Recognition of Traffic Congestion State using a Deep Convolutional Network

Abstract	This paper describes a two-phase deep learned neural network for the recognition of traffic congestion in surveillance camera imagery. The network is transfer-learned using GoogLeNet for image processing and a bespoke subnet for congestion recognition.
Authors	Chris Willis (BAE) Dan Harborne (Cardiff) Richard Tomsett (IBM UK) Moustafa Alzantot (UCLA)
Date	Sep-2017
Venue	1st Annual Fall Meeting of the DAIS ITA, 2017