Vehicle capsule networks
A system, comprising a computer that includes a processor and a memory, the memory storing instructions executable by the processor to detect and locate an object by processing video camera data with a capsule network, wherein training the capsule network includes determining routing coefficients with a scale-invariant normalization function. The computer can be further programmed to receive the detected and located object.
1. A method, comprising:
detecting and locating an object by processing video camera data with a capsule network that includes a plurality of capsules arranged in layers, wherein lower-level capsules include prediction tensors that include the size of an object feature vector, a number of the lower-level capsules in a lower-level layer that can be linked to one or more capsules in a higher-level layer, and the number of routing links to the one or more higher-level capsules in the higher-level layer; wherein training the capsule network includes determining routing coefficients with a scale-invariant normalization function which includes determining maximal routing coefficients using a Max-min( ) function and includes expectation-maximization to link the lower-level capsules to the one or more higher-level capsules; and
receiving the detected and located object at a computing device.
2. The method of claim 1 , wherein the capsule network includes further determining that the routing coefficients corresponding to routes between capsule layers.
3. The method of claim 1 , wherein routing coefficients are determined by grouping routes based on one or more of correlation or clustering following training based on a first training data set, wherein the grouped routes connect determined elements in a capsule layer with locations in a subsequent capsule layer.
4. The method of claim 1 , wherein training the capsule network further includes determining routing coefficients with the scale-invariant normalization function within a for-loop.
5. The method of claim 1 , further comprising operating a vehicle based on receiving the detected and located object.
6. The method of claim 5 , wherein operating the vehicle based on receiving the detected and located object includes determining a predicted location of the object in global coordinates.
7. The method of claim 1 , further comprising determining traffic data based on receiving the detected and located object.
8. The method of claim 1 , further comprising acquiring the video camera data with one or more of a stationary video camera included in a traffic infrastructure system and a mobile video camera included in one or more of a vehicle and a drone.
9. A system, comprising a processor; and
a memory, the memory including instructions to be executed by the processor to:
detect and locate an object by processing video camera data with a capsule network that includes a plurality of capsules arranged in layers, wherein lower-level capsules include prediction tensors that include the size of an object feature vector, a number of the lower-level capsules in a lower-level layer that can be linked to one or more capsules in a higher-level layer, and the number of routing links to the one or more higher-level capsules in the higher-level layer; wherein training the capsule network includes determining routing coefficients with a scale-invariant normalization function which includes determining maximal routing coefficients using a Max-min( ) function and includes expectation-maximization to link the lower-level capsules to the one or more higher-level capsules; and
receive a detected and located object at a computing device.
10. The system of claim 9 , the instructions including further instructions to determine that the routing coefficients corresponding to routes between capsule layers.
11. The system of claim 9 , the instructions including further instructions to determine routing coefficients by grouping routes based on one or more of correlation or clustering following training based on a first training data set, wherein the grouped routes connect determined elements in a capsule layer with locations in a subsequent capsule layer.
12. The system of claim 9 , wherein training the capsule network further includes determining routing coefficients with the scale-invariant normalization function within a for-loop.
13. The system of claim 9 , wherein operating a vehicle based on predicting an object location includes receiving the detected and located object.
14. The system of claim 13 , wherein operating the vehicle based on receiving the detected and located object includes determining a predicted location of the object in global coordinates.
15. The system of claim 9 , wherein determining traffic data includes receiving the detected and located object.
16. The system of claim 9 , the instructions including further instructions to acquire the video camera data with one or more of a stationary video camera included in a traffic infrastructure system and a mobile video camera included in one or more of a vehicle and a drone.
17. A system, comprising:
means for controlling vehicle steering, braking and powertrain; and
means for detecting and locating an object by processing video camera data with a capsule network that includes a plurality of capsules arranged in layers, wherein lower-level capsules include prediction tensors that include the size of an object feature vector, a number of the lower-level capsules in a lower-level layer that can be linked to one or more capsules in a higher-level layer, and the number of routing links to the one or more higher-level capsules in the higher-level layer; wherein training the capsule network includes determining routing coefficients with a scale-invariant normalization function which includes determining maximal routing coefficients using a Max-min( ) function and includes expectation-maximization to link the lower-level capsules to the one or more higher-level capsules; and
means for receiving the detected and located object at a computing device; and
operating a vehicle based on the detected and located object and the means for controlling vehicle steering, braking and powertrain.
18. The system of claim 17 , wherein the capsule network includes further determining routing coefficients corresponding to routes between capsule layers.
19. The system of claim 17 , wherein routing coefficients are determined by grouping routes based on one or more of correlation or clustering following training based on a first training data set, wherein the route connects determined elements in a capsule layer with locations in a subsequent capsule layer.
20. The system of claim 17 , wherein training the capsule network includes determining routing coefficients with the scale-invariant normalization function within a for-loop.