-harvesting wireless sensor networks. The importance of this study is underscored by its contribution to advancing the domain of energy-harvesting wireless sensor networks, thereby serving as a touchstone for forthcoming researchers in this specialized field. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

17 pages, 18522 KiB

Open AccessArticle

Deep Neural Network-Based Visual Feedback System for Nasopharyngeal Swab Sampling

by Suhun Jung, Yonghwan Moon, Jeongryul Kim and Keri Kim

Sensors 2023, 23(20), 8443; https://doi.org/10.3390/s23208443 - 13 Oct 2023

Cited by 2 | Viewed by 1531

Abstract

During the 2019 coronavirus disease pandemic, robotic-based systems for swab sampling were developed to reduce burdens on healthcare workers and their risk of infection. Teleoperated sampling systems are especially appreciated as they fundamentally prevent contact with suspected COVID-19 patients. However, the limited field [...] Read more.

During the 2019 coronavirus disease pandemic, robotic-based systems for swab sampling were developed to reduce burdens on healthcare workers and their risk of infection. Teleoperated sampling systems are especially appreciated as they fundamentally prevent contact with suspected COVID-19 patients. However, the limited field of view of the installed cameras prevents the operator from recognizing the position and deformation of the swab inserted into the nasal cavity, which highly decreases the operating performance. To overcome this limitation, this study proposes a visual feedback system that monitors and reconstructs the shape of an NP swab using augmented reality (AR). The sampling device contained three load cells and measured the interaction force applied to the swab, while the shape information was captured using a motion-tracking program. These datasets were used to train a one-dimensional convolution neural network (1DCNN) model, which estimated the coordinates of three feature points of the swab in 2D X–Y plane. Based on these points, the virtual shape of the swab, reflecting the curvature of the actual one, was reconstructed and overlaid on the visual display. The accuracy of the 1DCNN model was evaluated on a 2D plane under ten different bending conditions. The results demonstrate that the x-values of the predicted points show errors of under 0.590 mm from

P_{0}

, while those of

P_{1}

and

P_{2}

show a biased error of about −1.5 mm with constant standard deviations. For the y-values, the error of all feature points under positive bending is uniformly estimated with under 1 mm of difference, when the error under negative bending increases depending on the amount of deformation. Finally, experiments using a collaborative robot validate its ability to visualize the actual swab’s position and deformation on the camera image of 2D and 3D phantoms. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

16 pages, 3258 KiB

Open AccessArticle

Fault Classification for Cooling System of Hydraulic Machinery Using AI

by Haseeb Ahmed Khan, Uzair Bhatti, Khurram Kamal, Mohammed Alkahtani, Mustufa Haider Abidi and Senthan Mathavan

Sensors 2023, 23(16), 7152; https://doi.org/10.3390/s23167152 - 13 Aug 2023

Cited by 5 | Viewed by 3530

Abstract

Hydraulic systems are used in all kinds of industries. Mills, manufacturing, robotics, and Ports require the use of Hydraulic Equipment. Many industries prefer to use hydraulic systems due to their numerous advantages over electrical and mechanical systems. Hence, the growth in demand for [...] Read more.

Hydraulic systems are used in all kinds of industries. Mills, manufacturing, robotics, and Ports require the use of Hydraulic Equipment. Many industries prefer to use hydraulic systems due to their numerous advantages over electrical and mechanical systems. Hence, the growth in demand for hydraulic systems has been increasing over time. Due to its vast variety of applications, the faults in hydraulic systems can cause a breakdown. Using Artificial-Intelligence (AI)-based approaches, faults can be classified and predicted to avoid downtime and ensure sustainable operations. This research work proposes a novel approach for the classification of the cooling behavior of a hydraulic test rig. Three fault conditions for the cooling system of the hydraulic test rig were used. The spectrograms were generated using the time series data for three fault conditions. The CNN variant, the Residual Network, was used for the classification of the fault conditions. Various features were extracted from the data including the F-score, precision, accuracy, and recall using a Confusion Matrix. The data contained 43,680 attributes and 2205 instances. After testing, validating, and training, the model accuracy of the ResNet-18 architecture was found to be close to 95%. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

15 pages, 4287 KiB

Open AccessArticle

Intelligent Measuring of the Volume Fraction Considering Temperature Changes and Independent Pressure Variations for a Two-Phase Homogeneous Fluid Using an 8-Electrode Sensor and an ANN

by Ramy Mohammed Aiesh Qaisi, Farhad Fouladinia, Abdulilah Mohammad Mayet, John William Grimaldo Guerrero, Hassen Loukil, M. Ramkumar Raja, Mohammed Abdul Muqeet and Ehsan Eftekhari-Zadeh

Sensors 2023, 23(15), 6959; https://doi.org/10.3390/s23156959 - 5 Aug 2023

Cited by 8 | Viewed by 1773

Abstract

Two-phase fluids are widely utilized in some industries, such as petrochemical, oil, water, and so on. Each phase, liquid and gas, needs to be measured. The measuring of the void fraction is vital in many industries because there are many two-phase fluids with [...] Read more.

Two-phase fluids are widely utilized in some industries, such as petrochemical, oil, water, and so on. Each phase, liquid and gas, needs to be measured. The measuring of the void fraction is vital in many industries because there are many two-phase fluids with a wide variety of liquids. A number of methods exist for measuring the void fraction, and the most popular is capacitance-based sensors. Aside from being easy to use, the capacitance-based sensor does not need any separation or interruption to measure the void fraction. In addition, in the contemporary era, thanks to Artificial Neural Networks (ANN), measurement methods have become much more accurate. The same can be said for capacitance-based sensors. In this paper, a new metering system utilizing an 8-electrode sensor and a Multilayer Perceptron network (MLP) is presented to predict an air and water volume fractions in a homogeneous fluid. Some characteristics, such as temperature, pressure, etc., can have an impact on the results obtained from the aforementioned sensor. Thus, considering temperature changes, the proposed network predicts the void fraction independent of pressure variations. All simulations were performed using the COMSOL Multiphysics software for temperature changes from 275 to 370 degrees Kelvin. In addition, a range of 1 to 500 Bars, was considered for the pressure. The proposed network has inputs obtained from the mentioned software, along with the temperature. The only output belongs to the predicted void fraction, which has a low MAE equal to 0.38. Thus, based on the obtained result, it can be said that the proposed network precisely measures the amount of the void fraction. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

15 pages, 5423 KiB

Open AccessArticle

MFF-YOLO: An Accurate Model for Detecting Tunnel Defects Based on Multi-Scale Feature Fusion

by Anfu Zhu, Bin Wang, Jiaxiao Xie and Congxiao Ma

Sensors 2023, 23(14), 6490; https://doi.org/10.3390/s23146490 - 18 Jul 2023

Cited by 3 | Viewed by 2292

Abstract

Tunnel linings require routine inspection as they have a big impact on a tunnel’s safety and longevity. In this study, the convolutional neural network was utilized to develop the MFF-YOLO model. To improve feature learning efficiency, a multi-scale feature fusion network was constructed [...] Read more.

Tunnel linings require routine inspection as they have a big impact on a tunnel’s safety and longevity. In this study, the convolutional neural network was utilized to develop the MFF-YOLO model. To improve feature learning efficiency, a multi-scale feature fusion network was constructed within the neck network. Additionally, a reweighted screening method was devised at the prediction stage to address the problem of duplicate detection frames. Moreover, the loss function was adjusted to maximize the effectiveness of model training and improve its overall performance. The results show that the model has a recall and accuracy that are 7.1% and 6.0% greater than those of the YOLOv5 model, reaching 89.5% and 89.4%, respectively, as well as the ability to reliably identify targets that the previous model error detection and miss detection. The MFF-YOLO model improves tunnel lining detection performance generally. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

31 pages, 18282 KiB

Open AccessReview

Deep Learning for Optical Sensor Applications: A Review

by Nagi H. Al-Ashwal, Khaled A. M. Al Soufy, Mohga E. Hamza and Mohamed A. Swillam

Sensors 2023, 23(14), 6486; https://doi.org/10.3390/s23146486 - 18 Jul 2023

Cited by 21 | Viewed by 5661

Abstract

Over the past decade, deep learning (DL) has been applied in a large number of optical sensors applications. DL algorithms can improve the accuracy and reduce the noise level in optical sensors. Optical sensors are considered as a promising technology for modern intelligent [...] Read more.

Over the past decade, deep learning (DL) has been applied in a large number of optical sensors applications. DL algorithms can improve the accuracy and reduce the noise level in optical sensors. Optical sensors are considered as a promising technology for modern intelligent sensing platforms. These sensors are widely used in process monitoring, quality prediction, pollution, defence, security, and many other applications. However, they suffer major challenges such as the large generated datasets and low processing speeds for these data, including the high cost of these sensors. These challenges can be mitigated by integrating DL systems with optical sensor technologies. This paper presents recent studies integrating DL algorithms with optical sensor applications. This paper also highlights several directions for DL algorithms that promise a considerable impact on use for optical sensor applications. Moreover, this study provides new directions for the future development of related research. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

20 pages, 12807 KiB

Open AccessArticle

Application of Machine Learning Algorithms to Classify Peruvian Pisco Varieties Using an Electronic Nose

by Celso De-La-Cruz, Jorge Trevejo-Pinedo, Fabiola Bravo, Karina Visurraga, Joseph Peña-Echevarría, Angela Pinedo, Freddy Rojas and María R. Sun-Kou

Sensors 2023, 23(13), 5864; https://doi.org/10.3390/s23135864 - 24 Jun 2023

Cited by 3 | Viewed by 2451

Abstract

Pisco is an alcoholic beverage obtained from grape juice distillation. Considered the flagship drink of Peru, it is produced following strict and specific quality standards. In this work, sensing results for volatile compounds in pisco, obtained with an electronic nose, were analyzed through [...] Read more.

Pisco is an alcoholic beverage obtained from grape juice distillation. Considered the flagship drink of Peru, it is produced following strict and specific quality standards. In this work, sensing results for volatile compounds in pisco, obtained with an electronic nose, were analyzed through the application of machine learning algorithms for the differentiation of pisco varieties. This differentiation aids in verifying beverage quality, considering the parameters established in its Designation of Origin”. For signal processing, neural networks, multiclass support vector machines and random forest machine learning algorithms were implemented in MATLAB. In addition, data augmentation was performed using a proposed procedure based on interpolation–extrapolation. All algorithms trained with augmented data showed an increase in performance and more reliable predictions compared to those trained with raw data. From the comparison of these results, it was found that the best performance was achieved with neural networks. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

14 pages, 6522 KiB

Open AccessArticle

A LiDAR–Inertial SLAM Method Based on Virtual Inertial Navigation System

by Yunpiao Cai, Weixing Qian, Jiayi Dong, Jiaqi Zhao, Kerui Wang and Tianxiao Shen

Electronics 2023, 12(12), 2639; https://doi.org/10.3390/electronics12122639 - 12 Jun 2023

Cited by 8 | Viewed by 2142

Abstract

In scenarios with insufficient structural features, LiDAR-based SLAM may suffer from degeneracy, resulting in impaired robot localization and mapping and potentially leading to subsequent deviant navigation tasks. Therefore, it is crucial to develop advanced algorithms and techniques to mitigate the degeneracy issue and [...] Read more.

In scenarios with insufficient structural features, LiDAR-based SLAM may suffer from degeneracy, resulting in impaired robot localization and mapping and potentially leading to subsequent deviant navigation tasks. Therefore, it is crucial to develop advanced algorithms and techniques to mitigate the degeneracy issue and ensure the robustness and accuracy of LiDAR-based SLAM. This paper presents a LiDAR–inertial simultaneous localization and mapping (SLAM) method based on a virtual inertial navigation system (VINS) to address the issue of degeneracy. We classified different gaits and match each gait to its corresponding torso inertial measurement unit (IMU) sensor to construct virtual foot inertial navigation components. By combining an inertial navigation system (INS) with zero-velocity updates (ZUPTs), we formed the VINS to achieve real-time estimation and correction. Finally, the corrected pose estimation was input to the IMU odometry calculation procedure to further refine the localization and mapping results. To evaluate the effectiveness of our proposed VINS method in degenerate environments, we conducted experiments in three typical scenarios. The results demonstrate the high suitability and accuracy of the proposed method in degenerate scenes and show an improvement in the point clouds mapping effect. The algorithm’s versatility is emphasized by its wide applicability on GPU platforms, including quadruped robots and human wearable devices. This broader potential range of applications extends to other related fields such as autonomous driving. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

17 pages, 706 KiB

Open AccessArticle

A Robust Pedestrian Re-Identification and Out-Of-Distribution Detection Framework

by Abdelhamid Bouzid, Daniel Sierra-Sosa and Adel Elmaghraby

Drones 2023, 7(6), 352; https://doi.org/10.3390/drones7060352 - 27 May 2023

Cited by 1 | Viewed by 2069

Abstract

Pedestrian re-identification is an important field due to its applications in security and safety. Most current solutions for this problem use CNN-based feature extraction and assume that only the identities that are in the training data can be recognized. On the one hand, [...] Read more.

Pedestrian re-identification is an important field due to its applications in security and safety. Most current solutions for this problem use CNN-based feature extraction and assume that only the identities that are in the training data can be recognized. On the one hand, the pedestrians in the training data are called In-Distribution (ID). On the other hand, in real-world scenarios, new pedestrians and objects can appear in the scene, and the model should detect them as Out-Of-Distribution (OOD). In our previous study, we proposed a pedestrian re-identification based on von Mises–Fisher (vMF) distribution. Each identity is embedded in the unit sphere as a compact vMF distribution far from other identity distributions. Recently, a framework called Virtual Outlier Synthetic (VOS) was proposed, which detects OOD based on synthesizing virtual outliers in the embedding space in an online manner. Their approach assumes that the samples from the same object map to a compact space, which aligns with the vMF-based approach. Therefore, in this paper, we revisited the vMF approach and merged it with VOS to detect OOD data points. Experiment results showed that our framework was able to detect new pedestrians that do not exist in the training data in the inference phase. Furthermore, this framework improved the re-identification performance and holds a significant potential in real-world scenarios. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

18 pages, 1893 KiB

Open AccessArticle

MFPCDR: A Meta-Learning-Based Model for Federated Personalized Cross-Domain Recommendation

by Yicheng Di and Yuan Liu

Appl. Sci. 2023, 13(7), 4407; https://doi.org/10.3390/app13074407 - 30 Mar 2023

Cited by 15 | Viewed by 3000

Abstract

Cross-domain recommendation systems frequently require the use of rich source domain information to improve recommendations in the target domain, thereby resolving the data sparsity and cold-start problems, whereas the majority of existing approaches frequently require the centralized storage of user data, which poses [...] Read more.

Cross-domain recommendation systems frequently require the use of rich source domain information to improve recommendations in the target domain, thereby resolving the data sparsity and cold-start problems, whereas the majority of existing approaches frequently require the centralized storage of user data, which poses a substantial risk of privacy breaches. Compared to traditional recommendation systems with centralized data, federated recommendation systems with multiple clients trained collaboratively have significant privacy benefits in terms of user data. While users’ interests are often personalized, meta-learning can be used to learn users’ personalized preferences, and personalized preferences can help models make recommendations in cold-start scenarios. We use meta-learning to learn the personalized preferences of cold-start users. Therefore, we offer a unique meta-learning-based federated personalized cross-domain recommendation model that discovers the personalized preferences for cold-start users via a server-side meta-recommendation module. To avoid compromising user privacy, an attention mechanism is used on each client to find transferable features that contribute to knowledge transfer while obtaining embeddings of users and items; each client then uploads the weights to the server. The server accumulates weights and delivers them to clients for update. Compared to traditional recommendation system models, our model can effectively protect user privacy while solving the user cold-start problem, as we use an attention mechanism in the local embedding module to mine the source domain for transferable features that contribute to knowledge transfer. Extensive trials on real-world datasets have demonstrated that our technique effectively guarantees speed while protecting user privacy. Full article

(This article belongs to the Topic Artificial Intelligence in Sensors, 2nd Volume)

► Show Figures

Figure 1

Topic Menu

Topic Editors

Artificial Intelligence in Sensors, 2nd Volume

Topic Information

Keywords

Participating Journals

Published Papers (15 papers)

Further Information

Guidelines

MDPI Initiatives

Follow MDPI