Research Area:  Internet of Things
The success of Artificial Intelligence (AI) through Deep Learning (DL) and Computer Vision has inspired many researchers to work on many real-life and human-centered tasks. These current AI systems are in use to augment the intelligence of IoT. IoT devices are equipped with very low computing and fewer storage resources. In the case of visual computing, a massive number of images or video data are needed to be processed, which seems to be not feasible for an IoT device. Therefore, those data are needed to transfer to a cloud machine for computation. However, in this case, bandwidth scarcity is a huge problem. Real-time computation and security and privacy of data are also very challenging issues. To handle this problem, Mobile Edge Computing (MEC) is used in IoT to perform the real-time computation locally. Combining state-of-the-art computer vision algorithms such as DL, especially Deep Convolutional Neural Network (CNN) based algorithms and MEC, can be a smart solution for onsite visual computing. This chapter scholarly discussed how deep CNN through MEC could be a potential technique for IoT based solutions. It also discuss how a deep transfer learning procedure can be applied in this method. This chapter proposes how different layers of deep CNN can be split up among Edge devices, Fog gateway, and Cloud servers to do visual computing at IoTs. Relevant technical backgrounds, current state-of-the-art, and future scopes are also emphasized in this chapter.
Keywords:  
Author(s) Name:  Abu Sufian, Ekram Alam, Anirudha Ghosh, Farhana Sultana, Debashis De & Mianxiong Dong
Journal name:   Mobile Edge Computing
Conferrence name:  
Publisher name:  Springer
DOI:  10.1007/978-3-030-69893-5_18
Volume Information:  pp 443–471
Paper Link:   https://link.springer.com/chapter/10.1007/978-3-030-69893-5_18