Bike Lane Imagery analysis

Data collection includes images of bike lanes as taken by the OSC app. The next key objective towards a unified quality metric is to evaluate the images to measure bike lane conditions. Two approaches are implemented . The first approach uses Microsoft's Custom Vision \cite{microsoft_azure_custom_2017}, to classify bike lane imagery.  The second approach uses  a custom built algorithm that uses open source computer vision and image processing functions.
Microsoft Computer Vision is an online platform that allows users to label custom imagery as a training data set and then train a custom classifier on this data. A training process then reveals a prediction URL that can be used on test data to reveal probabilities of which labels best describe the test data.