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.