Action detection of objects devices using deep learning in IoT applications

Abstract

Internet of Things (IoT) technology is the communication and communication of smart technological devices with each other. However, with the development of the Internet of Things (IoT), the number of smart applications and interconnected devices is increasing day by day. Deep Learning (DL) method has become necessary to process the large amount of raw data collected and to further improve intelligence and application capabilities. It is seen that the majority of researchers focus on action detection. Standard Deep Learning techniques are difficult to use in IoT devices as Deep Learning applications require high CPU, RAM and storage. In this study, an action detection technique has been developed directly on the edge device by enabling the use of deep learning techniques in IoT devices. This technique, as a representation of neural networks, divides it into on-board computers. Visual action detection is one of the critical components of a smart city. High processing capacity and storage requirements severely limit comprehensive and precise monitoring within the IoT and edge computing framework. The structure proposed in this paper suggests the deployment of micro deep learning algorithms to the latest IoT and embedded devices, including the utilisation of minimal computing resources such as processor, power and memory, with a contribution to IoT and embedded device activities in action detection. The systematic analysis shows that many IoT devices can be applied to the proposed optimisation design. The proposed model is much smaller in size than existing models.