Poster: Accessible, Distributed Hydro-Surveillance Through Integrated End-Edge-Cloud Architecture

Flood damage is a devastating natural disaster that requires effective hydro-surveillance (HS) systems. However, the limited capabilities of edge servers (ES) make it challenging to develop such systems. Our study proposes an accessible, distributed HS and computing system to address flood damage. To increase inference efficiency on ES, we develop a HS model and combine it with a collaborative scheduling algorithm to construct an integrated end-edge-cloud (iEEC) computing pathway. Our system achieves 94.3% accuracy and a speed of 30.3 frames per second (FPS), outperforming state-of-the-art (SOTA) approaches, and can handle hard instances. The prototype system and simulation tool demonstrate the effectiveness of our approach at various scales.