Open Dataset Research Articles

A biomechanics and energetics dataset of neurotypical adults walking with and without kinematic constraints

Numerous studies have explored the biomechanics and energetics of human walking, offering valuable insights into how we walk. However, prior studies focused on changing external factors (e.g., walking speed) and examined group averages and trends rather than individual adaptations in the presence of internal constraints (e.g., injury-related muscle weakness). To address this gap, this paper presents an open dataset of human walking biomechanics and energetics collected from 21 neurotypical young adults. To investigate the effects of internal constraints (reduced joint range of motion), the participants are both the control group (free walking) and the intervention group (constrained walking - left knee fully extended using a passive orthosis). Each subject walked on a dual-belt treadmill at three speeds (0.4, 0.8, and 1.1 m/s) and five step frequencies ( − 10% to 20% of their preferred frequency) for a total of 30 test conditions. The dataset includes raw and segmented data featuring ground reaction forces, joint motion, muscle activity, and metabolic data. Additionally, a sample code is provided for basic data manipulation and visualisation.

Head motion-corrected eye gaze tracking with the da Vinci surgical system.

To facilitate the integration of point of gaze (POG) as an input modality for robot-assisted surgery, we introduce a robust head movement compensation gaze tracking system for the da Vinci Surgical System. Previous surgical eye gaze trackers require multiple recalibrations and suffer from accuracy loss when users move from the calibrated position. We investigate whether eye corner detection can reduce gaze estimation error in a robotic surgery context. A polynomial regressor is first used to estimate POG after an 8-point calibration, and then, using another regressor, the POG error from head movement is estimated from the shift in 2D eye corner location. Eye corners are computed by first detecting regions of interest using the You Only Look Once (YOLO) object detector trained on 1600 annotated eye images (open dataset included). Contours are then extracted from the bounding boxes and a derivative-based curvature detector refines the eye corner. Through a user study (n = 24), our corner-contingent head compensation algorithm showed an error reduction in degrees of visual angle of 1.20 (p = 0.037) for the left eye and 1.26 (p = 0.079) for the right compared to the previous gold-standard POG error correction method. In addition, the eye corner pipeline showed a root-mean-squared error of 3.57 (SD = 1.92) pixels in detecting eye corners over 201 annotated frames. We introduce an effective method of using eye corners to correct for eye gaze estimation, enabling the practical acquisition of POG in robotic surgery.

Ethical Competencies in Machine Learning from a Communicational Perspective in Educational Process

The term "artificial intelligence" (AI) refers to computer programs equipped with numerous competencies, such as making calculations, grouping and categorizing data, or communicating with the user in ethnic languages. On the other hand, artificial intelligence systems do not have certain properties, and among them, apart from the lack of "creative abilities", is indifference to the moral aspect of actions when searching and compiling data or cataloging phenomena. This study aims to discuss selected reasons for this state of affairs in the context of machine learning (ML) methodology, including the issues and applications of artificial intelligence from the perspective of scientific communication that occurs in the educational process. Given this goal, a research problem was formulated in the form of a question: How are ethical competencies developed in the machine learning process in the context of communication occurring in the educational process? In order to answer this research question, the text analysis method and the synthesis method were used. As a result of the research, it was determined that conducting machine learning with human participation, as well as using artificial intelligence systems previously learned with human participation, may enable the transmission of moral content in the normative sense to a cybernetic machine. Since human participation allows supervised learning of cybernetic machines, this type of learning, used as the sole method or in combination with another method, offers the opportunity to provide applications with the desired information about socio-cultural rules. Fully independent training of cybernetic machines does not ensure they collect information on ethical aspects desirable in communication during the educational process because open data sets on which machine learning takes place may contain harmful content, amplifying negative social phenomena.

Deep learning-based restoration of nonlinear motion blurred images for plant classification using multi-spectral images

There have been various plant image-based studies for segmentation, deblurring, super-resolution reconstruction, and classification. However, nonlinear motion blur in thermal images was not considered in the existing studies on plant classification. Nonlinear motion blur occurs in images due to camera or plant movements, and it causes the degradation of plant classification accuracy. Moreover, nonlinear motion blur in images gets worse when both camera and plant movements occur simultaneously. In this case, it becomes difficult to recognize plants, and the performance of plant image classification becomes very low. Therefore, to reduce the nonlinear motion blur, a thermal and visible light plant images-based deblurring network (TVPD-Net) is proposed in this study. In addition, a thermal and visible light plant images-based classification network (TVPC-Net) is also proposed to improve the plant classification performance on deblurred images. Experimental results revealed that the proposed TVPD-Net achieved 21.21 and 22.53 of the peak signal-to-noise ratio (PSNR), and 0.726 and 0.737 of the structural similarity index measure (SSIM) on both visible light and thermal plant image datasets which were self-collected, respectively. Moreover, the proposed TVPC-Net with deblurred images by TVPD-Net achieved 92.52 % (top-1 accuracy) and 87.73 % (harmonic mean of precision and recall (F1-score)). In addition, the experimental results on an open dataset named Hyperspectral Flower Dataset (HFD100) revealed that the proposed plant classification method achieved 90.94 % of top-1 accuracy and 86.21 % of F1-score. The accuracies of the proposed methods are greater than those of the state-of-the-art methods.

CaRDS - the statewide California Residential water Demand and Supply open dataset

As water scarcity becomes the new norm in the Western United States, states such as California have increased their efforts to improve water resilience. Achieving water security under climate change, population growth, and urbanization requires an integrated multi-sectoral approach, where adaptation strategies combine supply and demand management interventions. Yet, most studies consider supply-side and demand-side management strategies separately. Water conservation efforts are mainly driven by policy requirements and publicly available data to assess the effectiveness of demand- and supply-side management policies is often hard to find and unstructured. Here we present CaRDS - the statewide California Residential water Demand and Supply open dataset. CaRDS encompasses nine years (2013-2021) of monthly water supply and demand time series for 404 water suppliers in California, USA, compiled from different open-access data sources. Access to detailed temporal and spatial water supply operations and demands at the state-level can be useful to researchers and practitioners to realize applications such as evaluating the effectiveness of water conservation policies and discovering regional differences in water resilience measures.

Analysis of the alpha activity envelope in electroencephalography in relation to the ratio of excitatory to inhibitory neural activity.

Alpha waves, one of the major components of resting and awake cortical activity in human electroencephalography (EEG), are known to show waxing and waning, but this phenomenon has rarely been analyzed. In the present study, we analyzed this phenomenon from the viewpoint of excitation and inhibition. The alpha wave envelope was subjected to secondary differentiation. This gave the positive (acceleration positive, Ap) and negative (acceleration negative, An) values of acceleration and their ratio (Ap-An ratio) at each sampling point of the envelope signals for 60 seconds. This analysis was performed on 36 participants with Alzheimer's disease (AD), 23 with frontotemporal dementia (FTD) and 29 age-matched healthy participants (NC) whose data were provided as open datasets. The mean values of the Ap-An ratio for 60 seconds at each EEG electrode were compared between the NC and AD/FTD groups. The AD (1.41 ±0.01 (SD)) and FTD (1.40 ±0.02) groups showed a larger Ap-An ratio than the NC group (1.38 ±0.02, p<0.05). A significant correlation between the envelope amplitude of alpha activity and the Ap-An ratio was observed at most electrodes in the NC group (Pearson's correlation coefficient, r = -0.92 ±0.15, mean for all electrodes), whereas the correlation was disrupted in AD (-0.09 ±0.21, p<0.05) and disrupted in the frontal region in the FTD group. The present method analyzed the envelope of alpha waves from a new perspective, that of excitation and inhibition, and it could detect properties of the EEG, Ap-An ratio, that have not been revealed by existing methods. The present study proposed a new method to analyze the alpha activity envelope in electroencephalography, which could be related to excitatory and inhibitory neural activity.

Deepfake Video Detection Using Facial Feature Points and Ch-Transformer

With the development of Metaverse technology, the avatar in Metaverse has faced serious security and privacy concerns. Analyzing facial features to distinguish between genuine and manipulated facial videos holds significant research importance for ensuring the authenticity of characters in the virtual world and for mitigating discrimination, as well as preventing malicious use of facial data. To address this issue, the Facial Feature Points and Ch-Transformer (FFP-ChT) deepfake video detection model is designed based on the clues of different facial feature points distribution in real and fake videos and different displacement distances of real and fake facial feature points between frames. The face video input is first detected by the BlazeFace model, and the face detection results are fed into the FaceMesh model to extract 468 facial feature points. Then the Lucas-Kanade (LK) optical flow method is used to track the points of the face, the face calibration algorithm is introduced to re-calibrate the facial feature points, and the jitter displacement is calculated by tracking the facial feature points between frames. Finally, the Class-head (Ch) is designed in the transformer, and the facial feature points and facial feature point displacement are jointly classified through the Ch-Transformer model. In this way, the designed Ch-Transformer classifier is able to accurately and effectively identify deepfake videos. Experiments on open datasets clearly demonstrate the effectiveness and generalization capabilities of our approach.

Vision Transformer for Flood Detection Using Satellite Images from Sentinel-1 and Sentinel-2

Floods are devastating phenomena that occur almost all around the world and are responsible for significant losses, in terms of both human lives and economic damages. When floods occur, one of the challenges that emergency response agencies face is the identification of the flooded area so that access points and safe routes can be determined quickly. This study presents a flood detection methodology that combines transfer learning with vision transformers and satellite images from open datasets. Transformers are powerful models that have been successfully applied in Natural Language Processing (NLP). A variation of this model is the vision transformer (ViT), which can be applied to image classification tasks. The methodology is applied and evaluated for two types of satellite images: Synthetic Aperture Radar (SAR) images from Sentinel-1 and Multispectral Instrument (MSI) images from Sentinel-2. By using a pre-trained vision transformer and transfer learning, the model is fine-tuned on these two datasets to train the models to determine whether the images contain floods. It is found that the proposed methodology achieves an accuracy of 84.84% on the Sentinel-1 dataset and 83.14% on the Sentinel-2 dataset, revealing its insensitivity to the image type and applicability to a wide range of available visual data for flood detection. Moreover, this study shows that the proposed approach outperforms state-of-the-art CNN models by up to 15% on the SAR images and 9% on the MSI images. Overall, it is shown that the combination of transfer learning, vision transformers, and satellite images is a promising tool for flood risk management experts and emergency response agencies.

Discrete Morse theory segmentation on high-resolution 3D lithic artifacts

Abstract Motivated by the question of understanding the roots of tool making by anatomically modern humans and coexisting Neanderthals in the Paleolithic, a number of shape classification methods have been tested on photographs and drawings of stone tools. Since drawings contain interpretation and photographs fool both human and computational methods by color and shadows on the surface, we propose an approach using 3D datasets as best means for analyzing shape, and rely on first open access repositories on lithic tools. The goal is to not only analyze shape on an artifact level, but allow a more detailed analysis of stone tools on a scar and ridge level. A Morse-Smale complex (MS complex) extracted from the triangular mesh of a 3D model is a reduced skeleton consisting of linked lines on the mesh. Discrete Morse theory makes it possible to obtain such a MS complex from a scalar function. Thus, we begin with Multi-Scale Integral Invariant filtering on the meshes of lithic artifacts, which provides curvature measures for ridges, which are convex, and scars, which are concave. The resulting values on the vertices can be used as our discrete Morse function and the skeleton we get is build up from lines that will coincide with the ridges and, implicitly, contains the scars as enclosed regions of those lines on the mesh. As this requires a few parameters, we provide a graphical user interface (GUI) to allow altering the predefined parameters to quickly find a good result. In addition, a stone tool may have areas that do not belong to the scar/ridge class. These can be masked and we use conforming MS complexes to ensure that the skeleton keeps these areas whole. Finally, results are shown on real and open access datasets. The source code and manually annotated ground truth for the evaluation are provided as Open Access with a Creative Commons license.

Efficient flexible voxel-based two-stage network for 3D object detection in autonomous driving

3D object detection from the LiDAR point cloud plays an important role in autonomous driving. It is difficult to balance inference speed and detection accuracy when performing 3D point cloud object detection due to the large size of point cloud data and its unstructured storage, which makes it difficult to represent its features. To address the challenge, we propose a two-stage point cloud object detector, AFV-RCNN. In stage-1, the attention flexible voxel feature encoding layer is introduced, which utilizes flexible voxels to enhance feature encoding speed and focuses on foreground points to be detected through voxel attention. In stage-2, the multi-level and grid-based multi-scale RoI (Region of Interest) feature fusion module is designed. It directly extracts complete 3D structures from 3D region proposals and focuses on both local and global features through multi-scale partitioning. In the training stage, GHM-C Loss is applied to address the challenges associated with imbalanced target categories and the imbalance between difficult and easy samples in the classification task. We evaluate the model on the public KITTI Dataset and Waymo Open Dataset. The mAP in KITTI for 3D detection is 73.41% and inference on a single GPU reaches 30.0 fps. Compared with other state-of-the-art methods, AFV-RCNN achieves both the inference speed of a single-stage detector and the detection accuracy of a two-stage detector. It ensures higher detection accuracy while efficiently processing the point cloud.

Evaluating the sensitivity and robustness of occupancy models for building energy simulation during design

Occupancy modeling plays a crucial role in building energy simulation in building design, yet studies investigating the influence of model types and their input assumptions remain sparse. This study evaluates the sensitivity of parameters and model robustness to uncertainties in parameter ranges and data availability during the design phase. We assessed three occupancy models—deterministic, probabilistic, and agent-based—using an open dataset with occupancy data across four space types. Our findings reveal that the predictive performance of deterministic and probabilistic models is influenced heavily by a few of the most sensitive parameters. In contrast, more parameters affected the prediction performance of the agent-based model. This suggests a need to prioritize sensitive parameter estimation to enhance prediction accuracy in deterministic and probabilistic models. Specifically, the occupancy density and start/end time, followed by schedule fractions, emerged as highly sensitive due to their significant impact on average occupancy profiles. We also found that for hourly energy predictions, narrowing the range of the top five sensitive parameters decreases the coefficient of variation of the root mean square error (CV(RMSE)) by 3% and reduces its standard deviation by half. These results underscore the importance of careful parameter estimation to improve the reliability of occupancy models in building energy simulations.

Cognitive Workload Estimation in Conditionally Automated Vehicles Using Transformer Networks Based on Physiological Signals

Though driving automation promises to improve driving safety, drivers are still required to be ready to retake control in conditionally automated vehicles, which are defined by the Society of Automotive Engineers (SAE) as SAE L3 vehicles. Thus, drivers’ states can still affect driving safety in SAE L3 vehicles. Previous research found that a high cognitive load may impair drivers’ takeover performance. Thus, it is still necessary to estimate drivers’ cognitive load in SAE L3 vehicles. However, existing driver cognitive load estimation algorithms mostly focus on vehicles with a lower level of driving automation (e.g., SAE L0), which may not be relevant when estimating driver states in SAE L3 vehicles, given that drivers’ responsibilities are different, and several commonly used measures (e.g., driving performance) are unavailable when drivers are not continuously controlling the vehicle. Further, previous driver cognitive load estimation algorithms rarely considered the temporal information in the input features. Thus, we proposed a deep-learning algorithm to estimate driver cognitive load in SAE L3 vehicles, which integrated multiple physiological features (i.e., electrocardiogram, electrodermal activity, respiration) and considered the temporal correlation of the data using a transformer-encoder-based network. The performance of our algorithm was compared with baseline models on an open data set. Results showed that our algorithm outperformed baseline models and achieved an accuracy of 94.4% using within-subject data partition (proportionally splitting data from the same subject into the training and testing data sets) and an accuracy of 89% using across-subjects data partition (dividing the training and testing data sets based on individual subjects).

Fault vibration model driven fault-aware domain generalization framework for bearing fault diagnosis

Deep learning methods can learn effective representations from the data, simplifying the fault diagnosis process and improving accuracy. However, the lack of data presents a significant challenge to effectively implementing fault diagnosis driven by deep learning. Simulation models can generate rich source domain data, which can be combined with domain adaptation methods to realize fault diagnosis tasks under sample scarcity. Nevertheless, there are some problems with existing simulation-driven fault diagnosis frameworks: (1) The choice of parameters for simulation models is usually empirical, which can lead to domain-specific components differences between the simulation data and the actual data. (2) It is difficult for the simulation models to generate the background noise of the real equipment, which will cause fault information to be covered, severely limiting the robustness of the fault diagnosis network. (3) Studies based on domain adaptation methods require a certain amount of target domain data for training, which is difficult to meet in real industrial production. To address these issues, this paper proposes a fault-aware domain generalization bearing fault diagnosis framework driven by fault vibration model, an analytical simulation model. Firstly, a spectral background estimation algorithm based on adaptive clutter separation (ACS) and maximum spectral envelope (MSE), termed ACS-MSE, is designed to estimate the potential resonance frequencies in the real system, guiding the parameters setting of the fault diagnosis vibration model and ensuring consistency between the simulated data and the actual data in the label space. Secondly, the normal actual signal is combined with the fault simulation signal to create a mixed signal that considers real background noise, and a fault-aware autoencoder is developed to extract the fault-related impact components covered by the background noise. Finally, a multi resonance frequencies domain generalization fault diagnosis network based on contrastive learning is designed to extract generalized fault features crossing different resonance frequencies and realize diagnosis tasks without real fault data. Experimental results on an open dataset and a custom dataset showed that the proposed method achieves higher accuracy compared to some advanced domain generalization and even domain adaptation methods. Ablation studies verify the importance and irreplaceability of the individual components of the proposed method.

Feature selection based on contradictory state sequence for multi-scale interval valued decision table

In the context of selecting features for multi-scale interval valued decision table (MIVDT), conventional research approaches encounter difficulties including elevated data complexity, diminished computational efficiency, and limited model generalization capacity. To overcome these difficulties, feature selection methods based on contradictory state sequence (CSS) and fuzzy contradictory state sequence (FCSS) are proposed in this paper. Initially, MIVDT is established. According to the characteristics of interval value, a more thorough and impartial metric for assessing the inclusion degree is suggested, along with similarity relation based on the metric. Subsequently, the introduction of the first contradictory object allows for the delineation of contradictory state and fuzzy contradictory state, which serve to characterize the consistency of MIVDT. These two concepts are proposed to enhance the accuracy and efficiency of capturing key information in intricate data. Additionally, rapid feature selection algorithms are suggested, which rely on CSS and FCSS. In contrast to conventional feature selection approaches, the algorithms introduced in this study demonstrate superior computational efficiency and enhanced generalization capabilities when confronted with intricate datasets. In twelve open source datasets, the upper limit for the quantity of objects is 110341, and the average accuracy values of experiments under two parameter combinations are 99.54% and 97.08% respectively. The results of experiments demonstrate that the algorithms introduced are capable of efficiently identifying a novel feature subset from intricate datasets within a shorter timeframe.

Structure preserved fast dimensionality reduction

Many graph-based unsupervised dimensionality reduction techniques have raised concerns about their high accuracy. However, there is an urgent need to address the enormous time consumption problem in large-scale data scenarios. Therefore, we present a novel approach named Structure Preserved Fast Dimensionality Reduction (SPFDR). Firstly, the parameter-insensitive, sparse, and scalable bipartite graph is constructed to build the similarity matrix. Then, employing alternating iterative optimization, the linear dimensionality reduction matrix and the optimal similarity matrix preserved cluster structure are learned. The computational complexity of the conventional graph-based dimension reduction method costs O(n2d+d3), yet the proposed approach is O(ndm+nm2), wherein n, m, and d are the number of instances, anchors, and features, respectively. Eventually, experiments conducted with multiple open datasets will provide convincing evidence for how effective and efficient the proposed method is.

Elevated SLC1A5 associated with poor prognosis and therapeutic resistance to transarterial chemoembolization in hepatocellular carcinoma

BackgroundHepatocellular carcinoma (HCC) is a common malignant tumor, and glutamine is vital for tumor cells. The role of glutamine transporter SLC1A5 in tumor progression and transarterial chemoembolization (TACE) efficacy is under study. This research seeks to determine the impact of SLC1A5 expression on the prognosis and TACE efficacy of HCC and elucidate its mechanisms.MethodsSLC1A5 expression in HCC, correlation with patient outcomes, and response to TACE were studied in an open access liver cancer dataset and confirmed in our cohort. Additionally, the correlation between SLC1A5 expression and hypoxia, angiogenesis and immune infiltration was analyzed and verified by immunohistochemistry, immunofluorescence and transcriptome sequencing. Liver cancer cell lines with SLC1A5 expression knockdown or overexpression were constructed, and cell proliferation, colony formation, apoptosis, migration and drug sensitivity as well as in vivo xenograft tumor were measured. A gene set enrichment analysis was conducted to determine the signaling pathway influenced by SLC1A5, and a western blot analysis was performed to detect protein expression alterations.ResultsSLC1A5 expression was higher in HCC tissue and associated with poor survival and TACE resistance. Hypoxia could stimulate the upregulation of glutamine transport, angiogenesis and SLC1A5 expression. The SLC1A5 expression was positively correlated with hypoxia and angiogenesis-related genes, immune checkpoint pathways, macrophage, Tregs, and other immunosuppressive cells infiltration. Knockdown of SLC1A5 decreased proliferation, colony formation, and migration, but increased apoptosis and increased sensitivity to chemotherapy drugs. Downregulation of SLC1A5 resulted in a decrease in Vimentin and N-cadherin expression, yet an increase in E-cadherin expression. Upregulation of SLC1A5 increased Vimentin and N-cadherin expression, while decreasing E-cadherin. Overexpression of β-catenin in SLC1A5-knockdown HCC cell lines could augment Vimentin and N-cadherin expression, suppress E-cadherin expression, and increase the migration and drug resistance.ConclusionsElevated SLC1A5 was linked to TACE resistance and survival shortening in HCC patients. SLC1A5 was positively correlated with hypoxia, angiogenesis, and immunosuppression. SLC1A5 may mediate HCC cell migration and drug resistance via Epithelial-mesenchymal transition (EMT) pathway.

A global dataset for assessing nitrogen-related plant traits using drone imagery in major field crop species

Enhancing rapid phenotyping for key plant traits, such as biomass and nitrogen content, is critical for effectively monitoring crop growth and maximizing yield. Studies have explored the relationship between vegetation indices (VIs) and plant traits using drone imagery. However, there is a gap in the literature regarding data availability, accessible datasets. Based on this context, we conducted a systematic review to retrieve relevant data worldwide on the state of the art in drone-based plant trait assessment. The final dataset consists of 41 peer-reviewed papers with 11,189 observations for 11 major crop species distributed across 13 countries. It focuses on the association of plant traits with VIs at different growth/phenological stages. This dataset provides foundational knowledge on the key VIs to focus for phenotyping key plant traits. In addition, future updates to this dataset may include new open datasets. Our goal is to continually update this dataset, encourage collaboration and data inclusion, and thereby facilitate a more rapid advance of phenotyping for critical plant traits to increase yield gains over time.

Naïve Bayes Approach to Understanding Students’ Pychological Well-Being

High levels of stress and depression among Indonesian students are significant issues that demand special attention. According to data from the National Crime Information Center (Pusiknas) of the Indonesian National Police (Polri), there were 971 cases of suicide related to the mental health of students from January to October 18, 2023. This study utilizes an open dataset from Kaggle, which includes information on students' mental health conditions, and applies the Naive Bayes algorithm to determine accuracy, precision, and recall values. The Naive Bayes approach is employed to classify the mental health status of students and identify those who require special assistance. The results indicate that the Naive Bayes algorithm is effective in identifying students needing special help, with a high accuracy rate. Testing with RapidMiner yielded an accuracy of 90.00%, a recall rate of 89.66% for the 'No' class and 100.00% for the 'Yes' class. Precision for the 'No' label was 100.00%, while for the 'Yes' label, it was 25.00%. This approach can aid campuses, families, and professionals in identifying students who need special attention.

Automated Mapping of Vulnerability Advisories onto their Fix Commits in Open Source Repositories

The lack of comprehensive sources of accurate vulnerability data represents a critical obstacle to studying and understanding software vulnerabilities (and their corrections). In this article, we present an approach that combines heuristics stemming from practical experience and machine-learning (ML)—specifically, natural language processing (NLP)—to address this problem. Our method consists of three phases. First, we construct an advisory record object containing key information about a vulnerability that is extracted from an advisory, such as those found in the National Vulnerability Database (NVD). These advisories are expressed in natural language. Second, using heuristics, a subset of candidate fix commits is obtained from the source code repository of the affected project, by filtering out commits that can be identified as unrelated to the vulnerability at hand. Finally, for each of the remaining candidate commits, our method builds a numerical feature vector reflecting the characteristics of the commit that are relevant to predicting its match with the advisory at hand. Based on the values of these feature vectors, our method produces a ranked list of candidate fixing commits. The score attributed by the ML model to each feature is kept visible to the users, allowing them to easily interpret the predictions. We implemented our approach and we evaluated it on an open data set, built by manual curation, that comprises 2,391 known fix commits corresponding to 1,248 public vulnerability advisories. When considering the top-10 commits in the ranked results, our implementation could successfully identify at least one fix commit for up to 84.03% of the vulnerabilities (with a fix commit on the first position for 65.06% of the vulnerabilities). Our evaluation shows that our method can reduce considerably the manual effort needed to search open-source software (OSS) repositories for the commits that fix known vulnerabilities.

A deep learning-based super-resolution method for building height estimation at 2.5 m spatial resolution in the Northern Hemisphere

Building height is an important indicator for assessing the level of urban development along the vertical dimension. Existing large-scale building height estimation studies focus on coarse spatial resolution (e.g., 10, 500, and 1000 m), which cannot reveal height variations across buildings in urban areas. High-resolution images (e.g., < 5 m resolution) can support building-scale height estimation, but they usually have small spatial coverage and are not openly accessible. In this context, we proposed a deep learning-based super-resolution method to generate building height maps at a spatial resolution of 2.5 m using Sentinel-1/2 images. The proposed method consisted of two parts: 1) a super-resolution module (SR) for learning high-resolution features; and 2) a height stratification estimation module (HS) for guiding the network to learn different height levels to mitigate the imbalanced distribution of height values. We created an open building height dataset with 45,000 samples covering multiple urban areas in the Northern Hemisphere, including China, the conterminous United States (CONUS), and Europe. Experimental results showed that for height estimation at the pixel level, the proposed method obtained a root mean square error of 10.318 m in China, 5.654 m in CONUS, and 4.113 m in Europe, respectively. Predicted results provided rich spatial details, due to the inclusion of the super-resolution module, which was heavily missed by existing large-scale studies. Moreover, we calculated the mean and standard deviation of building height in 301 urban centers, each having at least a population of 500,000, and found that the buildings in China were the highest (11.353 m ± 9.543 m), followed by CONUS (8.487 m ± 6.202 m) and Europe (8.136 m ± 5.020 m). Ablation studies indicated that the joint use of Sentinel-1/2 images and the proposed modules (SR and HS) can effectively improve the performance of building height estimation. The building dataset we generated provides great potential in high-resolution database updating, urban planning, and natural disaster assessment, and indeed, a new perspective of how we can utilize cutting-edge satellite imaging technology in urban observation, measurement, monitoring, and management. The dataset and code of this study will be available at: https://github.com/lauraset/Super-resolution-building-height-estimation.