Search

According to WHO, Drowning is a very much serious and every year 3,74,000 people claiming the lives as a public health threat. Underwater human body detection through live video is a very challenging task because of movement of water currents continuously and changes the intensity of water. Detecting people from an underwater video is a very complex and challenging problem as the video can be affected by various factors such as undesired artifacts (e.g. noise), monitoring limitation of cameras, illumination variation etc. It is very important to build the appropriate model to address this problem. The major goal of this paper is to examine the possibility of detect humans in an underwater Ambient using one of the Faster R-Convolution neural network (Faster R-CNN) and data enhancement algorithms and acutely, recall and precise evaluation of the proposed model results.

Live video applications are getting popular, and content providers widely use adaptive bitrate (ABR) streaming to improve QoE while maintaining low latency. However, users' increasing preference to watch videos on mobile devices poses great challenges for ABR algorithm due to the dramatically varying cellular network. Existing learn-based ABR algorithms face difficulties to generalize to various network conditions because of their reliance on training traces, and model/rule-based ABR schemes suffer from rebuffering under low latency constraint since they cannot robustly control the buffer occupancy within a small range. To address it, this work proposes Cratus, a lightweight and robust ABR algorithm for mobile live streaming, which achieves high QoE and low latency by accurately regulating the buffer at a small level. To enhance the control ability, Cratus controls the buffer dynamic behavior rather than the buffer occupancy. By using sliding mode control approach, Cratus robustly controls the buffer dynamic and ensures that the buffer occupancy is bounded around the target level regardless of network uncertainties. Trace-driven experiments show that Cratus outperforms existing ABRs: average QoE is increased by 12.3% to 28.6%, and rebuffering time is limited within 0.8s on average, which is reduced by 53.5% to 92.3%.

Desde la existencia de los medios de comunicación masiva, esto es, prensa, radio, televisión e internet, existen líderes de opinión que influyen en la toma de decisiones de su audiencia. Con la adopción masiva de las plataformas de redes sociales, la ubicuidad que proporcionan los dispositivos móviles y los live videos, la barrera de información entre el líder de opinión y su audiencia se ha reducido significativamente (Carillo et al., 2017). Las empresas, al entender que el desplazamiento de la atención de los consumidores hacia lo digital es más permanente que transitorio, han destinado mayores recursos a la publicidad digital, particularmente, en el influencer marketing o marketing de influyentes (Insider Intelligence, 2021). Como asevera Voorveld (2019), el influencer marketing ofrece una oportunidad para llegar a diversos grupos de audiencia de manera más directa, personal e interactiva. Para entender el tema de influencer marketing, en particular el papel que juegan los influencers y líderes de opinión en la era digital, se analizará a fondo el tema en el Capítulo 1. La presente tesis doctoral buscó responder cómo las variables relativas a la relación que un seguidor experimenta con el influencer, siendo estas identificación, interacción parasocial y afinidad, anteceden la respuesta a los live videos publicados por los influencers. Además, contestar si los live videos publicados por influencers son efectivos para anunciar las marcas, esto a través de las variables engagament y dependencia hacia el video. Por último, se evaluaron las consecuencias de estas variables evaluando cómo éstas afectan el comportamiento y sentimiento hacia las marcas que anuncian los influencers a través de las variables Brand advocacy, Brand attachment e Intención de Compra. Para responder a las preguntas de investigación, se diseñó la investigación con un enfoque tipo empírico causal y se analizaron las relaciones entre las variables a través de un modelo estructural. El instrumento de medida con la medición de las variables fue un cuestionario, y se consiguió una muestra de n = 376 en Puerto Rico a través de un método de muestreo por conveniencia principalmente a través de correo electrónico y en las redes sociales Instagram y Facebook. Los resultados muestran que las variables relativas al influencer: la identificación, la interacción parasocial y la afinidad, afectan el resultado hacia el contenido publicado en un live video medido a través de las variables engagement y dependencia al medio. Esta relación con el contenido que publican en vivo los influencers tiene un efecto en las marcas que se anuncian, esos resultados siendo las variables brand advocacy, brand attachment e intención de compra. En general, los hallazgos de esta tesis doctoral presentan implicaciones teóricas y prácticas en el tema de marketing de influencers. Específicamente, los resultados aportan valor para una mejor comprensión del uso del live video en el marketing de influencers, un canal reciente en la literatura. Además, ofrece recomendaciones para la gestión de marketing en las empresas. En particular para gerentes y directores de marketing y comunicación, para utilizar los live videos en el marketing de influencers en las redes sociales Instagram y Facebook como alternativa publicitaria. Los resultados de la presente tesis sirven también como guía para los influencers, proveyendo recomendaciones para el uso de live videos en su estrategia de contenido y para manejar de manera más efectiva su relación con sus seguidores. Since the existence of the mass media, that is, the press, radio, television and the Internet, there have been opinion leaders who influence the decision-making of their audience. With the massive adoption of social networking platforms, the ubiquity provided by mobile devices and live videos, the information barrier between the opinion leader and their audience has been significantly reduced (Carillo et al., 2017). Companies, understanding that the shift in consumer attention to digital is more permanent than transitory, have allocated more resources to digital advertising, particularly in influencer marketing (Insider Intelligence, 2021). As Voorveld (2019) affirms, influencer marketing offers an opportunity to reach various audience groups in a more direct, personal and interactive way. To understand the subject of influencer marketing, in particular the role that influencers and opinion leaders play in the digital age, the subject will be analyzed in depth in chapter 1. This doctoral thesis pursued to answer how the variables related to the relationship that a follower experiences with the influencer, being these identification, parasocial interaction and affinity, precede the response to the live videos published by the influencers. Also, to answer if the live videos published by influencers are effective to advertise the brands, this through the variables engagament and dependence on the video. Finally, the consequences of these variables were analized, evaluating how they affect the behavior and sentiment towards the brands that influencers advertise through the variables Brand advocacy, Brand attachment and Purchase intention. To answer the research questions, the research was designed with a causal empirical approach and the relationships between the variables were analyzed through a structural model. The measurement instrument with the measurement of the variables was a questionnaire, and a sample of n = 376 was obtained in Puerto Rico through a convenience sampling method mainly through email and on the social networks Instagram and Facebook. The results show that the variables related to the influencer: identification, parasocial interaction and affinity, affect the result towards the content published in a live video measured through the variables engagement and dependence on the medium. This relationship with the content that influencers publish live has an effect on the brands that are advertised, these results being the variables brand advocacy, brand attachment and purchase intention. In general, the findings of this doctoral thesis present theoretical and practical implications on the subject of influencer marketing. Specifically, the results provide value for a better understanding of the use of live video in influencer marketing, a recent channel in the literature. In addition, it offers recommendations for marketing management in companies. In particular for marketing and communication managers and directors, to use live videos in influencer marketing on the social networks Instagram and Facebook as an advertising alternative. The results of this thesis also serve as a guide for influencers, providing recommendations for the use of live videos in their content strategy and to more effectively manage their relationship with their followers.

Long-distance transport of caribou (Rangifer tarandus) can result in morbidities and mortalities. This case report describes the use of a long-acting tranquilizer, zuclopenthixol acetate (ZA) and live video monitoring (LVM) to transport caribou over 2300 kilometers. Two groups of captive caribou were transported from Fort St. John, British Columbia to Dryden, Ontario (n=14; 28.5 h transport) and Anchorage, Alaska (n=11; 32 h transport). The day prior to transport, caribou were administered ZA at 1 mg/kg via deep intramuscular (IM) injection. Digital video cameras allowed for live observation of caribou during transport. Still images of videos from each compartment in the Ontario transport were analyzed for percentage (%) lying versus standing over three time periods (Day 1, Night, and Day 2). Overall, caribou spent 57% of the transport lying down, with the highest percentage occurring at night (73%). As group size and animal density decreased there was a trend for caribou to spend more time lying down. Three animals developed extrapyramidal effects to the ZA and were effectively treated with midazolam at 0.2 mg/kg IM. There were no significant visible injuries or mortalities during or up to 6 weeks post-transport. Zuclopenthixol acetate and LVM were used to successfully transport caribou over long distances and should be considered in future translocations to improve animal welfare during transport.

Although deep learning has reached a higher accuracy for video content analysis, it is not satisfied with practical application demands of porn streamer recognition in live video because of multiple parameters, complex structures of deep network model. In order to improve the recognition efficiency of porn streamer in live video, a deep network model compression method based on multimodal knowledge distillation is proposed. First, the teacher model is trained with visual-speech deep network to obtain the corresponding porn video prediction score. Second, a lightweight student model constructed with MobileNetV2 and Xception transfers the knowledge from the teacher model by using multimodal knowledge distillation strategy. Finally, porn streamer in live video is recognized by combining the lightweight student model of visualspeech network with the bullet screen text recognition network. Experimental results demonstrate that the proposed method can effectively drop the computation cost and improve the recognition speed under the proper accuracy.

Automatic Number Plate Recognition (ANPR) combines electronic hardware and complex computer vision software algorithms to recognize the characters on vehicle license plate numbers. Many researchers have proposed and implemented ANPR for various applications such as law enforcement and security, access control, border access, tracking stolen vehicles, tracking traffic violations, and parking management system. This paper discusses a live-video ANPR system using CNN developed on an Android smartphone embedded with a camera with limited resolution and limited processing power based on Malaysian license plate standards. In terms of system performance, in an ideal outdoor environment with good lighting and direct or slightly skewed camera angle, the recognition works perfectly with a computational time of 0.635 seconds. However, this performance is affected by poor lighting, extremely skewed angle of license plates, and fast vehicle movement.

Social presence is known to be important for distance education, and a common approach in online classes is to provide chat boxes and forums to provide the social presence. In such a class, however, learners must explicitly act beyond their normal learning activities, so often there is no social presence in the class even when there are several learners working on the same course material. In this paper, we develop an approach where learners can share the social presence without any explicit action; their normal learning activities would be used to provide visual cues for social presence. We present Cocode, a system designed for an online programming class that shows other learners' code editors and running output in the programming environment with minimum privacy issues. For evaluation, we ran two user studies with groups of participants who took an offline class and an online programming class from the university; results from the studies showed that learners felt less social presence in Cocode than in offline classes, but they felt significantly more social presence in Cocode than in online classes with live video lectures, forums, and chat sessions.

Videos take a lot of time to transport over the network, hence running analytics on the live video on embedded or mobile devices has become an important system driver. Considering such devices, e.g., surveillance cameras or AR/VR gadgets, are resource constrained, although there has been significant work in creating lightweight deep neural networks (DNNs) for such clients, none of these can adapt to changing runtime conditions, e.g., changes in resource availability on the device, the content characteristics, or requirements from the user. In this article, we introduce ApproxNet, a video object classification system for embedded or mobile clients. It enables novel dynamic approximation techniques to achieve desired inference latency and accuracy trade-off under changing runtime conditions. It achieves this by enabling two approximation knobs within a single DNN model rather than creating and maintaining an ensemble of models, e.g., MCDNN [MobiSys-16]. We show that ApproxNet can adapt seamlessly at runtime to these changes, provides low and stable latency for the image and video frame classification problems, and shows the improvement in accuracy and latency over ResNet [CVPR-16], MCDNN [MobiSys-16], MobileNets [Google-17], NestDNN [MobiCom-18], and MSDNet [ICLR-18].

Live video hosted by streamer is being sought after by more and more Internet users. A few streamers show inappropriate action in normal live video content for profit and popularity, who bring great harm to the network environment. In order to effectively regulate the streamer behavior in live video, a streamer action recognition method in live video with spatial-temporal attention and deep dictionary learning is proposed in this paper. First, deep features with spatial context are extracted by a spatial attention network to focus on action region of streamer after sampling video frames from live video. Then, deep features of video are fused by assigning weights with a temporal attention network to learn the frame attention from an action. Finally, deep dictionary learning is used to sparsely represent the deep features to further recognize streamer actions. Four experiments are conducted on a real-world dataset, and the competitive results demonstrate that our method can improve the accuracy and speed of streamer action recognition in live video.

Online video streaming is one of the channels of the current economic development of e-commerce. In this technology era, our lifestyle is gradually moving towards electronic development and the public use of e-commerce applications for their businesses. Online video streaming strategies, therefore, benefit marketers, and this is mainly due to their cost-effective properties. The main objective of this research is to investigate the influences of online video streaming on customers’ online purchase intention. Three independent variables include perceived credibility, perceived ease of use and perceived usefulness are used to study the influence of online video streaming on the consumers’ online purchase intention. A set of self-administered questionnaires was distributed to 215 respondents from three states in Malaysia which are Johor, Melaka and Selangor, who had the experience of online purchases that were affected by online video streaming. The collected data are then analysed using SPSS and Smart PLS software. The results of the analysis indicated that all three independent variables have a significant impact on consumers' online purchase intentions through online video streaming. In conclusion, this research is important to e-marketers in order to allow them to gain a better understanding of the online purchasing behaviour of their customers

In response to the COVID-19 pandemic, the emergency policy for online learning at home has been launched in China. Live video, prerecorded video, and their combination are the three main video lear...

In this paper, the authors develop an efficient face recognition algorithm from images or live video streaming for IoT systems based on K-nearest neighbor and support vector machine learning to recognize the person from the local database and extract the features of the face. Because of the complexity of the conditions, there might be some factors of facing errors like the size; the angle; the distance from the ear, nose, and eyes; etc. This sustainable machine learning-based IoT system is designed for sovereign face recognition with features extraction with improved accuracy near about 96%. The experimental study is done to test the performance of the face recognition in the changes of number of persons in video or images. Finally, this manuscript recognized persons from live video or images with accuracy approximately 96% by using the SVM and KNN classifiers and discussed with the block diagram and proposed algorithm.

The McCollough Effect is a color aftereffect produced by exposure to colored, oriented patterns. For example, following adaptation to vertical red and horizontal green stripes in alternation, vertical black and white patterns appear greenish, while horizontal black and white patterns appear reddish. The striking aspect of the McCollough Effect is that just a few minutes of adaptation can produce an aftereffect lasting days or weeks. Though this effect is easily induced, previous work has shown that stronger effects can be achieved with longer periods of adaptation. To allow especially long adaptation durations, the current work develops a novel method of induction of the McCollough Effect using live video feed, filtered by orientation, and viewed with a head-mounted display. Results showed that this “McCollough World” paradigm was as strong an inducer (per unit time) as traditional paradigms using gratings, while allowing observers to adapt comfortably for multiple hours. Two hours of McCollough World adaptation produced effects that were significantly larger than 20 min of traditional adaptation, which is close to the tolerance limits for gratings. This work provides insight into the features necessary for induction of the McCollough Effect and provides a strategy for creating especially strong and long-lasting color aftereffects.