Throughput reduction in autonomous vehicle camera sensors
Throughput reduction in autonomous vehicle camera sensors, including: generating, by a camera sensor, a frame; selecting an area of focus for the frame; and generating, by the camera sensor from the frame, a downsampled frame and a cropped frame, wherein the cropped frame is based on the area of focus.
1. A method for throughput reduction in high resolution camera sensors, comprising:
generating, by a camera sensor, a frame;
selecting, by the camera sensor, an area of focus for the frame;
generating, by the camera sensor from the frame, a downsampled frame and a cropped frame, wherein the cropped frame is based on the area of focus; and
providing, by the camera sensor to a computing system of an autonomous vehicle via a data link having the camera sensor at a first end and the computing system at the second end, the downsampled frame and the cropped frame instead of the frame.
2. The method of claim 1 , further comprising:
selecting another area of focus for the frame;
generating, from the frame, another cropped frame based on the other area of focus; and
wherein providing the downsampled frame and the cropped frame further comprises providing the other cropped frame.
3. The method of claim 1 , wherein selecting the area of focus is based on one or more models.
4. The method of claim 3 , wherein selecting the area of focus comprises selecting, based on the one or more models, as the area of focus, an area of the image having one or more of: a degree of motion exceeding a threshold or a highest degree of motion.
5. The method of claim 3 , wherein selecting the area of focus comprises selecting, based on the one or more models, as the area of focus, an area of the image having a highest spatial density.
6. The method of claim 3 , wherein selecting the area of focus comprises selecting, based on the one or more models, as the area of focus, an area of the image having a highest object density.
7. The method of claim 1 , wherein the camera sensor is included in a plurality of camera sensors of the autonomous vehicle, and wherein generating the frame, selecting the area of focus, and providing the downsampled frame and the cropped frame is performed by the plurality of camera sensors of the autonomous vehicle.
8. The method of claim 7 , wherein the plurality of camera sensors comprise a plurality of camera sensor pairs each corresponding to a different facing of the autonomous vehicle.
9. The method of claim 8 , wherein each first camera sensor in a given sensor pair of the plurality of camera sensor pairs is configured to coordinate their selected area of focus with a second camera sensor in the given camera sensor pair.
10. The method of claim 9 , wherein each camera sensor pair of the plurality of camera sensor pairs is in a stereoscopic configuration.
11. The method of claim 1 , wherein providing, via the data link, the downsampled frame and the cropped frame instead of the frame comprises providing the downsampled frame and the cropped frame as separate frames via the data link.
12. The method of claim 1 , wherein providing, via the data link, the downsampled frame and the cropped frame instead of the frame comprises:
mapping the cropped frame to the downsampled frame to generate a new frame; and
providing the new frame via the data link.
13. An autonomous vehicle camera sensor for throughput reduction in high resolution camera sensors, the autonomous vehicle camera sensor configured to perform steps comprising:
generating a frame;
selecting an area of focus for the frame;
generating a downsampled frame and a cropped frame, wherein the cropped frame is based on the area of focus: and
providing, to an automation computing system of an autonomous vehicle via a data link having the camera sensor at a first end and the automation computing system at the second end, the downsampled frame and the cropped frame instead of the frame.
14. The autonomous vehicle camera sensor of claim 13 , wherein the steps further comprise:
selecting another area of focus for the frame;
generating, from the frame, another cropped frame based on the other area of focus; and
wherein providing the downsampled frame and the cropped frame further comprises providing the other cropped frame to the automation computing system.
15. The autonomous vehicle camera sensor of claim 13 , wherein selecting the area of focus is based on one or more models.
16. The autonomous vehicle camera sensor of claim 13 , wherein the autonomous vehicle camera sensor is included in a plurality of autonomous vehicle camera sensors of the autonomous vehicle.
17. The autonomous vehicle camera sensor of claim 16 , wherein the plurality of autonomous vehicle camera sensors comprise a plurality of camera sensor pairs each corresponding to a different facing of the autonomous vehicle.
18. An autonomous vehicle for throughput reduction in high resolution camera sensors, comprising:
an automation computing system;
an autonomous vehicle camera sensor configured to perform steps comprising:
generating a frame;
selecting an area of focus for the frame;
generating a downsampled frame and a cropped frame, wherein the cropped frame is based on the area of focus; and
providing, to the automation computing system via a data link having the camera sensor at a first end and the automation computing system at the second end, the downsampled frame and the cropped frame instead of the frame.
19. The autonomous vehicle of claim 18 , wherein the steps further comprise:
selecting another area of focus for the frame;
generating, from the frame, another cropped frame based on the other area of focus; and
wherein providing the downsampled frame and the cropped frame further comprises providing the other cropped frame to the automation computing system.
20. The autonomous vehicle of claim 18 , wherein selecting the area of focus is based on one or more models.