Automatic event annotation

1 . A fleet management server comprising:

a memory;

a neural network stored in the memory, the neural network trained on both pre-post event (PPE) data collected from a plurality of vehicle sensors for a plurality of past vehicle events and from image data corresponding to the PPE data for each of the plurality of past vehicle events; and

a processor, coupled to the memory, wherein the processor includes a program stored on computer readable medium configured to:

receive new vehicle event data, including new pre-post event (PPE) data collected from a plurality of vehicle sensors, representing a new vehicle event that occurred when a vehicle condition occurred while a driver was driving a vehicle;

execute the neural network using the new PPE data as inputs to the neural network to automatically predict and assign at least one annotative classification, selected among an array of pre-defined annotative classifications associated with a vehicle event, to the new vehicle event based on the new PPE data, and

output, to a user display, a representation of the new event data based on the automatically assigned annotative classification for the new vehicle event to reduce the user's time spent reviewing the new event data.

2 . The fleet management server as set forth in claim 1 , further including a further processor configured to:

receive past vehicle event data, including both the past image data and the past PPE data, associated with at least one vehicle and at least one driver, the past vehicle event data representing at least one respective past vehicle event that occurred while one of the drivers was driving one of the vehicles; and

train the neural network using the past vehicle event data.

3 . The fleet management server as set forth in claim 2 , wherein the further processor is further configured to:

receive an annotative classification for a past vehicle event that was previously manually assigned based on the respective past vehicle event data of one or more of the past vehicle events.

4 . The fleet management server as set forth in claim 3 , wherein the processor is further configured to:

execute the neural network to automatically predict and assign the annotative classification to the new vehicle event using the new PPE data as a proxy for new image data of the new vehicle event data and without using the actual new image data of the new vehicle event.

5 . The fleet management server as set forth in claim 4 , wherein the processor is further configured to:

output, to the user, a ranking for the new vehicle event based on the automatically assigned annotative classification.

6 . The fleet management server as set forth in claim 5 , wherein the processor is further configured to:

output the representation, to the user, in an order based on the ranking of the automatically assigned annotative classification.

7 . The fleet management server as set forth in claim 2 , wherein the processor is further configured to:

receive a video data recording from the forward-facing video camera when the vehicle condition occurred;

analyze the video data recording to determine the expected driver behavior during the event; and

output, to a user display, a representation of the expected driver behavior.

8 . The fleet management server as set forth in claim 1 , wherein the processor is further configured to:

execute the neural network to automatically predict and assign the annotative classification to the new vehicle event using the new PPE data as a proxy for new image data of the new vehicle event data and without using the actual new image data of the new vehicle event.

9 . The fleet management server as set forth in claim 1 , wherein the processor is further configured to:

output, to the user, a ranking for the new vehicle event based on the automatically assigned annotative classification.

10 . The fleet management server as set forth in claim 1 , wherein the processor is further configured to:

output the representation, to the user, in an order based on the ranking of the automatically assigned annotative classification.

11 . The fleet management server as set forth in claim 1 , wherein:

the memory and the processor are located remote from the fleet management server.

12 . The fleet management server as set forth in claim 1 , wherein:

the memory and the processor are located on the vehicle from which the new vehicle event data was received.

13 . The fleet management server as set forth in claim 1 , wherein the new PPE data includes:

at least one of a speed of the vehicle, a steering angle of the vehicle, a braking force of the vehicle and a distance to a forward vehicle.

14 . The fleet management server as set forth in claim 1 , wherein:

the PPE data extends from 5 seconds before the new vehicle event to 5 seconds after the new vehicle event.

15 . The fleet management server as set forth in claim 1 , wherein:

the representation of the new event data is color-coded based on a severity of the new event.

16 . A method of automatically predicting and assigning an annotative classification to a new vehicle event, the method comprising:

using a processor to train a neural network stored in a computer memory on both pre-post event (PPE) data collected from a plurality of vehicle sensors for a plurality of past vehicle events and from image data corresponding to the PPE data for each of the plurality of past vehicle events;

receiving at a processor, new event data, including new pre-post event (PPE) data collected from a plurality of vehicle sensors, representing a new vehicle event that occurred when a vehicle condition occurred while a driver was driving a vehicle;

executing, on a processor, the neural network to automatically predict and assign an annotative classification, selected among an array of annotative classifications associated with a vehicle event, to the new vehicle event; and

outputting, to a user display, a representation of the new event data based on the automatically assigned annotative classification for the new vehicle event to reduce a user's time spent reviewing the new event data.

17 . The method of automatically predicting and assigning an annotative classification to a new vehicle event as set forth in claim 16 , further including:

receiving at the processor the trains the neural network, past vehicle event data, including both the past image data and the past PPE data, associated with at least one vehicle and at least one driver, the past event data representing at least one respective past vehicle event that occurred when the vehicle condition occurred while one of the drivers was driving one of the vehicles.

18 . The method of automatically predicting and assigning an annotative classification to a new vehicle event as set forth in claim 17 , further including:

receiving at the processor that trains the neural network, an annotative classification that was previously manually assigned based on the past image data of the respective past vehicle event data of the past vehicle events.

19 . The method of automatically predicting and assigning an annotative classification to a new vehicle event as set forth in claim 18 , further including:

executing the neural network on a processor to automatically predict and assign the annotative classification to the new vehicle event using the new PPE data as a proxy for new image data of the new vehicle event data and without using the actual new image data of the new vehicle event.

20 . The method of automatically predicting and assigning an annotative classification to a new vehicle event as set forth in claim 19 , further including:

ranking, via a processor, the new vehicle event based on the automatically assigned annotative classification; and

outputting, from the processor to the user, the ranking of the new vehicle event.

21 . The method of automatically predicting and assigning an annotative classification to a new vehicle event as set forth in claim 20 , further including:

outputting the representation, to the user, in an order based on the ranking of the automatically assigned annotative classification.

22 . The method of automatically predicting and assigning an annotative classification to a new vehicle event as set forth in claim 16 , further including:

executing the neural network to automatically predict and assign the annotative classification to the new vehicle event using the new PPE data as a proxy for new image data of the new vehicle event data and without using the actual new image data of the new vehicle event.

23 . The method of automatically predicting and assigning an annotative classification to a new vehicle event as set forth in claim 16 , further including:

outputting, to the user, a ranking for the new vehicle event based on the automatically assigned annotative classification; and

outputting the representation, to the user, in an order based on the ranking.

24 . The method of automatically predicting and assigning an annotative classification to a new vehicle event as set forth in claim 16 , further including:

identifying in the PPE data, using a processor, at least one of a speed of the vehicle, a steering angle of the vehicle, a braking force of the vehicle and a distance to a forward vehicle.

25 . The method of automatically predicting and assigning an annotative classification to a new vehicle event as set forth in claim 16 , further including:

obtaining the PPE data from 5 seconds before the new vehicle event to 5 seconds after the new vehicle event.

26 . The method of automatically predicting and assigning an annotative classification to a new vehicle event as set forth in claim 16 , further including:

representing the new event data as color-coded based on a severity of the new event.

27 . A system of automatically predicting and assigning an annotative classification to a new vehicle event, the system comprising:

a memory on a vehicle; and

a processor, on the vehicle, coupled to the memory, wherein the processor is configured to receive new vehicle event data, including new pre-post event (PPE) data collected from a plurality of vehicle sensors, representing a new vehicle event that occurred when a vehicle condition occurred while a driver was driving a vehicle;

a server including a server memory and receiving the PPE data from the processor, wherein the server is configured to execute a neural network stored in the server memory using the new vehicle event data to automatically predict and assign an annotative classification associated with a vehicle event to the new vehicle event, wherein the neural network is trained on both pre-post event (PPE) data collected from a plurality of vehicle sensors for a plurality of past vehicle events and from image data corresponding to the PPE data for each of the plurality of past vehicle events; and

an output display, communicatively coupled to the server, that outputs a representation of the new event data, to a user, based on the automatically assigned annotative classification for the new vehicle event to reduce a user's time spent reviewing the new event data.

28 . The system as set forth in claim 27 , wherein the server is remote from the vehicle.

29 . The system as set forth in claim 28 , further including:

a transmitter, on the vehicle, that communicates with the processor and transmits the new PPE data; and

a receiver, on the server, that receives the new PPE data from the transmitter.

30 . A system of automatically predicting and assigning an annotative classification to a new vehicle event, the system comprising:

a memory on a vehicle;

a processor, on the vehicle, coupled to the memory, wherein the processor is configured to receive new vehicle event data, including new pre-post event (PPE) data collected from a plurality of vehicle sensors, representing a new vehicle event that occurred when a vehicle condition occurred while a driver was driving a vehicle; and

means for automatically predicting and assigning an annotative classification associated with a vehicle event to the new vehicle event based on a previous manually assigned annotative classification, the means including a neural network stored in a computer memory trained on both pre-post event (PPE) data collected from a plurality of vehicle sensors for a plurality of past vehicle events and from image data corresponding to the PPE data for each of the plurality of past vehicle events; and

an output display that outputs a representation of the new event data, to a user, based on the automatically assigned annotative classification for the new vehicle event to reduce a user's time spent reviewing the new event data.