Multi-Modal Transformer Network for Classroom Instruction Analysis

Creativity is a significant element in design education, and frequently a significant competency during recruitment for design professions. Group work and individual work are widely employed in higher education. Many studies have highlighted the merits of employing group work in design education, cultivating collaborative design abilities and fostering sought-after employability skills. Although the benefits of group work in design practice and education are widely recognised, few studies have shown evidence that group work outperforms individual work regarding creative design activities in higher education contexts. Therefore, the aim of this research is to explore whether group or individual work is more beneficial for fostering students in generating creative designs in STEM design education. A case study, involving two cohorts of second-year undergraduate students studying a UK Engineering degree Industrial Design programme, is reported. The case study compares the design outputs produced by the two cohorts tackling the same design challenge in a product design module but employing individual and group work, respectively. The case study results show that no significant differences have been found between the design outputs produced by group work and individual work, considering novelty, usefulness and overall creativity. Further analysis reveals that a student’s academic performance is not significantly related to the level of creativity of the design produced. This research indicates design educators should employ both group and individual work to complement each other in design education, and suggests potential solutions to enhance students’ design creativity.

Automatic identification of human activity has led to a possibility of providing personalised services in different domains i.e. healthcare, security and sport etc. With advancement in sensor technology, automatic activity recognition can be done in an unobtrusive and non-intrusive way. The placement of the sensor and wearability are ones of vital keys in the successful activity recognition of free space livings. Experiments were carried out to investigate the use of a single wrist-worn accelerometer for automatic activity classification. The performances of two classification algorithms namely Decision Tree C4.5 and Artificial Neural Network were compared using four different sets of features to classify five daily living activities. The result revealed that Decision Tree C4.5 has outperformed Neural Network regardless of the different sets of features used. The best classification result was achieved using the set containing the most popular and accurate features i.e. mean, minimum, energy and sample differences etc. The best accuracy of 94.13% was achieved using only wrist-worn accelerometer showing a possibility of automatic activity classification with no movement constrain, discomfort and stigmatisation caused by the sensor.

Teaching and learning occurs in three main instructional settings, i.e., whole class, small group or individual work. Each instructional setting provides affordances and constraints in terms of the types of teaching and learning strategies that can be implemented effectively, which in turn impacts on students’ opportunity to learn. Variation between teachers’ use of instructional settings within a culture is likely to be a reflection of their beliefs and instructional style whilst variation in teachers’ use of instructional settings across cultures is likely to reflect the cultural framing of teaching and learning, and curriculum emphases within those cultures.

Research on procrastination mostly focuses on person-related antecedents and neglects situational and social factors, such as group work. Prior research indicates that conjunctive and additive group work may increase individual effort and performance as compared to individual work. Based on these findings, we investigate whether conjunctive and additive group work may also help reduce procrastination as compared to individual work. In a registered field experiment, N = 218 students with high levels of trait procrastination worked on an academic task over the course of 10 days in one of three conditions (individual work vs. conjunctive group work vs. additive group work). Dependent variables comprised task procrastination, task performance, and positive and negative task-related affect. Regarding conjunctive group work, results are mixed, with some evidence that conjunctive group work leads to lower procrastination as compared to individual work. Both types of group work resulted in higher negative task-related affect when assessed prospectively. No other effects were found. The findings contribute to the idea that targeted changes in the learning environment, such as the implementation of group work, may help reduce procrastination.

In mixed-focus collaboration, users continuously switch between and work. We have developed a new two-person interaction mechanism, coupled tele-desktops, that is, arguably, not biased towards individual or group work. We evaluate this mechanism, and the general idea of mixed-focus collaboration, using a new quantitative framework consisting of (a) a set of precisely-defined coupling modes determining the extent of individual and group work, and (b) the times spent in, durations of, and number of transitions among these modes. We describe a new visualization scheme for compactly displaying these metrics in an individual collaborative session. We use this framework to characterize about forty six person hours of use of coupled tele-desktops, most of which involved collaborative use of a UI builder. Our results include (a) quantitative motivation for coupled tele-desktops, and (b) several new quantitative observations, and quantification of several earlier qualitative observations regarding mixed-focus collaboration.

In organizations, decisions usually involve multiple persons working together, spending considerable time in meetings. Group decision support systems (GDSS) are designed to help groups in meetings reach consensus. Such systems started with facilities (called decision rooms) in which people work together at the same time and in the same place. Such decision rooms contain individual computers where people can do private work, public screens seen by all, and networks and software to support both group and individual work. Over time, GDSS has expanded to include people located in different places and at different times. Although GDSS hardware is mostly off-the-shelf, specialized software is used for generating organizing, and prioritizing ideas, for organizational memory, and other tasks associated with group work. This chapter discusses the nature of GDSS, includes brief descriptions of early decision rooms, and considers major software vendors. The chapter concludes that GDSS is now a mature technology, many of whose concepts are now embedded in the way organizations work, and the major legacy of University research is to practice what was learned about individual and group behaviors in computer-based environments.

The methodology and content of the project activities of students in biology is relevant in a modern school and should be applied in educational activities. Project activities have a fruitful effect on the personality of students, include them in social and research work. This work provides for the implementation of a system of sequential actions, makes it possible to express one's opinion and realize individual goals and objectives, to realize ideas, forms competencies aimed at searching and analyzing information, collecting data, observing and experimenting, helps develop communication skills and contribute to the product by group or individual work. Also, students most often acquire competence in the use of ICT. Each student acts simultaneously as an organizer, executor, expert of activity. Since the biology curriculum includes knowledge from many academic disciplines, the project activity clearly implements interdisciplinary connections. The knowledge acquired by students in the course of project activities is more solid. The analysis of work on project activities shows that students increase the level of formation of key competencies (problem setting, goal setting, planning and evaluation of the result, search and processing of information, written communication, oral presentation, group work), which is recommended in the Federal State Educational Standard.Based on the data of various events for students in the Samara region, an analysis of the topics, content and quality of projects in biology and ecology was carried out, the main shortcomings were identified, and recommendations were given for their elimination. Developed, implemented and presented as examples of projects in grades 5-9 on the general topic «Medicinal plants». The proposed study is of practical importance, since the materials and conclusions can be used in the organization of educational and scientific and educational work of schoolchildren in the process of bioecological education and methodological training of students in the course of industrial pedagogical practices on the basis of Samara State University of Social Sciences and Education.

Automatic action recognition in videos is a challenging computer vision task that has become an active research area in recent years. Existing strategies usually use kernel-based learning algorithms that considers a simple combination of different features completely disregarding how such features should be integrated to fit the given problem. Since a given feature is most suitable to describe a given image/video property, the adaptive weighting of such features can improve the performance of the learning algorithm. In this paper, we investigated the use of the Multiple Kernel Learning (MKL) algorithm to adaptive search for the best linear relation among the considered features. MKL is an extension of the support vector machines (SVMs) to work with a weighted linear combination of several single kernels. This approach allows to simultaneously estimate the weights for the multiple kernels combination as well as the underlying SVM parameters. In order to prove the validity of the MKL approach, we considered a descriptor composed of multiple features aligned with dense trajectories. We experimented our approach on a database containing 36 cooking actions. Results confirm that the use of MKL improves the classification performance.

This paper presents a novel action recognition method for observing human workers using interactions between actions and related objects on an internal construction site. This method can be used to measure work rates for labour productivity monitoring. This monitoring is critical because the performance of a construction project is significantly impacted by labour productivity. However, construction sites are generally crowded with a large number of workers and objects. Such congestion disrupts the accurate, automatic recognition of construction workers’ actions. This congestion is one reason that existing automatic action recognition studies of construction areas mainly focus on workers’ actions themselves. However, the crowded conditions mean that sites could offer a great deal of clues that could be used for automatic action recognition. According to psychological studies, interactions clearly take place between human actions and related objects, such as between hammering and a hammer. Humans use these interactions to recognize actions or objects more accurately. On the construction site, workers, materials, tools, and equipment are carefully planned out ahead of actual construction. The categories of workers and objects are pre-defined and, as noted, specific interactions define relations between worker actions and objects. In this paper, the interactions are limited to human workers and their hand-held objects. Action recognition results can be combined with hand-held object information to improve recognition accuracy. With the limited interactions, experiments in this paper show a significant improvement in action recognition. This paper describes the utilization of these interactions to improve construction action recognition accuracy based on human skeleton data and 2D color video from Microsoft KINECT sensor.

The advancement in technologies such as Artificial Intelligence (AI) and the Internet of Things (IoT) has enabled the development of a variety of healthcare applications and services such as Remote Health Monitoring (RHM). Developing an RHM system requires integrating various medical sensors that generate a vast amount of data. The integration of AI and Machine Learning (ML) models in the RHM system for analysis of the medical data can enable the development of a decision support system that can reduce costs and save valuable time for clinicians and caregivers. This thesis is focused on leveraging ML and AI technique sto address the challenges in RHM with a specific emphasis on improving the performance of AI-based decision support systems. This is achieved by addressing limitations in current AI-based models and developing novel Random Neural Network (RNN)-based models for fall detection and epilepsy detection and prediction. Initially the thesis presents a literature review on state-of-the-art machine learning techniques used in remote healthcare solutions, automatic fall activity recognition for elderly populations, and epileptic seizure detection and prediction. Furthermore, the study presents a novel fall detection method based on RNN. The results obtained from the proposed model are compared with traditional classification algorithms such as Artificial Neural Networks, Convolutional Neural Networks, and state-of-the-art fall detection schemes. The proposed RNN-based method demonstrates significant improvements in accuracy compared to other methods. The proposed scheme achieved an overall accuracy of 98% in detecting falls. Additionally, several other parameters such as precision, recall, specificity, and F-measure show that the proposed algorithm has better generalization capabilities when compared with other traditional machine learning schemes. Furthermore, the proposed RNN is also compared with a recent scheme and the obtained results demonstrate the superiority of the proposed scheme. Additionally, the thesis presents two novel RNN-based classification schemes for detecting epileptic seizures. In the first scheme statistical features are extracted from raw EEG data, while the second method incorporates frequency analysis techniques such as Discrete Wavelet Transform (DWT) and Fast Fourier Transform. The performance of these models is evaluated and compared with other classification algorithms, demonstrating the effectiveness of RNN-based models for epilepsy detection. Finally, a novel RNN-based epilepsy prediction model is proposed for classification of EEG data into three classes i.e. ictal, pre-ictal, and inter-ictal. The model pre-processes the EEG data by decomposing the signals into five frequency sub-bands using DWT. After pre-processing, six statistical features are extracted from each frequency sub-band which a reused as an input vector to train the RNN model. The proposed scheme achieves an overall accuracy of 95.66% in seizure prediction, outperforming other state-of-the-art models.

Well integrity monitoring, especially annulus pressure surveillance, is essential for detecting abnormalities and maintaining safe production. Modern sensors enable real-time measurement of pressure, temperature, and other parameters in the well and at the wellhead. Even so, human intervention is currently required to interpret data trends, flag issues and make decisions. Machine learning may strengthen this process by automating activity recognition and detecting anomalies in near real time, reducing reliance on manual analysis on research datasets, subject to further validation before any operational use. We compare two machine learning approaches: an Inference System and Long Short-Term Memory (LSTM), by applying them to time series data for production activity recognition and abnormality detection. The dataset, consisting of minute-by-minute measurements of pressure, temperature, and valve statuses from four oil production wells in the Gulf of America, was pre-processed through cleaning and imputation before being used for model training and evaluation. The inference system combines expert knowledge with statistically derived insights to create rules as part of a rule-based method. In contrast, the LSTM model learned from labeled input data. Both approaches first classified production activities into categories such as well shut-in-steady, well shut-in-transient, production-steady, production-ramp-up, production-ramp-down, production-tubing-side or production-annulus-side valve status change, and other well activities. This classification helped better interpret real-time well activities and increase the accuracy of anomaly detection. Pressure and temperature abnormalities primarily detected during steady-state activities, such as well shut-in-steady and production-steady, were used to assess well integrity. Here, the rule-based method relied on statistical analysis to define rules for abnormality detection, while the LSTM model achieved the same task by learning from supervised data. On the evaluated research dataset, results show that the LSTM model outperformed the inference system, achieving an F1 score of 94% for well activity classification and 91% for abnormality detection. These results reflect performance on a limited dataset and do not imply predictive accuracy in field operations. The machine-learning methods and results presented in this paper are research-oriented and do not represent the operator´s engineering standards and operating procedures. The methods detailed in this paper illustrate how automatic well activity recognition may be approached, enhances abnormality detection, supports earlier well integrity issue identification, minimizes dependence on human judgment, and supports root-cause identification of well integrity issues when combined with other techniques. Such automated routines suggest feasibility as future digital tools for a well intervention operation may play a role in future research and system development.

Markerless human motion tracking can be employedin many applications such as automatic surveillance, motioncapture, human-machine interface and activity recognition. Thisproblem has been extensively studied in the computer visionresearch community in the last years. In this context, the presentpaper presents an approach for 3D markerless human motiontracking based on a skeletal kinematic model of the human body.This method is applied over a 3D probabilistic occupancy gridof the environment, which is constructed by means of a Bayesianfusion of images from multiple synchronized sensoring cameras.Although the use of kinematic models in 3D human tracking iswidely employed, its use over 3D probabilistic occupancy gridsstill was not vastly investigated in the literature. The experimentswere performed using a public dataset with video sequences ofpeople in motion. The results show that the method is capableof dealing adequately with the 3D markerless human motiontracking problem.

In last few decades, technological revolution has accelerated the deployment of large scale surveillance systems on almost all public places such as malls, hospitals, airports, railways, bus stations, roads, etc. These intelligent surveillance systems can play crucial role in governance of situations, collective security and safety, mitigating as well as prevention of adversaries. With gradual increase in multi camera surveillance systems enclosing multi angle views of same as well as different scenes has increased complexity of monitoring the systems by manual inspection. Abnormalities also known as anomalies or outliers are inevitable part of the existence and presumed to be rare in occurrence. Manual monitoring of such abnormalities is susceptible to errors and limited by human capabilities such as inattention and tiresome. Hence in the field of computer vision, automated abnormal activity recognition (AAR) from surveillance systems is emerging research area. The intent of this research is to shed a light on recent innovations and developments that have made a mark in abnormal activity recognition (AAR) involving deep learning. This paper also includes conventional categorization of anomalies based on different perspectives which can provide better understanding to young researchers. Though recent developments still poses many real time challenges in automatic abnormal activity recognition, some of them are enclosed in this paper.

Introduction: Automatic surgical activity recognition in the operating room (OR) is mandatory to enable assistive surgical systems to manage the information presented to the surgical team. Therefore the purpose of our study was to develop and evaluate an activity recognition model. Material and methods: The system was conceived as a hierarchical recognition model which separated the recognition task into activity aspects. The concept used radio frequency identification (RFID) for instrument recognition and accelerometers to infer the performed surgical action. Activity recognition was done by combining intermediate results of the aspect recognition. A basic scheme of signal feature generation, clustering and sequence learning was replicated in all recognition subsystems. Hidden Markov models (HMM) were used to generate probability distributions over aspects and activities. Simulated functional endoscopic sinus surgeries (FESS) were used to evaluate the system. Results and discussion: The system was able to detect surgical activities with an accuracy of 95%. Instrument recognition performed best with 99% accuracy. Action recognition showed lower accuracies with 81% due to the high variability of surgical motions. All stages of the recognition scheme were evaluated. The model allows distinguishing several surgical activities in an unconstrained surgical environment. Future improvements could push activity recognition even further.

This study presents a deep learning model devoted to the analysis of swimming using a single Inertial Measurement Unit (IMU) attached to the sacrum. Gyroscope and accelerometer data were collected from 35 swimmers with various expertise levels during a protocol including the four swimming techniques. The proposed methodology took high inter- and intra-swimmer variability into account and was set up for the purpose of predicting eight swimming classes (the four swimming techniques, rest, wallpush, underwater, and turns) at four swimming velocities ranging from low to maximal. The overall F1-score of classification reached 0.96 with a temporal precision of 0.02 s. Lap times were directly computed from the classifier thanks to a high temporal precision and validated against a video gold standard. The mean absolute percentage error (MAPE) for this model against the video was 1.15%, 1%, and 4.07%, respectively, for starting lap times, middle lap times, and ending lap times. This model is a first step toward a powerful training assistant able to analyze swimmers with various levels of expertise in the context of in situ training monitoring.