Automotive detection of urban features from multi-scale imagery using Deep Learning

Mobility is an essential component of routine life for everyone. Walking and biking are two major transportation modes that help residents meet most or all of their daily needs. However, several mobility-related challenges occur as a result of weak maintenance of urban infrastructure such as sidewalks, bike lanes, and trails, which can decrease the safety of residents and may lead them to opt for vehicle-based transportation instead. One of the main reasons behind poor infrastructure maintenance is inefficiency of traditional inspection methods. We are proposing a Machine Learning method that uses multiple observation sources including street-level and UAV images to be employed in lieu of traditional inspection methods.

Faculty Supervisor:

Claude Duguay

Student:

Amin Gharebaghi

Partner:

GeoMate

Discipline:

Environmental sciences

Sector:

Professional, scientific and technical services

University:

University of Waterloo