SEMANTIC SEGMENTATION AND CONCURRENT TASK OFFLOADINGMETHOD OF DETECTING REAL TIME OBJECTS IN EDGE CLOUD COMPUTING

Abstract

Edge computing has the power of doing computation and storage neared to the point of data
production. To lowering the latency and improv the real-time processing capabilities, edge
computing is used. A cloud-based system refers to technology that allows computing
resources to be delivered through the internet. It eliminates the need that regional
infrastructure by providing on-demand availability of storage, computing power, and
applications. The detection of objects is an automated vision approach for detecting and
localizing items in pictures or movies. It entails identifying and categorizing many objects,
as well as supplying box boundaries and labels. The main issue for the object detection is
the processing speed and the storage needed for training. It cannot be achieved by
centralized cloud server alone to enhance the performance of the system the edge server is
integrated with cloud server. So, the concurrent and task off-loading algorithm is used for
this purpose and the object is detected using semantic segmentation algorithm. The
concurrent and task offloading algorithm makes every client request as a separate process
and all process run simultaneously. The semantic segmentation algorithm is used in both
the server for object detection which follows the process of Input Processing, Network
Forward Pass, Object Detection, Non- Maximum Suppression and finally the object is
interpreted. It is concluded from the analysis the response time of the concurrent edge server
is directly proportional to the concurrent cloud. The precision and recall value of edge
server is 0.5% better than the cloud server. shows whereas the accuracy of the object
detected in cloud computing is 4% higher than that of the edge computing detection.