Using deep appearance to extract distinctive features from road users in traffic video

The partner organization, Brisk Synergies, produces software tools for use by Urban Planners in doing traffic monitoring and analysis. Their software detects traffic objects in video, such as cars, trucks, buses, motorcycles, pedestrians and bicyclists. This project aims to enhance Brisk Synergies software by creating deep learning algorithms that can detect and track traffic objects. The project seeks to identify suitable object appearance features, such as object colour and texture, generated by the deep neural networks that will increase the object tracking performance. The appearance features will be combined with motion features (e.g. speed of moving cars and pedestrians) to help in tracking of objects when they are briefly occluded by other objects.

Faculty Supervisor:

James Clark

Student:

Partner:

Transoft Solutions Tech Corp

Discipline:

Engineering

Sector:

Professional, scientific and technical services

University:

McGill University

Program: