… Although the object detection and recognition have been well studied in computer vision …
Edge servers are usually less powerful than cloud servers. A single edge server may be …

… practicability in edge vision tasks. Our framework can help with everyday vision tasks like OD…
or embedded systems, without relying heavily on cloud computing or external servers. Such …

… edge computing, we can offload MAR tasks to nearby servers. Prior studies focus on either
single-task MAR applications offloading or dependent tasks … tasks, over a generic cloud-edge …

… [27] allocates computer vision tasks on edge servers by modeling and then scheduling
directed acyclic graphs. Unlike us, they do not focus on traffic surveillance, but on tasks given by …

… ], mobile devices offload tasks to an edge server. At the same time, an overloaded edge
server can offload tasks to other edge servers / edge clouds. The authors devised an edge-first …

… to connect edge computing architecture to the cloud server. … been at the forefront in investments
in edge technology [67]: • … to perform real-time computer vision tasks (eg, video, image …

… edge computing systems design is to efficiently integrate the functional modules, including
interactions with edge servers … It is designed for computer vision tasks on video streams. The …

… with all vision DNNs and thus our approach could be applied to other computer vision tasks.
… would yield substantial performance gain on these tasks as compared to existing baselines. …

… an edge-based detection stage. In the first stage, we establish connections between edge
devices and cloud servers to … to complete low-level vision tasks. Following the general steps to …

… task interference exists in concurrent DNN tasks, further degrading the timeliness of urgent
tasks and throughput of non-urgent tasks… for DNN inference task scheduling on edge servers. …