View Video Research Articles

How Does Short Video Viewing Influence Young Children’s Everyday Language Practices? A Case Study of China

How Does Short Video Viewing Influence Young Children’s Everyday Language Practices? A Case Study of China

Deep Learning Phenotyping of Tricuspid Regurgitation for Automated High Throughput Assessment of Transthoracic Echocardiography.

Diagnosis of tricuspid regurgitation (TR) requires careful expert evaluation. This study developed an automated deep learning pipeline for assessing TR from transthoracic echocardiography. An automated deep learning workflow was developed using 47,312 studies (2,079,898 videos) from Cedars-Sinai Medical Center (CSMC) between 2011 and 2021. The pipeline was tested on a temporally distinct test set of 2,462 studies (108,138 videos) obtained in 2022 at CSMC and a geographically distinct cohort of 5,549 studies (278,377 videos) from Stanford Healthcare (SHC). In the CSMC test dataset, the view classifier demonstrated an AUC of 1.000 (0.999 - 1.000) and identified at least one A4C video with colour Doppler across the tricuspid valve in 2,410 of 2,462 studies with a sensitivity of 0.975 (0.968-0.982) and a specificity of 1.000 (1.00-1.000). In the CSMC test cohort, moderate-or-severe TR was detected with an AUC of 0.928 (0.913 - 0.943) and severe TR was detected with an AUC of 0.956 (0.940 - 0.969). In the SHC cohort, the view classifier correctly identified at least one TR colour Doppler video in 5,268 of the 5,549 studies, resulting in an AUC of 0.999 (0.998 - 0.999), a sensitivity of 0.949 (0.944 - 0.955) and specificity of 0.999 (0.999 - 0.999). The AI model detected moderate-or-severe TR with an AUC of 0.951 (0.938 - 0.962) and severe TR with an AUC of 0.980 (0.966 - 0.988). We developed an automated pipeline to identify clinically significant TR with excellent performance. This approach carries potential for automated TR detection and stratification for surveillance and screening. Can an automated deep learning model assess tricuspid regurgitation severity from echocardiography? We developed and validated an automated tricuspid regurgitation detection algorithm pipeline across two healthcare systems with high volume echocardiography labs. The algorithm correctly identifies apical-4-chamber view videos with colour Doppler across the tricuspid valve and grades clinically significant TR with strong agreement to expert clinical readers. A deep learning pipeline could automate TR screening, facilitating reproducible accurate assessment of TR severity, allowing rapid triage or re-review and expand access in low-resource or primary care settings.

Understanding urban perception with visual data: A systematic review

Visual characteristics of the built environment affect how people perceive and experience cities. For a long time, many studies have examined visual perception in cities. Such efforts have accelerated in recent years due to advancements in technologies and the proliferation of relevant data (e.g., street view imagery, geo-tagged photos, videos, virtual reality, and aerial imagery). There has not been a comprehensive systematic review paper on this topic to reveal an overarching set of research trends, limitations, and future research opportunities. Such omission is plausibly due to the difficulty in reviewing a large number of relevant papers on this popular topic. In this study, we utilized machine learning techniques (i.e., natural language processing and large language models) to semi-automate the review process and reviewed 393 relevant papers. Through the review, we found that these papers can be categorized into the physical aspects of cities: greenery and water, street design, building design, landscape, public space, and the city as a whole. We also revealed that many studies conducted quantitative analyses with a recent trend of increasingly utilizing big data and advanced technologies, such as combinations of street view imagery and deep learning models. Limitations and research gaps were also identified as follows: (1) a limited scope in terms of study areas, sample size, and attributes; (2) low quality of subjective and visual data; and (3) the need for more controlled and sophisticated methods to infer more closely examined impacts of visual features on human perceptions. We suggest that future studies utilize and contribute to open data and take advantage of existing data and technologies to examine the causality of visual features on human perception. The approach developed to accelerate this review proved to be accurate, efficient, and insightful. Considering its novelty, we also describe it to enable replications in the future.

Assessing inclusion and representativeness on digital platforms for health education: Evidence from YouTube

BackgroundStudies confirm that significant biases exist in online recommendation platforms, exacerbating pre-existing disparities and leading to less-than-optimal outcomes for underrepresented demographics. We study issues of bias in inclusion and representativeness in the context of healthcare information disseminated via videos on the YouTube social media platform, a widely used online channel for multi-media rich information. With one in three US adults using the Internet to learn about a health concern, it is critical to assess inclusivity and representativeness regarding how health information is disseminated by digital platforms such as YouTube. MethodsLeveraging methods from fair machine learning (ML), natural language processing and voice and facial recognition methods, we examine inclusivity and representativeness of video content presenters using a large corpus of videos and their metadata on a chronic condition (diabetes) extracted from the YouTube platform. Regression models are used to determine whether presenter demographics impact video popularity, measured by the video’s average daily view count. A video that generates a higher view count is considered to be more popular. ResultsThe voice and facial recognition methods predicted the gender and race of the presenter with reasonable success. Gender is predicted through voice recognition (accuracy = 78%, AUC = 76%), while the gender and race predictions use facial recognition (accuracy = 93%, AUC = 92% and accuracy = 82%, AUC = 80%, respectively). The gender of the presenter is more significant for video views only when the face of the presenter is not visible while videos with male presenters with no face visibility have a positive relationship with view counts. Furthermore, videos with white and male presenters have a positive influence on view counts while videos with female and non − white group have high view counts. ConclusionPresenters’ demographics do have an influence on average daily view count of videos viewed on social media platforms as shown by advanced voice and facial recognition algorithms used for assessing inclusion and representativeness of the video content. Future research can explore short videos and those at the channel level because popularity of the channel name and the number of videos associated with that channel do have an influence on view counts.

Comparing the efficacy of online vs. offline methods in modulating emotional responses through mental stress tasks

Emotion regulation and stress represent two complex, intertwined processes. Yet, there is a paucity of studies examining the effect of stress on emotional regulation among Vietnamese students, specifically the influence of online and offline experimental conditions. This study investigates these dynamics by subjective assessment of the emotional responses of the subjects participating in this study. The research was structured in three primary phases: preliminary, online (conducted via the Internet), and offline (conducted in person) experiments. The preliminary phase (N=16) aimed to validate the appropriateness of the International Affective Picture System (IAPS) picture library and emotional stimulation videos from previous studies. Participants in the online (N=30) and offline (N=45) experiments were divided into control and stress groups, distinguished by the difficulty and urgency of the mental arithmetic task (MAT). Participants were then asked to view photos and videos and to rate their subjective feelings using the Self-assessment Manikin (SAM). The online experimental session revealed that subjects under acute stress exhibited heightened sensitivity, as indicated by increased valence scores in the positive and neutral groups, and decreased scores for negative imagery (Positive: MD=0.292, p=0.0113; Neutral: MD=0.245, p=0.001; Negative: MD=-0.435, p<0.0001). An increase in arousal scores in the stress group was also observed for positive (p<0.0001) and neutral sets (p<0.0001). However, these findings were not replicated in the offline experiment session. In conclusion, the IAPS pictures have proven suitable for emotional studies among Vietnamese students.

Choice of architecture of video surveillance systems of State Border Guard of Ukraine based on cloud and fog technologies

Modern video surveillance systems have evolved from closed-circuit television systems to the level of complex systems that operate as part of complex integrated systems and solve the tasks of not only recording events in the sectors of view of video surveillance cameras but also analysing the received video information. The increase in the amount of information circulating in modern systems requires the creation of new concepts. In recent years, the attention of scientists has been drawn to distributed information processing technologies. The concept of cloud computing is developing rapidly, and the basic ideas that were laid down in its construction can no longer solve the problems faced by the technology. This has led to the emergence of so-called post-cloud architectures that extend and complement the capabilities of cloud computing. These architectures include Mist, Edge, Fog, etc. The purpose of the article is to study video surveillance systems based on the concept of cloud and fog computing and criteria for assessing their effectiveness. The research was conducted using the methods of analysis and generalisation, modelling, and experimentation. This approach made it possible to conduct a comparative study of video surveillance systems built on two architectures. The results obtained indicate that fog computing technology has the advantage of reducing latency, minimising the need for repeated requests to the cloud by performing calculations at the cloud edge. In a model based on the cloud computing concept, services that use cloud resources lead to an increase in network load. At the same time, fog computing technology allows you to relieve the network load by performing part of the computation by fog nodes. The results of the experimental study show the advantages of fog computing for networks that are sensitive to delays. However, if we consider a video surveillance system with the main task of recording events in the sectors of view of CCTV cameras and the ability to view video information in real time, it is obvious that fog nodes will not provide long-term storage of video information, and delays will not be critical. In the case of a video surveillance system with video analytics functions, fog nodes will be able to perform part of the video analytics algorithms, thus unloading the cloud. Therefore, an urgent task is to study the effectiveness of building video surveillance systems with video analytics functions based on fog architecture.

Assessment of the Quality of Sinus Elevation with Lateral Window Approach Procedure Videos on YouTube: A Content-quality Analysis.

Although the lateral window approach allows for greater graft material delivery and bone formation, it is more challenging and invasive, prompting keen interest among dentists to master this method. YouTube is increasingly used for medical training; however, concerns regarding the quality of instructional videos exist. This study proposes new criteria for evaluating YouTube videos on maxillary sinus elevation with the aim of establishing standards for assessing instructional content in the field. We sourced 100 maxillary sinus elevation videos from YouTube and, following exclusion criteria, analyzed 65 remaining videos. The video characteristics, content quality, and newly developed criteria were evaluated. Statistical analyses, employing ordinal logistic regression, identified the factors influencing the quality of instructional videos and evaluated the significance of our new criteria. Although video interaction and view rate exhibited positive relations to content quality, they were not significant ( P =0.818 and 0.826, respectively). Notably, videos of fair and poor quality showed a significant negative relation ( P <0.001). Audio commentary, written commentary, and descriptions of preoperative data displayed positive but statistically insignificant relationships ( P =0.088, 0.228, and 0.612, respectively). The comparison of video evaluation results based on the developed criteria with content quality scores revealed significant negative relationships for good, fair, and poor videos ( P <0.001, Exp(B)=-4.306, -7.853, -10.722, respectively). Among the various video characteristics, only image quality showed a significant relationship with content quality. Importantly, our newly developed criteria demonstrated a significant relationship with video content quality, providing valuable insights for assessing instructional videos on maxillary sinus elevation and laying the foundation for robust standards.

YouTube Video Characteristics as Indicators for Adherence to Accurate Water Birth Guidelines [ID 2683502

INTRODUCTION: The accessibility of YouTube videos has made it a popular platform for seeking and sharing pregnancy advice, especially experiences related to “natural” delivery methods like water births. Therefore, it is important to understand which video characteristics amassed more views as well as explore the relationship between characteristics and accuracy of information. We aim to assess what components of water birth YouTube videos are predictive of more accurate content with evidence based on American College of Obstetricians and Gynecologists (ACOG) guidelines. METHODS: A YouTube search on “water birth” was conducted using the filter “most viewed.” The first 100 videos were selected using a strict inclusion criteria, and video characteristics were recorded. Video views ranged from 457,296 to 78,035,874. Reviewers scored videos in accordance with published guidelines using the “ACOG Water Birth Rubric.” Chi-squared goodness-of-fit and independence tests were performed to assess significance of differences between groups. RESULTS: Out of 96 videos, 13.1% from North America were scored as accurate, versus 15% from other continents (P=.5754). Only 14.4% of videos from personal channels were accurate, despite comprising 71.8% of all videos in the top 100. All videos from doulas and midwives scored neutral, and there were no videos from health care professionals with advanced degrees. Commercial videos had the highest amount of accurate videos at 16.6%. Videos with a supportive figure in the thumbnail had significantly less videos graded as inaccurate (3.5%) compared to those without (18%; P=.0003). CONCLUSION: Patients use YouTube videos as a resource for pregnancy information, particularly for alternative birthing practices. North America has some of the most popular videos and personal channels that often deviate from ACOG guidelines. Our findings highlight an urgent need for health care professionals to guide patients towards evidence-based resources, especially for alternative birth practices, and contribute accurate and informative content on such platforms.

Adaptation of Postural Sway in a Standing Position during Tilted Video Viewing Using Virtual Reality: A Comparison between Younger and Older Adults.

This study aimed to investigate the effects of wearing virtual reality (VR) with a head-mounted display (HMD) on body sway in younger and older adults. A standing posture with eyes open without an HMD constituted the control condition. Wearing an HMD and viewing a 30°-tilt image and a 60°-tilt image in a resting standing position were the experimental conditions. Measurements were made using a force plate. All conditions were performed three times each and included the X-axis trajectory length (mm), Y-axis trajectory length (mm), total trajectory length (mm), trajectory length per unit time (mm/s), outer peripheral area (mm2), and rectangular area (mm2). The results showed a significant interaction between generation and condition in Y-axis trajectory length (mm) and total trajectory length (mm), with an increased body center-of-gravity sway during the viewing of tilted VR images in older adults than in younger adults in both sexes. The results of this study show that body sway can be induced by visual stimulation alone with VR without movement, suggesting the possibility of providing safe and simple balance training to older adults.

Dacryocystorhinostomy videos on YouTube as a source of patient education

BackgroundTo determine the quality and reliability of DCR YouTube videos as patient education resources and identify any associated factors predictive of video quality.MethodsA YouTube search was conducted using the terms “Dacryocystorhinostomy, DCR, surgery” on 12th of January 2022, with the first 50 relevant videos selected for inclusion. For each video, the following was collected: video hyperlink, title, total views, months since the video was posted, video length, total likes/dislikes, authorship (i.e. surgeon, patient experience or media companies) and number of comments. The videos were graded independently by a resident, a registrar and an oculoplastic surgeon using three validated scoring systems: the Journal of the American Medical Association (JAMA), DISCERN, and Health on the Net (HON).ResultsThe average number of video views was 22,992, with the mean length being 488.12 s and an average of 18 comments per video. The consensus JAMA, DISCERN and HON scores were 2.1 ± 0.6, 29.1 ± 8.8 and 2.7 ± 1.0, respectively. This indicated that the included videos were of a low quality, however, only DISCERN scores had good interobserver similarity. Videos posted by surgeons were superior to non-surgeons when considering mean JAMA and HON scores. No other factors were associated with the quality of educational content.ConclusionThe quality and reliability of DCR related content for patient education is relatively low. Based on this study’s findings, patients should be encouraged to view videos created by surgeons or specialists in preference to other sources on YouTube.

Video double-lumen tube for one lung ventilation: implementation and experience in 343 cases of routine clinical use during the first 20 months of the SARS-CoV-2 pandemic

BackgroundDouble-lumen tubes (DLTs) are the preferred device for lung isolation. Conventional DLTs (cDLT) need a bronchoscopic position control. Visualisation of correct DLT positioning could be facilitated by the use of a video double-lumen tube (vDLT). During the SARS-CoV-2-pandemic, avoiding aerosol-generation was suggesting using this device. In a large retrospective series, we report both general and pandemic related experiences with the device.MethodsAll anesthesia records from patients aged 18 years or older undergoing surgery from April 1st, 2020 to December 31st, 2021 in the department of thoracic surgery requiring intraoperative lung isolation were analyzed retrospectively.ResultsDuring the investigation period 343 left-sided vDLTs (77.4%) and 100 left-sided cDLTs (22.6%) were used for one lung ventilation. In the vDLT group bronchoscopy could be reduced by 85.4% related to the cDLT group. Additional bronchoscopy to reach or maintain correct position was needed in 11% of the cases. Other bronchoscopy indications occured in 3.6% of the cases. With cDLT, in 1% bronchoscopy for other indications than conforming position was observed.ConclusionsThe Ambu® VivaSight™ vDLT is an efficient, easy-to-use and safe airway device for the generation of one lung ventilation in patients undergoing thoracic surgery. The vDLT implementation was achieved easily with full interchangeability to the left-sided cDLT. Using the vDLT can reduce the need for aerosol-generating bronchoscopic interventions by 85.4%. Continuous video view to the carina enabling position monitoring of the DLT without need for bronchoscopy might be beneficial for both employee’s and patient’s safety.

Temporal segmentation of motion propagation in response to an external impulse

In high-density crowds, local motion can propagate, amplify, and lead to macroscopic phenomena, including ”density waves”. These density waves only occur when individuals interact, and impulses are transferred to neighbours. How this impulse is passed on by the human body and which effects this has on individuals is still not fully understood. To further investigate this, experiments focusing on the propagation of a push were conducted within the EU-funded project CrowdDNA. In the experiments the crowd is greatly simplified by five people lining up in a row. The rearmost person in the row was pushed forward in a controlled manner with a punching bag. The intensity of the push, the initial distance between participants and the initial arm posture were varied. Collected data included side view and top view video recordings, head trajectories, 3D motion using motion capturing (MoCap) suits as well as pressure measured at the punching bag. With a hybrid tracking algorithm, the MoCap data are combined with the head trajectories to allow an analysis of the motion of each limb in relation to other persons.The observed motion of the body in response to the push can be divided into three phases. These are (i) receiving an impulse, (ii) receiving and passing on an impulse, and (iii) passing on an impulse. Using the 3D MoCap data, we can identify the start and end times of each phase. To determine when a push is passed on, the forward motion of the person in front has to be considered. The projection of the center of mass relative to the initial position of the feet is a measure of the extent to which a person is displaced from the rest position. Specifying the timing of these phases is particularly important to distinguish between different types of physical interactions. Our results contribute to the development and validation of a pedestrian model for identifying risks due to motion propagation in dense crowds.

Analysis of Instagram Posts with the Hashtag Preventivedentistry

This study aimed to analyze the posts published with the #preventivedentistry on Instagram. In this cross-sectional retrospective study, the #preventivedentistry was searched on Instagram. Data were collected between 01/10/2023 and 01/20/2023. A total of 940 posts were analyzed for type, purpose, and source. The number of likes of the posts and the number of video views were determined. Of the posts examined, 869 were images, and 71 were videos. Of the posts, 19.7% (n = 185) with the #preventivedentistry were about preventive dentistry. The content of 80.3% (n = 755) included advertisements, celebrations, information about other dental practices, patient satisfaction, and personal sharing. Among specialist dentists, pediatric dentists shared the most posts about preventive dentistry. Regardless of content, most of the posts had 0-25 likes, and only a small percentage had above 300 likes. People searching with #preventivedentistry on Instagram cannot find sufficient information about preventive dentistry. Dental professionals should use Instagram more actively to increase awareness of preventive dentistry.

Accuracy to identify young adults with chronic ankle instability using a virtual reality - Balance error scoring system: A cross-sectional study

IntroductionThe Balance Error Scoring System (BESS) assesses the ability to control postural stability by performing 3 different stances on two-type surfaces during closed eyes. Virtual reality technology combined with the BESS test (VR-BESS) may be used to disrupt visual inputs instead of closing the eyes, which may improve the sensitivity of diagnosing patients with chronic ankle instability (CAI). ObjectiveThis study aimed to evaluate the accuracy to identify individuals with CAI of the VR-BESS test comparing with the original BESS test. MethodsThe BESS and VR-BESS tests were administered to 68 young adults (34 participants with CAI and 34 without CAI). Frontal and lateral video views were used to measure the participant's performance errors. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve was computed to determine the diagnostic test's overall accuracy. ResultsThe total score of the BESS test and the VR-BESS test were statistically significant in comparison to the AUC of no discrimination at 0.5, with AUC values of 0.63 and 0.64, respectively. The cut-off scores for the BESS and VR-BESS tests were 12 and 15, respectively. There was no significant difference between the ROC curves of the BESS and the VR-BESS test for identifying individuals with CAI. ConclusionThe BESS and VR-BESS tests may be utilized interchangeably to identify individuals with CAI.

Patient experiences and perceptions of kidney stone surgery: what lessons can be learned from TikTok?

The aim of this study was to perform an evaluation of patient experiences and perceptions regarding kidney stone surgery on the social media platform TikTok. An increasing number of the public use social media (SoMe) as a platform to share their views regarding their experiences related to surgical treatment. Using the hashtag #kidneystonesurgery, the 100 most recent video posts as of 01.01.2024 on TikTok were included. As well as demographic data such as gender and location, thematic content was also collected. To achieve this, a previously published framework was used and adapted for application in the setting of kidney stone surgery. This was piloted on 20 sample videos to assess its feasibility before revision and establishment of the final framework. This included the following key areas: Pain, Complications, Anxiety, Recovery, Return to work, Finances, Treatment delays, Diet and Prevention and stent complaints. The majority of posts (95%) were from North America, 80% by females and the mean number of video views was 92,826 (range: 261-2,000,000). 76% of the videos discussed ureteroscopy (URS). 49% were filmed at the hospital, which was named in 9% of the videos. Top three topics discussed were: Recovery (65%), pain (62%) and stents (55%). This was followed by anxiety (39%) and complications (24%). 12% of these videos uploaded by lay people included basic medical information that was wholly incorrect. More than half of the posts (51%) were negative in tone. Treatment delays (5%) and a lack of sufficient preoperative information (4%) were also raised, that appeared to contribute to the negative reports. However, the main cause for negative tone owed to the 80% of the patients (n = 44) who discussed stents that focused their video on the pain suffered from the post operative stent. There is a high level of usership and engagement on TikTok on the subject of kidney stone surgery. The proportion of negative videos is high and much of this is related to the bothersome stent symptoms and complications. This could easily lead to misperceptions among potential patients about the true burden of such adverse events.

Ambient identity construction via massive anonymous danmu comments

This article contributes to the scholarship on digital identity work by examining the linguistic mechanisms of ambient identity construction in fully anonymous online environments. It investigates how video viewers construct textual personae by leaving massive anonymous comments in danmu, a viewing-and-commenting system that synchronously posts comments onto a video screen as it plays. Drawing on the sociological concept of homophily and the linguistics-informed Appraisal framework, this study systematically tracks the patterns in the attitudinal orientations among massive anonymous comments left over a high-profile Chinese video featuring a teacher's home visit. The article argues that the technological affordances of danmu lead to the inherent collectiveness of anonymous digital identity construction. It reports two attitudinal meaning-making mechanisms through which massive anonymous comments converge into a homogeneous mass and describes the viewers' collective ambient identities revealed in their comments. This project brings clarity to the dynamics of ambient digital identity construction by deploying computational tools to linguistic analysis and has practical implications for marketing research, social media monitoring, and community building.

Rage against the machine? Framing societal threat and efficacy in YouTube videos about artificial intelligence.

Artificial intelligence (AI) has become a part of the mainstream public discourse beyond expert communities about its risks, benefits, and need for regulation. In particular, since 2014, the news media have intensified their coverage of this emerging technology and its potential impact on most domains of society. Although many studies have analyzed traditional media coverage of AI, analyses of social media, especially video-sharing platforms, are rare. In addition, research from a risk communication perspective remains scarce, despite the widely recognized potential threats to society from many AI applications. This study aims to detect recurring patterns of societal threat/efficacy in YouTube videos, analyze their main sources, and compare detected frames in terms of reach and response. Using a theoretical framework combining framing and risk communication, the study analyzed the societal threat/efficacy attributed to AI in easily accessible YouTube videos published in a year when public attention to AI temporarily peaked (2018). Four dominant AI frames were identified: the balanced frame, the high-efficacy frame, the high-threat frame, and the no-threat frame. The balanced and no-threat frames were the most prevalent, with predominantly positive and neutral AI narratives that neither adequately address the risks nor the necessary societal response from a normative risk communication perspective. The results revealed the specific risks and benefits of AI that are most frequently addressed. Video views and user engagement with AI videos were analyzed. Recommendations for effective AI risk communication and implications for risk governance were derived from the results.

When Video Improves Learning in Higher Education

The use of video in education has become ubiquitous as technological developments have markedly improved the ability and facility to create, deliver, and view videos. The concomitant pedagogical transformation has created a sense of urgency regarding how video may be used to advance learning. Initial reviews have suggested only limited potential for the use of video in higher education. More recently, a systematic review of studies on the effect of video use on learning in higher education, published in the journal Review of Educational Research, found, overall, effects to be positive. In the present paper, we critique this study. We reveal significant gaps in the study methodology and write-up and use a cognitive processing lens to critically assess and re-analyse study data. We found the results of this study to be only applicable to learning requiring lower-level cognitive processing and conclude, consistent with prior research, that claims of a universal benefit are not yet warranted.

Gait recognition of camouflaged people based on UAV infrared imaging

As a unique behavioral feature that can be obtained at distance, gait plays an essential role in special target surveillance, security prevention, and more. If gait recognition can be deployed on unmanned aerial vehicles (UAVs), it will open novel research perspectives for special operations. However, existing researches are dominated by ground view video recognition in constrained scenes rather than aerial view. In addition, gait recognition in complex environments remains challenging due to the similarity of targets and backgrounds, especially in scenarios where individuals intentionally hide identities through camouflage. To address these problems, we utilize the UAV carrying the infrared camera for gait acquisition and propose a novel gait recognition framework of camouflaged people. Firstly, an infrared gait dataset captured by the UAV flying at different altitudes and perspectives is constructed in this paper, named IR-150. Then, a novel hierarchical framework that integrates the advantages of CNN and Transformer is designed for gait recognition, named CTGait. Specially, depth-wise convolution and residual structures are applied to the local aggregation unit and feed-forward networks to introduce the locality mechanism. By means of diagonal masking and learnable temperature parameters, the attention mechanism is redesigned to enhance the recognition accuracy further. Various experiments prove that the CTGait is superior to the state-of-the-art methods, which achieves an average Rank-1 accuracy of 97.74 % and 82.08 % on the datasets CASIA-C and IR-150, respectively. Tests on the public datasets CASIA-B and GREW demonstrate that the CTGait exhibits significant robustness even in complex scenarios.

Salient object detection in egocentric videos

AbstractIn the realm of video salient object detection (VSOD), the majority of research has traditionally been centered on third‐person perspective videos. However, this focus overlooks the unique requirements of certain first‐person tasks, such as autonomous driving or robot vision. To bridge this gap, a novel dataset and a camera‐based VSOD model, CaMSD, specifically designed for egocentric videos, is introduced. First, the SalEgo dataset, comprising 17,400 fully annotated frames for video salient object detection, is presented. Second, a computational model that incorporates a camera movement module is proposed, designed to emulate the patterns observed when humans view videos. Additionally, to achieve precise segmentation of a single salient object during switches between salient objects, as opposed to simultaneously segmenting two objects, a saliency enhancement module based on the Squeeze and Excitation Block is incorporated. Experimental results show that the approach outperforms other state‐of‐the‐art methods in egocentric video salient object detection tasks. Dataset and codes can be found at https://github.com/hzhang1999/SalEgo.