Training a machine to recognize a motor vehicle driver using a mobile device

A system for identification of a driver who is using a mobile device while driving, the system comprising:

• a camera;
• a processor;
• a memory;
• wherein the processor is configured to capture a plurality of images from the camera of a driver of a motor vehicle through a side window of the motor vehicle while the driver is driving the motor vehicle in a road environment, wherein the images are storable in the memory;
• wherein a plurality of positive images are selected from the images, wherein the positive images show a driver using a mobile device while driving a motor vehicle;
• wherein a plurality of negative images are selected from the images, wherein the negative images show a driver is not using a mobile device while driving a motor vehicle;
• a machine;
• wherein a plurality of training images are selected from both the positive images and the negative images,
• wherein the machine is trained using the selected training images to recognize that a test image, captured through a window of the motor vehicle, shows a driver using a mobile device while driving the motor vehicle;
• wherein a test image is captured through a side window of the motor vehicle;
• wherein the trained machine is used to classify that a test image shows a side view of a driver using a mobile device while driving the motor vehicle;
• and
• wherein a driver is identified using a mobile device while driving the motor vehicle in a road environment.

A set of training images produced by a process comprising:

• capturing a plurality of images of a driver of a motor vehicle through a side window of the motor vehicle;
• selecting from the captured images, a plurality of positive images which show a driver using a mobile device while driving a motor vehicle;
• selecting from the captured images, a plurality of negative images which show a driver is not using a mobile device while driving a motor vehicle;
• selecting a plurality of training images from both the positive images and the negative images;
• by inputting to the machine the selected training images and respective labels indicating that the selected training images are positive images or negative images, training a machine to classify that a test image, captured through a side window of the motor vehicle, shows a driver using a mobile device while driving the motor vehicle;
• optimizing the training of the machine by adding or removing a portion of the training images to produce thereby an adjusted set of training images and repeating the training using the adjusted set of training images, wherein a goal of the optimizing is to increase above a previously determined threshold a fraction of true positive classifications.