student profile: Miss Julia Ju


Thesis work

Thesis title: Seamless Handover for Cooperative Inference Between Devices and Edge Environment

Supervisors: Dong YUAN , Wei BAO

Thesis abstract:

The modern development of Deep Neural Network (DNN) makes the thriving of Artificial Intelligence (AI). Specifically, it has shown a great potential to provide a more intelligent approach in video processing, such as self-driving cars. However, before we can achieve this technology, it is obvious that artificial intelligence technology remains a huge concern as the potential safety hazards are still very likely to happen. The challenge remains how to reduce latency without sacrificing accuracy of the inference of video processing. This is crucially important in the fields where neither the latency nor the accuracy cannot be sacrificed. Previous literatures suggested that by properly partitioning the DNN model, it is possible to accelerate the inference process without scarifying accuracy. This is because of the processing time difference in different environments such as cloud server, edge server and mobile devices. However, to the best of my knowledge, previous papers have yet to considered how to continue video processing when out of services happens, such as handover and disconnection. This is of crucial importance, because not only the latency time is usually not negligible, but also the failure of service caused by it might be damaging. Motivated by the above observation, the project is aimed to find a seamless handover approach in video processing that can save as many frames as possible during the handover period.

Note: This profile is for a student at the University of Sydney. Views presented here are not necessarily those of the University.