Workload Offloading in Object Detection using Edge Optimization and Deep Learning Model

Read full text Summarise with VIRAT
Authors: Shabariram Chokkalingam Palaniappan, P. Priya Ponnuswamy, Bhuvana Shanmugam, Sivaganesan Dhandapani
Cited by: 0
References: 13

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In modern times, the development of the Industrial Internet of Things allows edge-enabled applications.Workload offloading plays a significant role in enhancing the edge computing environment.The offloading process transferring the computationally intensive tasks to the external server.The strategy optimizes the processing time.The study proposes a solution for object detection using edge optimization and deep learning models.The object detection algorithm using custom lightweight YOLO-based model is deployed on an edge-cloud system.The process involves shifting the object detection workload to a cloud server based on the model's accuracy and memory constraints in the edge device.The system efficiency of the object detection system is enhanced using dynamic workload offloading.The optimal workload offloading scheme considers the computation capabilities of edge devices.Based on the device condition, the object detection task is executed in the local server or offloaded to the cloud server.The mathematical model is defined to find the optimal offloading decisions.The memory management in the edge devices is monitored using the low memory management tool.TensorFlow based deep learning model is deployed on cloud server.The edge optimized version of the lightweight deep learning model using TensorFlow Lite is hosted on edge devices.The results indicate that edge-cloud object detection is an effective approach with the involvement of workload offloading.The integrated model achieves precision of 90% for all the object classes considered.The dynamic quantization provides system accuracy as 98.15% for edge optimized model and 99.19% for deep learning model.

Fields of Study & Subjects

IoT and Edge/Fog Computing Advanced Neural Network Applications Internet of Things and AI

Publication details

Journal / source
Journal of Engineering Science and Technology Review
Volume
19
Issue
1
Pages
184–191
Publication date
1 February 2026
Language
EN