Building Energy Efficient Semantic Segmentation in Intelligent Edge Computing

Semantic segmentation is a critical area in computer vision, which needs voluminous image data streaming from user devices. Usually, it is challenging to process semantic segmentation tasks in user devices due to the Limited computation power and battery life. Intelligent edge computing effectively enhances the accuracy of semantic segmentation tasks by offloading computations to nearby devices, providing lower latency and improved responsiveness. However, inefficient offloading brings additional energy consumption due to the irregular relationship between task requirements and offloading settings. In this paper, we attempt to improve energy efficiency for processing semantic segmentation tasks in the edge environment by leveraging energy consumption and task requirements. We first investigate the power consumption with different offloading settings in a real intelligent edge environment. Based on the investigation, we formulate the offloading setting as a restricted multi-armed bandit problem and solve it by enhancing the upper confidence bound algorithm. Comprehensive simulation results show that the proposed solution significantly improves the energy efficiency for offloading semantic segmentation tasks in a given intelligent edge environment.