Title:	DNNs for filtering batch videos on the edge
DNr:	Berzelius-2022-179
Project Type:	LiU Berzelius
Principal Investigator:	Ahmed Ali-Eldin Hassan <ahmed.hassan@chalmers.se>
Affiliation:	Chalmers tekniska högskola
Duration:	2022-09-07 – 2023-04-01
Classification:	10206
Keywords:

Abstract

Massive video analysis is becoming one of the main use-cases of edge computing. However, with the ever increasing number of streams, and the ever increasing complexity of DNN analytics, specialized models are needed to filter these video batches to decide which video frames should be processed locally, which ones need to be offloaded, and which ones should be processed on hefty GPUs. This project aims to train and test models for massive video filteration. This allocation will enable us to finalize the two papers we are working on.