Task Distribution of Object Detection Algorithms in Fog-Computing Framework

Advancements in deep neural networks has led to the extensive implementation of machine learning models for inferencing and analytics on data especially in smart city projects. Object detection algorithm is one of well-known application of deep neural network. Given how computationally expensive these operations are, there is a growing need for methods to reduce the effort of running these complex algorithms on resource-constrained embedded devices which are typically used in IoT applications. Recently, a computing paradigm called fog computing which extends the cloud computing paradigm to the network edge has captured the attention of researchers and industrial organizations alike. This paper investigates the possibilities of implementing Fog Computing using a novel layer-wise partitioning scheme as a solution to reduce the effort of running deep inferencing for object detection algorithms on embedded IoT devices. Results show that the proposed solution is potential in comparison with cloud and single node based system.