Driver behavior monitoring

1. A method comprising:

receiving, by at least one processor of a computing device, an image captured by a camera mounted on or in a vehicle;

detecting, by the at least one processor, an object in an image from the camera using a neural network trained to output a bounding box surrounding the object in the image and a class identification data for the object;

determining, by the at least one processor, whether the object applies to the vehicle;

determining, by the at least one processor, an event based on movement of the vehicle in the presence of the object when the object applies to the vehicle; and

determining, by the at least one processor, an extent to which a driver of the vehicle performs a particular action associated with the event.

2. The method according to claim 1 , wherein classifying comprises determining a probability that the object belongs to a class.

3. The method according to claim 1 , wherein the output comprises a distance of the object from a specified point on the vehicle.

4. The method according to claim 3 , wherein the specified point is a front bumper of the vehicle.

5. The method according to claim 1 , wherein the class identification data comprises pixel values corresponding to an inside of the bounding box.

6. The method according to claim 1 , wherein the image comprises a cropped image for the detected object.

7. The method according to claim 1 , further comprising increasing, by the at least one processor, confidence associated with detecting the object in a single frame upon detecting the object in a plurality of frames.

8. The method according to claim 1 , further comprising decreasing, by the at least one processor, confidence associated with detecting the object in a single frame when the object is not detected in a consecutive frame.

9. The method according to claim 1 , wherein determining the extent comprises determining a responsibility of the driver using a machine learning model that outputs a fault of the driver.

10. The method according to claim 9 , further comprising inputting, by the at least one processor, a plurality of feature vectors for a plurality of frames into the machine learning model.

11. The method according to claim 10 , wherein the plurality of feature vectors comprises at least one of a vehicle speed, a lane position, a relative position of another vehicle, or a relative position to a pedestrian.

12. The method according to claim 9 , wherein the machine learning model is a deep neural network or a support vector machine.

13. The method according to claim 1 , wherein the event represents lane driving, tailgating, or obeying a traffic sign.

14. The method according to claim 1 , wherein the neural network is a fast region-based convolutional network.

15. The method according to claim 1 , wherein the neural network generates the bounding box.

16. A computer program product comprising a non-transitory computer-readable medium having instructions stored thereon, the instructions being executable by one or more processors configured to:

receive an image captured by a camera mounted on or in a vehicle;

detect an object in an image from the camera using a neural network trained to output a bounding box surrounding the object in the image and a class identification for the object;

determine whether the object applies to the vehicle;

determine an event based on movement of the vehicle in the presence of the object when the object applies to the vehicle; and

determine an extent to which a driver of the vehicle performs a particular action associated with the event.

17. The computer-program product according to claim 16 , wherein the one or more processors are configured to determine a probability that the object belongs to a class.

18. The computer-program product according to claim 16 , wherein the output comprises a distance of the object from a specified point on the vehicle.

19. The computer-program product according to claim 18 , wherein the specified point is a front bumper of the vehicle.

20. The computer-program product according to claim 16 , wherein the class identification data comprises pixel values corresponding to an inside of the bounding box.

21. The computer-program product according to claim 16 , wherein the image comprises a cropped image for the detected object.

22. The computer-program product according to claim 16 , wherein the one or more processors are configured to increase confidence associated with detecting the object in a single frame upon detecting the object in a plurality of frames.

23. The computer-program product according to claim 16 , wherein the one or more processors are configured to decrease confidence associated with detecting the object in a single frame when the object is not detected in a consecutive frame.

24. The computer-program product according to claim 16 , wherein the one or more processors are configured to determine a responsibility of the driver using a machine learning model that outputs a fault of the driver.

25. The computer-program product according to claim 24 , wherein the one or more processors are configured to input a plurality of feature vectors for a plurality of frames into the machine learning model.

26. The computer-program product according to claim 25 , wherein the plurality of feature vectors comprises at least one of a vehicle speed, a lane position, a relative position of another vehicle, or a relative position to a pedestrian.

27. The computer-program product according to claim 24 , wherein the machine learning model is a deep neural network or a support vector machine.

28. The computer-program product according to claim 16 , wherein the event represents lane driving, tailgating, or obeying a traffic sign.

29. The computer-program product according to claim 16 , wherein the neural network is a fast region-based convolutional network.

30. The computer-program product according to claim 16 , wherein the neural network generates the bounding box.