Video Annotation Research Articles

Enhancing Learning Engagement in the Flipped Classroom using a Video-Annotation Tool

In the realm of academia, there has been a burgeoning interest in examining student engagement within the context of the flipped classroom approach, particularly at the higher education level. Nevertheless, research pertaining to a pivotal facet of this pedagogical methodology—namely, student engagement during the individualized, at-home preparatory phase—remains limited. Within this void, the primary objective of this study is to delve into the prospective benefits offered by a video annotation tool in augmenting students’ level of engagement during this autonomous study phase. Throughout the course of this investigation, a bespoke questionnaire was crafted, grounded in a multifaceted framework for evaluating student engagement, with a specific emphasis on dissecting four distinct dimensions: behavioral, emotional, agentive, and cognitive. This questionnaire is geared towards gauging student engagement within the domain of the Flipped Classroom model that relies on video-based instructional content—a facet heretofore poorly explored within the scholarly literature. Both quantitative and qualitative data were collected by administering the aforementioned questionnaire to a cohort of 68 undergraduate engineering students at Politecnico di Milano (Italy), all within the authentic context of a case study. The outcomes stemming from this empirical inquiry showcase a noteworthy enhancement in student engagement, with particular prominence accorded to the realms of emotional and agentive engagement. Moreover, this study establishes that the interactivity and proactive involvement necessitated by the video annotation tool do not obstruct students’ behavioral and cognitive engagement levels. In summation, this research endeavors to illuminate the potentiality of bridging a pivotal juncture in the Flipped Classroom paradigm, specifically the phase characterized by independent at-home study. This bridge is facilitated by the utilization of a video annotation tool designed to heighten student engagement. This transformative approach effectively transmutes the traditionally passive at-home study phase of the Flipped Classroom into an active experience, thereby enhancing the overall efficacy of this pedagogical approach.

A socially interactive communication skills training for competent health educators

Abstract Background Every contact with expectant and new parents can be a “teachable moment” for promoting the health of mother-baby-family; yet, healthcare providers may not effectively perform their educational role. Previous studies have described various setting-specific challenges. Procedural knowledge, resulting in lack of confidence and avoidance, is a universal theme, challenging current communications skills training. Given the evolving role of midwives under the new Cyprus National Healthcare System, there is need for innovative practices. Objectives “Baby Buddy Communicators” designed, implemented and piloted tested a socially interactive digital environment on the Edubreak platform, where practicing (N = 8) and in-training (N = 7) Midwives completed an 8-week iterative hybrid communication skills course along with service-users. Results Four situational-based scenarios were co-created with ‘holders of the experience’ tapping on key priorities (elective caesarean, breastfeeding challenges, baby in NICU, postnatal depression). Participants produced video artefacts of role-play consultations and engaged in peer-to-peer and service-user feedback via the in-built video annotations and blog writing tools. Deductive content analysis of the rich material using the 18-item Health Professionals Communications Skills (HPCS) scale identified recurrent patterns tapping on both informational aspects (e.g. incomplete information gathering, overload without assessing comprehension) as well as social skills (e.g. agenda setting, early reassurance, assertiveness). Conclusions This cross-sectoral transfer of social video from sports training to healthcare is a “proof of concept”. Inverted (i.e. making current practices visible) and experiential learning (e.g. feedforward) with cross-sectoral (service-user involvement) and cross-module dimensions (i.e. communication skills and evidence-based practice) provided the basis for evaluation and a source of shared learning and reflection. Key messages • Several areas of improvement in communication skills of maternal healthcare providers were identified while peer-to-peer and service-user feedback was a source of reflection. • The use of digital technology and social video allowed immersive role-playing, observation and constructive feedback loops providing a new paradigm in health communication skills training.

Can Video Annotation Software With Minimal Standardization Be Used as a Back-Up for a Marker-Free Motion Tracking System when Monitoring Barbell Thruster Velocity During a Repeated Circuit?

Mangine, GT, Hines, A, Grazer, J, Esmat, TA, and McLester, J. Can video annotation software with minimal standardization be used as a back-up for a marker-free motion tracking system when monitoring barbell thruster velocity during a repeated circuit? J Strength Cond Res XX(X): 000-000, 2024-This study examined the agreement between a marker-free motion tracking system (PERCH) and a free video annotation software (KINOVEA) for measuring barbell thruster velocity using minimal standardization. For a larger placebo-controlled supplement intervention, 10 men (28.7 ± 8.0 years, 175 ± 8 cm, 91.0 ± 12.0 kg) and 10 women (31.3 ± 5.2 years, 167 ± 4 cm, 70 ± 13 kg) completed a circuit of rowing, 6 barbell thrusters (men: 43.1 kg, women: 29.5 kg), and 3 box jumps for "as many repetitions as possible" (AMRAP) in either 5 or 15 minutes. Barbell velocity (m × second -1 ) was monitored by PERCH from the frontal plane, whereas all sessions were video recorded from the same view and subsequently analyzed by KINOVEA using natural barbell landmarks. Fifty repetitions were randomly and equally drawn from each participant ( n = 1,000 repetitions) across all workout and supplement conditions. Although Wilcoxon signed-rank test revealed no differences (mean difference = 0.03 m × second -1 , p = 0.332), KINOVEA only explained 2.6% of PERCH variance, and Spearman's correlation analysis of Bland-Altman plots suggested that differences between technologies were exacerbated by the magnitude of velocity ( ρ = -0.62, p < 0.001). Cross-validated, regression analysis using KINOVEA and readily available participant demographics (arm reach and biological sex) improved prediction by 13.4% but agreement remained inconsistent ( ρ = 0.73, p < 0.001). Neither KINOVEA nor KINOVEA combined with participant demographics are interchangeable with PERCH for monitoring barbell velocity during an AMRAP. Still, researchers and athletes might use this free and accessible software as back-up for whenever PERCH fails to detect repetitions, as some information is better than none.

Validation of two novel human activity recognition models for typically developing children and children with Cerebral Palsy.

Human Activity Recognition models have potential to contribute to valuable and detailed knowledge of habitual physical activity for typically developing children and children with Cerebral Palsy. The main objective of the present study was to develop and validate two Human Activity Recognition models. One trained on data from typically developing children (n = 63), the second also including data from children with Cerebral Palsy (n = 16), engaging in standardised activities and free play. Our data was collected using accelerometers and ground truth was established with video annotations. Additionally, we aimed to investigate the influence of window settings on model performance. Utilizing the Extreme gradient boost (XGBoost) classifier, twelve sub-models were created, with 1-,3- and 5-seconds windows, with and without overlap. Both Human Activity Recognition models demonstrated excellent predictive capabilities (>92%) for standardised activities for both typically developing and Cerebral Palsy. From all window sizes, the 1-second window performed best for all test groups. Accuracy was slightly lower (>75%) for the Cerebral Palsy test group performing free play activities. The impact of window size and overlap varied depending on activity. In summary both Human Activity Recognition models effectively predict standardised activities, surpassing prior models for typically developing and children with Cerebral Palsy. Notably, the model trained on combined typically developing children and Cerebral Palsy data performed exemplary across all test groups. Researchers should select window settings aligned with their specific research objectives.

Consistent effect of eating rate on food and energy intake across twenty-four ad libitum meals.

Foods consumed at lower eating rates (ER) lead to reductions in energy intake. Previous research has shown that texture-based differences in eating rateER can reduce meal size. The effect size and consistency of these effects across a wide range of composite and complex meals differing considerably in texture and varying in meal occasion have not been reported. We determined how consistently texture-based differences in ER can influence food and energy intake across a wide variety of meals. In a crossover design, healthy participants consumed twelve breakfast and twelve lunch meals that differed in texture to produce a fast or slow ER. A breakfast group (n = 15) and lunch group (n = 15) completed twelve ad libitum meal sessions each (six 'fast' and six 'slow' meals), where intake was measured and behavioural video annotation was used to characterise eating behaviour. Liking did not differ significantly between fast and slow breakfasts (P = 0·44) or lunches (P = 0·76). The slow meals were consumed on average 39 % ± 9 % (breakfast) and 45 % ± 7 % (lunch) slower than the fast meals (both P < 0·001). Participants consumed on average 22 % ± 5 % less food (84 g) and 13 % ± 6 % less energy (71 kcal) from slow compared with fast meals (mean ± SE; P < 0·001). Consuming meals with a slower ER led to a reduction in food intake, where an average decrease of 20 % in ER produced an 11 % ± 1 % decrease in food intake (mean ± SE). These findings add to the growing body of evidence showing that ER can be manipulated using food texture and that this has aits consistent effect on food and energy intake across a wide variety of Hedonically equivalent meals.

Which Gait Tasks Produce Reliable Outcome Measures of Freezing of Gait in Parkinson's Disease?

Measurement of freezing of gait (FOG) relies on the sensitivity and reliability of tasks to provoke FOG. It is currently unclear which tasks provide the best outcomes and how medication state plays into this. To establish the sensitivity and test-retest reliability of various FOG-provoking tasks for presence and severity of FOG, with (ON) and without (OFF) dopaminergic medication. FOG-presence and percentage time frozen (% TF) were derived from video annotations of a home-based FOG-provoking protocol performed in OFF and ON. This included: the four meter walk (4MW), Timed Up and Go (TUG) single (ST) and dual task (DT), 360° turns in ST and DT, a doorway condition, and a personalized condition. Sensitivity was tested at baseline in 63 definite freezers. Test-retest reliability was evaluated over 5 weeks in 26 freezers. Sensitivity and test-retest reliability were highest for 360° turns and higher in OFF than ON. Test-retest intra-class correlation coefficients of % TF varied between 0.63-0.90 in OFF and 0.18-0.87 in ON, and minimal detectable changes (MDCs) were high. The optimal protocol included TUG ST, 360° turns ST, 360° turns DT and a doorway condition, provoking FOG in all freezers in OFF and 91.9% in ON and this could be done reliably in 95.8% (OFF) and 84.0% (ON) of the sample. Combining OFF and ON further improved outcomes. The highest sensitivity and reliability was achieved with a multi-trigger protocol performed in OFF + ON. However, the high MDCs for % TF underscore the need for further optimization of FOG measurement.

Video Annotations Contribute to Coach and Teacher Conversations during Coaching Cycles

Video Annotations Contribute to Coach and Teacher Conversations during Coaching Cycles

Landmark palpation with video annotation tool for measuring scapular dyskinesis: An intra-tester reliability study

Landmark palpation with video annotation tool for measuring scapular dyskinesis: An intra-tester reliability study

Using expansive learning to design and implement video-annotated peer feedback in an undergraduate general education module

AbstractExisting studies have measured the effect of video-based feedback on student performance or satisfaction. Other issues are underacknowledged or merit further investigation. These include sociocultural aspects which may shape the design and implementation of video-based feedback, the ways students use technology to engage in feedback, and the processes through technology may transform learning. This study investigates the design and implementation of a video-annotated peer feedback activity to develop students’ presentation skills and knowledge of climate science. It explores how their use of a video annotation tool re-mediated established feedback practices and how the systematic analysis of contradictions in emerging practices informed the subsequent redesign and reimplementation of the approach. Employing a formative intervention design, the researchers intervened in the activity system of a first-year undergraduate education module to facilitate two cycles of expansive learning with an instructor and two groups of Hong Kong Chinese students (n = 97, n = 94) across two semesters. Instructor interviews, student surveys, and video annotation and system data were analysed using Activity Theory-derived criteria to highlight contradictions in each system and suggest how these could be overcome. The findings highlight the critical importance of active instructor facilitation; building student motivation by embedding social-affective support and positioning peer feedback as an integrated, formative process; and supporting students’ use of appropriate cognitive scaffolding to encourage their interactive, efficient use of the annotation tool. Conclusions: In a field dominated by experimental and quasi-experimental studies, this study reveals how an Activity Theory-derived research design and framework can be used to systemically analyse cycles of design and implementation of video-annotated peer feedback. It also suggests how the new activity system might be consolidated and generalised.

Video Unsupervised Domain Adaptation with Deep Learning: A Comprehensive Survey

Video analysis tasks such as action recognition have received increasing research interest with growing applications in fields such as smart healthcare, thanks to the introduction of large-scale datasets and deep learning-based representations. However, video models trained on existing datasets suffer from significant performance degradation when deployed directly to real-world applications due to domain shifts between the training public video datasets (source video domains) and real-world videos (target video domains). Further, with the high cost of video annotation, it is more practical to use unlabeled videos for training. To tackle performance degradation and address concerns in high video annotation cost uniformly, the video unsupervised domain adaptation (VUDA) is introduced to adapt video models from the labeled source domain to the unlabeled target domain by alleviating video domain shift, improving the generalizability and portability of video models. This paper surveys recent progress in VUDA with deep learning. We begin with the motivation of VUDA, followed by its definition, and recent progress of methods for both closed-set VUDA and VUDA under different scenarios, and current benchmark datasets for VUDA research. Eventually, future directions are provided to promote further VUDA research. The repository of this survey is provided at https://github.com/xuyu0010/awesome-video-domain-adaptation.

A microdiscectomy surgical video annotation framework for supervised machine learning applications.

Lumbar discectomy is among the most common spine procedures in the US, with 300,000 procedures performed each year. Like other surgical procedures, this procedure is not excluded from potential complications. This paper presents a video annotation methodology for microdiscectomy including the development of a surgical workflow. In future work, this methodology could be combined with computer vision and machine learning models to predict potential adverse events. These systems would monitor the intraoperative activities and possibly anticipate the outcomes. A necessary step in supervised machine learning methods is video annotation, which involves labeling objects frame-by-frame to make them recognizable for machine learning applications. Microdiscectomy video recordings of spine surgeries were collected from a multi-center research collaborative. These videos were anonymized and stored in a cloud-based platform. Videos were uploaded to an online annotation platform. An annotation framework was developed based on literature review and surgical observations to ensure proper understanding of the instruments, anatomy, and steps. An annotated video of microdiscectomy was produced by a single surgeon. Multiple iterations allowed for the creation of an annotated video complete with labeled surgical tools, anatomy, and phases. In addition, a workflow was developed for the training of novice annotators, which provides information about the annotation software to assist in the production of standardized annotations. A standardized workflow for managing surgical video data is essential for surgical video annotation and machine learning applications. We developed a standard workflow for annotating surgical videos for microdiscectomy that may facilitate the quantitative analysis of videos using supervised machine learning applications. Future work will demonstrate the clinical relevance and impact of this workflow by developing process modeling and outcome predictors.

Auxiliary audio–textual modalities for better action recognition on vision-specific annotated videos

Most current audio–visual datasets are class-relevant, where audio and visual modalities are annotated. Thus, current audio–visual recognition methods apply cross-modality attention or modality fusion. However, leveraging the audio modality effectively in vision-specific videos for human activity recognition is of particular challenge. We address this challenge by proposing a novel audio–visual recognition framework that effectively leverages audio modality in any vision-specific annotated dataset. The proposed framework employs language models (e.g., GPT-3, CPT-text, BERT) for building a semantic audio–video label dictionary (SAVLD) that serves as a bridge between audio and video datasets by mapping each video label to its most K-relevant audio labels. Then, SAVLD along with a pre-trained audio multi-label model are used to estimate the audio–visual modality relevance. Accordingly, we propose a novel learnable irrelevant modality dropout (IMD) to completely drop the irrelevant audio modality and fuse only the relevant modalities. Finally, for the efficiency of the proposed multimodal framework, we present an efficient two-stream video Transformer to process the visual modalities (i.e., RGB frames and optical flow). The final predictions are re-ranked with GPT-3 recommendations of the human activity classes. GPT-3 provides high-level recommendations using the labels of the detected visual objects and the audio predictions of the input video. Our framework demonstrated a remarkable performance on the vision-specific annotated datasets Kinetics400 and UCF-101 by outperforming most relevant human activity recognition methods.

A deep-learning-based threshold-free method for automated analysis of rodent behavior in the forced swim test and tail suspension test

BackgroundThe forced swim test (FST) and tail suspension test (TST) are widely used to assess depressive-like behaviors in animals. Immobility time is used as an important parameter in both FST and TST. Traditional methods for analyzing FST and TST rely on manually setting the threshold for immobility, which is time-consuming and subjective. New methodWe proposed a threshold-free method for automated analysis of mice in these tests using a Dual-Stream Activity Analysis Network (DSAAN). Specifically, this network extracted spatial information of mice using a limited number of video frames and combined it with temporal information extracted from differential feature maps to determine the mouse’s state. To do so, we developed the Mouse FSTST dataset, which consisted of annotated video recordings of FST and TST. ResultsBy using DSAAN methods, we identify immobility states at accuracies of 92.51 % and 88.70 % for the TST and FST, respectively. The predicted immobility time from DSAAN is nicely correlated with a manual score, which indicates the reliability of the proposed method. Importantly, the DSAAN achieved over 80 % accuracy for both FST and TST by utilizing only 94 annotated images, suggesting that even a very limited training dataset can yield good performance in our model. Comparison with existing method(s)Compared with DBscorer and EthoVision XT, our method exhibits the highest Pearson correlation coefficient with manual annotation results on the Mouse FSTST dataset. ConclusionsWe established a powerful tool for analyzing depressive-like behavior independent of threshold, which is capable of freeing users from time-consuming manual analysis.

Standardizing male pelvic anatomy teaching using a clinical enrichment video.

Evolving trends in medical education and modern curricular changes have reduced the amount of time and resources allocated for anatomy education. As the amount of dedicated anatomy education time decreases, more self-directed learning is required. Cadaveric dissection and didactic teaching have been supplemented with multimedia, clinical anatomy, and imaging for over 20 years, with mixed results. Specifically, the use of video-based anatomy teaching increases medical learning, if done methodically. A 20-minute video was produced highlighting surgical anatomy using the following operative cases: perineal anatomy (artificial urinary sphincter case), inguinal and testicular anatomy (scrotal orchidopexy for acute testicular torsion), prostate anatomy (robotic radical prostatectomy), and bladder anatomy (endoscopy). The annotated video was shown to first-year medical students. Pre- and post-video multiple choice question quizzes were given to the students. Once submitted, the students completed a survey. Overall, 191 first-year medical students participated in our study. Average scores were similar between each quiz (50±16% vs. 49±17%) and there was no statistically significant change. Seventy-seven percent of participants felt the video improved their knowledge of urologic anatomy and 83% agreed the video should be shown to future classes. Sixty percent of participants felt the video solidified their anatomy knowledge, 78% felt the video was stimulating and entertaining, and 43% of the students felt the video increased their interest in pursuing urology as a career choice. Anatomy teaching can be supplemented using surgical videos, especially in a time when in-person anatomy teaching is limited. Further study is required to determine whether this teaching modality improves long-term anatomy knowledge retention.

Differences in small-breed dogs' body language and vocalizations in a negative context

A dog's emotional state is communicated primarily through body language, readily observable by humans and other animals. Additionally, dogs express visual and auditory signals differently according to the situation they are experiencing. However, there is a need for more research on smaller breeds since they tend to display more behavioral issues than larger breeds. The objective of the present study was to identify the behavioral characteristics of 25 dogs of three different small breeds (Chihuahua, Toy Poodle, and Miniature Schnauzer) in a negative context. We used a web-based tool called the Dog Actions Video Annotation Interface (DAVAI) to evaluate the movements of the tail, legs, snout, and vocalizations when the dogs were exposed to unfamiliar people. Our results showed no significant differences in dog behavior associated with breed. However, we identified variations in dog vocalizations explained by the sex, with male dogs whining more frequently than females. We highlight the importance of understanding dog body language and vocalizations, regardless of their breed. Our study is particularly relevant in negative contexts, where dogs may exhibit distress signals that require prompt intervention to alleviate their suffering.

Assessing two methods of webcam-based eye-tracking for child language research.

We assess the feasibility of conducting web-based eye-tracking experiments with children using two methods of webcam-based eye-tracking: automatic gaze estimation with the WebGazer.js algorithm and hand annotation of gaze direction from recorded webcam videos. Experiment 1 directly compares the two methods in a visual-world language task with five to six year-old children. Experiment 2 more precisely investigates WebGazer.js' spatiotemporal resolution with four to twelve year-old children in a visual-fixation task. We find that it is possible to conduct web-based eye-tracking experiments with children in both supervised (Experiment 1) and unsupervised (Experiment 2) settings - however, the webcam eye-tracking methods differ in their sensitivity and accuracy. Webcam video annotation is well-suited to detecting fine-grained looking effects relevant to child language researchers. In contrast, WebGazer.js gaze estimates appear noisier and less temporally precise. We discuss the advantages and disadvantages of each method and provide recommendations for researchers conducting child eye-tracking studies online.

Improving the Accuracy of Recognition and Evaluation of Technical Movements of Basketball Players Using Deep Learning Algorithms

Deep learning algorithms are revolutionizing the analysis of technical movements in basketball players by extracting intricate patterns and insights from vast amounts of video data. By training neural networks on annotated video sequences of basketball games, these algorithms can automatically detect and classify various technical movements such as dribbling, shooting, passing, and defensive maneuvers. The use of deep learning enables the identification of subtle nuances in player movements, facilitating more accurate performance assessment and actionable feedback for athletes and coaches. This paper introduces a novel approach for analyzing basketball player movements utilizing Anthropometrical Variable Assessment Deep Learning (AVADL). By integrating anthropometric variables with deep learning algorithms, AVADL offers a comprehensive framework for accurately recognizing and evaluating technical movements on the basketball court. We present experimental results demonstrating the effectiveness of AVADL across various dataset sizes and player profiles, showcasing high accuracy and performance metrics. The incorporation of anthropometric measurements provides valuable context into the diverse physical attributes of basketball players, enhancing our understanding of their playing style and performance. Experimental results demonstrate the effectiveness of AVADL across various dataset sizes and player profiles, with training accuracies ranging from 90% to 97% and testing accuracies from 85% to 92%. Precision, recall, and F1-Score metrics consistently show values above 0.80, indicating the robustness of the approach. The incorporation of anthropometric measurements provides valuable context into the diverse physical attributes of basketball players, enhancing our understanding of their playing style and performance.

Structured feedback and operative video debriefing with critical view of safety annotation in training of laparoscopic cholecystectomy: a randomized controlled study

BackgroundThe learning curve in minimally invasive surgery (MIS) is lengthened compared to open surgery. It has been reported that structured feedback and training in teams of two trainees improves MIS training and MIS performance. Annotation of surgical images and videos may prove beneficial for surgical training. This study investigated whether structured feedback and video debriefing, including annotation of critical view of safety (CVS), have beneficial learning effects in a predefined, multi-modal MIS training curriculum in teams of two trainees.MethodsThis randomized-controlled single-center study included medical students without MIS experience (n = 80). The participants first completed a standardized and structured multi-modal MIS training curriculum. They were then randomly divided into two groups (n = 40 each), and four laparoscopic cholecystectomies (LCs) were performed on ex-vivo porcine livers each. Students in the intervention group received structured feedback after each LC, consisting of LC performance evaluations through tutor-trainee joint video debriefing and CVS video annotation. Performance was evaluated using global and LC-specific Objective Structured Assessments of Technical Skills (OSATS) and Global Operative Assessment of Laparoscopic Skills (GOALS) scores.ResultsThe participants in the intervention group had higher global and LC-specific OSATS as well as global and LC-specific GOALS scores than the participants in the control group (25.5 ± 7.3 vs. 23.4 ± 5.1, p = 0.003; 47.6 ± 12.9 vs. 36 ± 12.8, p < 0.001; 17.5 ± 4.4 vs. 16 ± 3.8, p < 0.001; 6.6 ± 2.3 vs. 5.9 ± 2.1, p = 0.005). The intervention group achieved CVS more often than the control group (1. LC: 20 vs. 10 participants, p = 0.037, 2. LC: 24 vs. 8, p = 0.001, 3. LC: 31 vs. 8, p < 0.001, 4. LC: 31 vs. 10, p < 0.001).ConclusionsStructured feedback and video debriefing with CVS annotation improves CVS achievement and ex-vivo porcine LC training performance based on OSATS and GOALS scores.

Man and the Machine: Effects of AI-assisted Human Labeling on Interactive Annotation of Real-time Video Streams

AI-assisted interactive annotation is a powerful way to facilitate data annotation—a prerequisite for constructing robust AI models. While AI-assisted interactive annotation has been extensively studied in static settings, less is known about its usage in dynamic scenarios where the annotators operate under time and cognitive constraints, e.g., while detecting suspicious or dangerous activities from real-time surveillance feeds. Understanding how AI can assist annotators in these tasks and facilitate consistent annotation is paramount to ensure high performance for AI models trained on these data. We address this gap in interactive machine learning (IML) research, contributing an extensive investigation of the benefits, limitations, and challenges of AI-assisted annotation in dynamic application use cases. We address both the effects of AI on annotators and the effects of (AI) annotations on the performance of AI models trained on annotated data in real-time video annotations. We conduct extensive experiments that compare annotation performance at two annotator levels (expert and non-expert) and two interactive labeling techniques (with and without AI assistance). In a controlled study with \(N=34\) annotators and a follow-up study with 51,963 images and their annotation labels being input to the AI model, we demonstrate that the benefits of AI-assisted models are greatest for non-expert users and for cases where targets are only partially or briefly visible. The expert users tend to outperform or achieve similar performance as the AI model. Labels combining AI and expert annotations result in the best overall performance as the AI reduces overflow and latency in the expert annotations. We derive guidelines for the use of AI-assisted human annotation in real-time dynamic use cases.

Functional Analysis of English Carriers and Related Resources of Cultural Communication in Internet Media

Abstract In the Internet intelligent teaching platform, students’ demand for English cultural content is increasingly obvious. To help students quickly locate the overall content of resources in online autonomous learning, this study constructs a video annotation model for online teaching. This method classifies text by designing an optimized Bidirectional Encoder Representation from the Transformers model and designs a Text Rank keyword extraction model that integrates external knowledge and semantic feature weights. The extraction of knowledge points contained in audio and video resources can be realized. In the experimental data set, a relatively complete video content summary could be obtained by combining the first three sentences with the last two sentences. The F1 value of the classification model was up to 91.3%. In addition, the BERT-T model proposed in this article had the best effect on the experiment. Compared with the original BERT model, the macro-F1 was 0.8% higher and 0.5% higher than the Ro BERTA model. In the keyword extraction experiment, B-Text Rank was 2.19 and 2.85% higher than the traditional Text Rank in the two datasets. The experiment shows that the BERT-Text Rank network resource annotation model has excellent application performance in English online autonomous teaching and could guide students to learn.