Search

In Human action recognition, the identification of actions is a system that can detect human activities. The types of human activity are classified into four different categories, depending on the complexity of the steps and the number of body parts involved in the action, namely gestures, actions, interactions, and activities [1]. It is challenging for video Human action recognition to capture useful and discriminative features because of the human body's variations. To obtain Intelligent Solutions for action recognition, it is necessary to training models to recognize which action is performed by a person. This paper conducted an experience on Human action recognition compare several deep learning models with a small dataset. The main goal is to obtain the same or better results than the literature, which apply a bigger dataset with the necessity of high-performance hardware. Our analysis provides a roadmap to reach the training, classification, and validation of each model.

Despite recent efforts, accuracy in group emotion recognition is still generally low. One of the reasons for these underwhelming performance levels is the scarcity of available labeled data which, like the literature approaches, is mainly focused on still images. In this work, we address this problem by adapting an inflated ResNet-50 pretrained for a similar task, activity recognition, where large labeled video datasets are available. Audio information is processed using a Bidirectional Long Short-Term Memory (Bi-LSTM) network receiving extracted features. A multimodal approach fuses audio and video information at the score level using a support vector machine classifier. Evaluation with data from the EmotiW 2020 AV Group-Level Emotion sub-challenge shows a final test accuracy of 65.74% for the multimodal approach, approximately 18% higher than the official baseline. The results show that using activity recognition pretraining offers performance advantages for group-emotion recognition and that audio is essential to improve the accuracy and robustness of video-based recognition.

With the appearance of Shared Autonomous Vehicles there will no longer be a driver responsible for maintaining the car interior and well-being of passengers. To counter this, it is imperative to have a system that is able to detect any abnormal behaviors, more specifically, violence between passengers. Traditional action recognition algorithms build models around known interactions but activities can be so diverse, that having a dataset that incorporates most use cases is unattainable. While action recognition models are normally trained on all the defined activities and directly output a score that classifies the likelihood of violence, video anomaly detection algorithms present themselves as an alternative approach to build a good discriminative model since usually only non-violent examples are needed. This work focuses on anomaly detection and action recognition algorithms trained, validated and tested on a subset of human behavior video sequences from Bosch's internal datasets. The anomaly detection network architecture defines how to properly reconstruct normal frame sequences so that during testing, each sequence can be classified as normal or abnormal based on its reconstruction error. With these errors, regularity scores are inferred showing the predicted regularity of each frame. The resulting framework is a viable addition to traditional action recognition algorithms since it can work as a tool for detecting unknown actions, strange/violent behaviors and aid in understanding the meaning of such human interactions.

Violence detection behavior is a particular problem regarding the great problem action recognition. In recent years, the detection and recognition of violence has been studied for several applications, namely in surveillance. In this paper, we conducted a recent systematic review of the literature on this subject, covering a selection of various researched papers. The selected works were classified into three main approaches for violence detection: video, audio, and multimodal audio and video. Our analysis provides a roadmap to guide future research to design automatic violence detection systems. Techniques related to the extraction and description of resources to represent behavior are also reviewed. Classification methods and structures for behavior modelling are also provided., European Structural and Investment Funds in the FEDER component, through the Operational Competitiveness and Internationalization Programme (COMPETE 2020) [Project n ∘ 039334; Funding Reference: POCI-01-0247-FEDER-039334]. This work has been supported by national funds through FCT – Fundação para a Ciência e Tecnologia through project UIDB/04728/2020

Violence is a word that encompasses several meanings ranging from an actual fight to theft and several types of harassment. Therefore, violence detection through surveillance systems can be a quite difficult yet important task. The increasing use of carpooling services and vehicle sharing brought the need to implement a sufficient general surveillance system for monitoring these vehicles for assuring the passengers’ safety during the ride. This paper raised the literature for this matter, finding fewer research papers than it was expected for the in-vehicle perspective, noticeably to sexual harassment. Most of the research papers focused on out-vehicle issues such as runs over and vehicle theft. In-vehicle electronic components security and cockpit user experience were perceived as major concern areas. This paper discusses these findings and presents some insights about in-vehicle surveillance., This work is supported by the European Structural and Investment Funds in the FEDER component, through the Operational Competitiveness and Internationalization Programme (COMPETE 2020) [Project no 039334; Funding Reference: POCI-01-0247-FEDER-039334]., This work has been supported by national funds through FCT – Fundação para a Ciência e Tecnologia through project UIDB/04728/2020.

Driverless vehicles are more and more becoming a reality. However, people still have some concerns in using them, the main concern is fear, hence the importance of creating a surveillance system inside those vehicles. For the detection and classification of human movements to be possible it is necessary to train the system with data representative enough for all kinds of possibilities. Although the production of large quantities of data becomes an expensive process and adds the problem of data protection, the use of synthetic data once they are artificially generated allows lower costs and eliminates the problem of data protection. A bibliographic study was carried out in this paper with articles from 2017 or later on the use of synthetic data. In these studies, it is noted that synthetic data is widely used with good results. As far as image capture is concerned, they show that 3D cameras have better results, but they are more expensive, so 2D cameras are more often used with later conversion to 3D images. The stitched puppet (SP) model is capable of adapting to the most difficult poses having obtained good results in its application in the FAUST dataset.

An in-depth study of knowledge and technologies was made related to the various scientific, technical, and industrial domains necessary for the acquisition of skills and capabilities for the design and development of a multisensory fusion system for vehicle cockpits. After an extensive literature review, it was possible to determine the baselines of the solution to be developed and obtain a pipeline prototype.

Introduction Empathy is the mediating role of the physician–patient relationship. Through this process the practitioner seeks to understand the patient's frame of reference and also to establish a relationship of openness, mutual respect, trust and deep understanding. Objective This study aims to analyze the perceptions of medical students on the importance of empathy in the doctor–patient relationship, and to analyze the gender differences and in the different years of the course. It is also intended to identify a taxonomy of students based on their perceptions of empathy. Methods This cross-sectional study was conducted on undergraduate medical students. A total of 208 medical students responded to the Jefferson Scale of Physician Empathy–Student Portuguese Version (JSPE-spv). Principal Components Analysis with varimax rotation was used to identify the number and compositions of emerged factors. The scores of the retained factors were submitted to a cluster analysis to identify different groups of students, based on the dimension of empathy. A One-way Anova analysis and post-hoc tests of Tukey supported the identified cluster. Results This study shows statistically significant differences between genders, i.e., female individuals show higher scores than male individuals. Empathy increases over the course, but with no statistically significant differences. Six factors emerged from factor analysis, and three of them are empathy dimensions found in other studies: “adoption of perspectives”, “compassionate care”, and “standing in the patients shoes”. Cluster analysis divided the students into five groups according to the dimensions of empathy identified. Conclusions In this medical school in Portugal, it may be concluded that women are more empathetic than men, and empathy grows throughout the course. It is recommended to carry out a longitudinal study, in order to follow the evolution of these students, as well as to analyze factors that contribute to the increase in empathy characteristics.