Measuring confidence in deep neural networks

1. A system comprising a computer including a processor and a memory, the memory including instructions such that the processor is programmed to:

receive a first prediction generated by a deep neural network based on sensor data;

disable at least one neuron of a plurality of neurons of the deep neural network after the first prediction;

then provide the sensor data to the deep neural network and receive a second prediction generated by the deep neural network based on the sensor data;

calculate a distribution of a plurality of predictions that include the first prediction and the second prediction; and

determine at least one of a measurement or a classification corresponding to an object based on the distribution,

wherein the deep neural network generates each prediction of the plurality of predictions with a different number of neurons determined by randomly disabling some of the plurality of neurons.

2. The system of claim 1 , wherein the processor is further programmed to:

compare a standard deviation of the distribution with a predetermined variation threshold; and

transmit, to a server, the sensor data when the standard deviation is greater than the predetermined variation threshold.

3. The system of claim 2 , wherein the process is further programmed to:

disable an autonomous vehicle mode of a vehicle when the distribution is greater than the predetermined distribution variation threshold, wherein the vehicle includes the computer.

4. The system of claim 1 , wherein the processor is further programmed to:

receive the sensor data from a vehicle sensor of a vehicle; and

provide the sensor data to the deep neural network.

5. The system of claim 1 , wherein the deep neural network comprises at least one of a convolutional neural network, a recurrent neural network, or a combination of the convolutional neural network and the recurrent neural network.

6. The system of claim 1 , wherein the processor is further programmed to:

provide an image captured by an image sensor of a vehicle to the deep neural network;

receive a first prediction generated by the deep neural network based on the image;

disable at least one neuron of plurality of neurons within the deep neural network after the first prediction;

provide the image to the deep neural network;

receive a second prediction generated by the deep neural network;

calculate an average based on at least the first prediction and the second prediction; and

determine the at least one of the measurement or the classification based on the average.

7. The system of claim 1 , wherein the object comprises at least a portion of a trailer connected to a vehicle and the measurement comprises a trailer angle.

8. The system of claim 1 , wherein the processor is further programmed to:

actuate a vehicle based on the at least one of the measurement or the classification.

9. A system comprising:

a server; and

a vehicle including a vehicle system, the vehicle system comprising a computer including a processor and a memory, the memory including instructions such that the processor is programmed to:

receive a first prediction generated by a deep neural network based on sensor data;

disable at least one neuron of a plurality of neurons of the deep neural network after the first prediction;

then provide the sensor data to the deep neural network and receive a second prediction generated by the deep neural network;

calculate, at the vehicle, a distribution of a plurality of predictions that include the first prediction and the second prediction; and

determine at least one of a measurement or a classification corresponding to an object detected by the sensor data,

wherein the deep neural network generates each prediction of the plurality of predictions with a different number of neurons determined by randomly disabling some of the plurality of neurons.

10. The system of claim 9 , wherein the processor is further programmed to:

compare a standard deviation of the distribution with a predetermined variation threshold; and

transmit, to the server, the sensor data when the standard deviation is greater than the predetermined variation threshold.

11. The system of claim 10 , wherein the process is further programmed to:

disable an autonomous vehicle mode of the vehicle when the distribution is greater than the predetermined distribution variation threshold.

12. The system of claim 9 , wherein the processor is further programmed to:

receive the sensor data from a vehicle sensor of the vehicle; and

provide the sensor data to the deep neural network.

13. The system of claim 9 , wherein the deep neural network comprises at least one of a convolutional neural network, a recurrent neural network, or a combination of the convolutional neural network and the recurrent neural network.

14. The system of claim 9 , wherein the processor is further programmed to:

provide an image captured by an image sensor of a vehicle to the deep neural network;

receive a first prediction generated by the deep neural network based on the image;

disable at least one neuron of plurality of neurons within the deep neural network after the first prediction;

provide the image to the deep neural network;

receive a second prediction generated by the deep neural network;

calculate an average based on at least the first prediction and the second prediction; and

determine the at least one of the measurement or the classification based on the average.

15. The system of claim 9 , wherein the object comprises at least a portion of a trailer connected to the vehicle and the measurement comprises a trailer angle.

16. The system of claim 9 , wherein the processor is further programmed to:

actuate the vehicle based on the at least one of the measurement or the classification.

17. A method, comprising:

receiving a first prediction generated by a deep neural network based on sensor data;

disabling at least one neuron of a plurality of neurons of the deep neural network after the first prediction;

then providing the sensor data to the deep neural network and receive a second prediction generated by the deep neural network;

calculating a distribution of a plurality of predictions that include the first prediction and the second prediction; and

determining at least one of a measurement or a classification corresponding to an object based on the distribution,

wherein the deep neural network generates each prediction of the plurality of predictions with a different number of neurons determined by randomly disabling some of the plurality of neurons.