CNN based Smart Surveillance System: A Smart IoT Application Post Covid-19 Era

Abstract

There are many applications available in the face detection algorithm but very limited applications are identified for further processing. When it comes to identifying faces in the crowd and that too in all-weather conditions then it's too difficult a task to be conducted. Considering this challenge, most of the surveillance systems are not automated. In the sense the CCTV deployed are used only for bufferingpurposes. Very rarely an event is brought to the notice and later CCTV footage is used as a tool for legal issues. There are a lot of concerns about deploying CCTV's in Public places as well. S electing this system, our work have made sure that the CCTV's wil l be used to process the video being taken, and whenever an event triggers, notice(s) has to be provided for officials. Thus, the proposed system is for automatic detection and recognition of human faces for finding criminals/suspects/missing persons for surveillance. The intended method first detects a face in the video using a face detection algorithm and then it will search whether the face is present in the data centre. The method provides the ability to detect, extract features, and recognize a face from inputs taken by camera or video automatically. Recognizing faces under different natural conditions can be done by training the system on a limited number of facial images. Also, this system is tested for Covid-19 Post situations, wherein wearing masks in public places is compulsory thus face recognition in these aspects is also tested, and encouraging results have been achieved.

AIoT Used for COVID-19 Pandemic Prevention and Control

“…The system uses the DPSSD face detector to perform the detection function and uses the integrated CNN to perform the positioning function. Afterward, Moorthy et al [ 39 ] propose to use face detection algorithm for remote monitoring and tracking. Therefore, it is still difficult to use face recognition system to monitor patients remotely, because of great differences between faces, such as expression, posture, skin color, and position.…”

“…They conducted experiments using the framework PAR and found out that the platform could be used flexibly and continuously for remote intensive care. Moorthy et al [ 39 ] built an AIoT-based system of intelligent devices and sensors for remote intensive care, which is able to track a large number of diseases and conditions. Although the above methods can effectively help to carry out remote monitoring, it remains to be verified whether wrong or missed diagnosis would occur if it is only based on the algorithm or face recognition system monitoring.…”

Auto-Tiler: Variable-Dimension Autoencoder with Tiling for Compressing Intermediate Feature Space of Deep Neural Networks for Internet of Things

“…Artificial Intelligence (AI) is rapidly spreading into Internet of Things (IoT) devices, including face recognition for smart security systems [1][2][3], voice assistant with AI speakers [4][5][6], and smart cars [7,8]. IoT edge devices, however, do not have sufficient resources to perform inference of complex deep neural networks (DNN) in a timely manner.…”