Inferring transportation modes from user-generated street view videos

This work investigates the potential of analysing user-generated street view imagery from videos to extract transportation modes, for example whether individuals are walking or cycling, using deep neural-networks. It addresses existing data gaps with respect to monitoring the sustainability of the mix of transport modes in cities, and complements existing methods based on stationary count stations or Google Street View. Preliminary results show the potential of the proposed workflow to automatically extract forms of transportation from videos as alternative data form.