Pedestrian countdown signal classification to increase pedestrian behavior legibility
Example embodiments relate to pedestrian countdown signal classification to increase pedestrian behavior legibility. An example embodiment includes a method that includes obtaining, by a computing system of a vehicle, a camera image patch. The method further includes determining, by the computing system, using the camera image patch and a pedestrian countdown signal classifier model, a state of a pedestrian countdown signal. The method also includes determining, by the computing system based on the state of the pedestrian countdown signal, a prediction of whether a pedestrian will enter a crosswalk governed by the pedestrian countdown signal. And the method includes, based on the prediction, causing, by the computing system, the vehicle to perform an invitation action that invites a pedestrian to enter the crosswalk.
1. A computer-implemented method comprising:
obtaining, by a computing system of a vehicle, a camera image patch;
determining, by the computing system, using the camera image patch and a pedestrian countdown signal classifier model, a state of a pedestrian countdown signal;
determining, by the computing system based on the state of the pedestrian countdown signal, a prediction of whether a pedestrian will enter a crosswalk governed by the pedestrian countdown signal; and
based on the prediction, causing, by the computing system, the vehicle to perform an invitation action that invites a pedestrian to enter the crosswalk.
2. The computer-implemented method of claim 1 , wherein the invitation action comprises turning toward the crosswalk and then stopping before entering the crosswalk.
3. The computer-implemented method of claim 1 , wherein the invitation action comprises illuminating a light source.
4. The computer-implemented method of claim 1 , wherein the invitation action comprises displaying a message.
5. The computer-implemented method of claim 1 , wherein the invitation action comprises outputting a sound.
6. The computer-implemented method of claim 1 , further comprising detecting, by the computing system, a transition of the pedestrian countdown signal from: i) a do not walk state to ii) a walk state or a countdown state,
wherein the prediction is further based on the detecting of the transition.
7. The computer-implemented method of claim 1 , further comprising determining a length of the crosswalk,
wherein the prediction is further based on the length of the crosswalk.
8. The computer-implemented method of claim 1 , wherein:
the state of the pedestrian countdown signal is a walk state or a countdown state,
the method further comprises determining that the state of the pedestrian countdown signal is governed by a pedestrian control, and
the prediction is further based on the determining that the state of the pedestrian countdown signal is governed by the pedestrian control.
9. The computer-implemented method of claim 8 , wherein the determining the prediction comprises determining that an occluded pedestrian's intent is to enter the crosswalk based on the state of the pedestrian countdown signal and the determining that the pedestrian countdown signal is governed by the pedestrian control.
10. The computer-implemented method of claim 1 , further comprising:
obtaining an audio signal; and
determining that the audio signal is indicative of the state of the pedestrian countdown signal being a walk state or a countdown state,
wherein the prediction is further based on the determining that the audio signal is indicative of the state of the pedestrian countdown signal being the walk state or the countdown state.
11. The computer-implemented method of claim 1 , wherein the pedestrian countdown signal classifier model is a classification convolutional neural network that is configured to classify the state of the pedestrian countdown signal as being one of multiple states.
12. The computer-implemented method of claim 1 , wherein obtaining the camera image patch comprises:
obtaining an image using a camera coupled to the vehicle; and
using a mapped position of the pedestrian countdown signal to select from within the image a camera image patch in which the pedestrian countdown signal is visible.
13. The computer-implemented method of claim 12 , further comprising:
determining that an orientation of the pedestrian countdown signal with respect to an orientation of the vehicle satisfies a threshold condition; and
obtaining the image based on the determining that the orientation of the pedestrian countdown signal with respect to the orientation of the vehicle satisfies the threshold condition.
14. A vehicle configured to be operated in an autonomous mode, the vehicle comprising:
a memory;
a processor coupled to the memory; and
instructions stored in the memory and executable by the process to perform functions comprising:
obtaining a camera image patch;
determining, using the camera image patch and a pedestrian countdown signal classifier model, a state of a pedestrian countdown signal;
determining, based on the state of the pedestrian countdown signal, a prediction of whether a pedestrian will enter a crosswalk governed by the pedestrian countdown signal; and
based on the prediction, causing the vehicle to perform an invitation action that invites a pedestrian to enter the crosswalk.
15. The vehicle of claim 14 , wherein the invitation action comprises turning toward the crosswalk and then stopping before entering the crosswalk.
16. The vehicle of claim 14 , wherein the invitation action comprises illuminating a light source.
17. The vehicle of claim 14 , wherein the invitation action comprises displaying a message.
18. The vehicle of claim 14 , wherein the invitation action comprises outputting a sound.
19. The vehicle of claim 14 , wherein:
the functions further comprise detecting a transition of the pedestrian countdown signal from: i) a do not walk state to ii) a walk state or a countdown state, and
wherein the prediction is further based on the detecting of the transition.
20. A non-transitory computer-readable medium having stored therein instructions executable by a computing system to cause the computing system to perform functions comprising:
obtaining a camera image patch;
determining, using the camera image patch and a pedestrian countdown signal classifier model, a state of a pedestrian countdown signal;
determining, based on the state of the pedestrian countdown signal, a prediction of whether a pedestrian will enter a crosswalk governed by the pedestrian countdown signal; and
based on the prediction, causing a vehicle to perform an invitation action that invites a pedestrian to enter the crosswalk.