Internet of Things (IoT) has rapidly developed in multidisciplinary research topics, particularly in Cyber-Physical infrastructures, such as smart-health care, transportation systems, vehicle management surveillance systems. The smart-video surveillance system has become an essential part of almost all security applications, including academic institutions. University campuses have rich video repositories comprising almost all kinds of academic and non-academic activities. Researchers have introduced many state-of-art activity recognition methods for various application domains with the availability of several activity data sets. Unfortunately, none of these data sets or methods have been developed explicitly for academia and do not cover academic activities. With the advancement of deep learning and IoT, the processing of large-scale video data has become convenient for performing various video analysis tasks. Thus, in this work, an automated deep learning-based academic activities recognition system is presented in smart-cyber infrastructure. We explore a new academic campus domain for research and proposed a novel Convolutional Neural Network (CNN) model for academic activities recognition utilizing a realistic campus dataset. The video database typically contains long, 24-hour video streams recorded by surveillance cameras installed in campus environments. The proposed model's efficiency is tested through extensive experimentation in terms of accuracy, computation time, and memory requirement. The experimental results reveal that the proposed method attains good results with an accuracy of 98%. INDEX TERMSCyber-Physical System, Academic Activity recognition , Campus Data set, CNN, Deep Learning Model I. INTRODUCTION N OWADAYS, Internet of Things (IoT) enables the Cyber-Physical components to interact with other devices and to communicate with safety-critical systems. With the rapid spread of smart portable devices, sensors, and smartphones, the IoT-enabled technology is emerging infrastructure from a traditional operation and supporting custom designs to more effective, smart, sustainable, and resilient systems [1] & [2]. The successful adaptation of IoT-enabled technology in Cyber-Physical Infrastructure Systems enables them to be more proactive and faster, have a lower expense, offer enhanced industry practices and improve sustainability. Advanced IoT-enabled infrastructures might produce timely information communication and effective decision-making for intelligent society and industry applications. IoT smart applications are practiced without human interference in different domains such as traffic congestion monitoring, healthcare, household waste recycling, water consumption monitoring, security and surveillance systems in cities.Smart video surveillance has become an obligatory component of every security application and is one of the primary sources of generating video data. The government and private organizations use video surveillance for various purposes like monitoring industrial assets, infrastru...

Like other business domains, digital monitoring has now become an integral part of almost every academic institution. These surveillance systems cover all the routine activities happening on the campus while producing a massive volume of video data. Selection and searching the desired video segment in such a vast video repository is highly time-consuming. Effective video summarization methods are thus needed for fast navigation and retrieval of video content. This paper introduces a keyframe extraction method to summarize academic activities to produce a short representation of the target video while preserving all the essential activities present in the original video. First, we perform fine-grain activity recognition using a realistic Campus Activities Dataset (CAD) by modeling activity attention scores using a deep CNN model. In the second phase, we use the generated attention scores for each activity category to extract significant video frames. Finally, we evaluate the inter-frame similarity index used to reduce the number of redundant frames and extract only the representative keyframes. The proposed framework is tested on different videos, and the experimental results show the performance of the proposed summarization process.

Mental health is an important part of a successful life for a person whether elderly, children, or young. Alzheimer’s is a fatal brain disease that severely damages the human brain, especially in the elderly. One way to prevent Alzheimer's disease is by detecting it early. The proposed research employs a deep learning methodology using a 3D convolutional neural network (3D CNN) that has been proposed to detect Alzheimer's disease at an early stage. The proposed model is primarily evaluated using three-dimensional brain images. A series of preprocessing have been applied that is an advanced normalization tool (ANT). The underlying pattern has a size of 128×128×64 and is passed to 17 layers of a neural network that is 3D-CNN. Another contribution of this study is the conversion of a 3D Alzheimer’s image into a 2D image. A 2D convolutional neural network such that RestNET50 and VGG16 are proposed to be used for Alzheimer’s detection. The proposed model has attained the highest of 78.07% accuracy using 3D CNN.