Method and system for training a neural network to classify objects or events

摘要

A method includes receiving a first set of sensor data including data representing an object or an event in a monitored environment, receiving a second set of sensor data representing a corresponding time period as a time period represented by the first set of sensor data, inputting to a tutor classifier data representing the first set of data and including data representing the object or the event, generating a classification of the object or event in the tutor classifier, receiving the second set of sensor data at an apprentice classifier training process, receiving the classification generated in the tutor classifier at the apprentice classifier training process, and training the apprentice classifier in the apprentice classifier training process using the second set of sensor data as input and using the classification received from the tutor classifier as a ground-truth for the classification of the second set of sensor data.