Image-enhanced depth sensing via depth sensor control
Systems and methods are disclosed for computing depth maps. One method includes capturing, using a camera, a camera image of a runtime scene. The method may also include analyzing the camera image of the runtime scene to determine a plurality of target sampling points at which to capture depth of the runtime scene. The method may further include adjusting a setting associated with a low-density depth sensor based on the plurality of target sampling points. The method may further include capturing, using the low-density depth sensor, a low-density depth map of the runtime scene at the plurality of target sampling points. The method may further include generating a computed depth map of the runtime scene based on the camera image of the runtime scene and the low-density depth map of the runtime scene.
1. A method of computing depth maps, the method comprising: capturing, using a camera, a camera image of a runtime scene; analyzing the camera image of the runtime scene to determine a plurality of target sampling points at which to capture depth of the runtime scene, wherein the plurality of target sampling points are determined using a neural network having been previously trained using a camera image of a training scene captured by the camera and a high-density depth map of the training scene captured by a high-density depth sensor; adjusting a setting associated with a low-density depth sensor based on the plurality of target sampling points such that the low-density depth sensor is configured to capture depths of the runtime scene at each of the plurality of target sampling points; capturing, using the low-density depth sensor, a low-density depth map of the runtime scene at the plurality of target sampling points, wherein adjusting the setting includes causing a first subset of a plurality of pixel receivers of the low-density depth sensor to be powered and a second subset of the plurality of pixel receivers to be unpowered, wherein the first subset corresponds to the plurality of target sampling points, and wherein powering the first subset causes the first subset to receive reflected signals from the plurality of target sampling points; and generating a computed depth map of the runtime scene based on the camera image of the runtime scene and the low-density depth map of the runtime scene.
2. The method of claim 1 , wherein analyzing the camera image of the runtime scene to determine the plurality of target sampling points includes: detecting one or more interest points in the camera image of the runtime scene; and determining the plurality of target sampling points based on the one or more interest points.
3. The method of claim 1 , wherein analyzing the camera image of the runtime scene to determine the plurality of target sampling points includes: generating, using the neural network, the plurality of target sampling points based on the camera image of the runtime scene.
4. The method of claim 1 , wherein adjusting the setting associated with the low-density depth sensor based on the plurality of target sampling points includes: causing a transmitter of the low-density depth sensor to transmit signals to the plurality of target sampling points.
5. The method of claim 1 , wherein adjusting the setting associated with the low-density depth sensor based on the plurality of target sampling points includes: causing a transmitter of the low-density depth sensor to sequentially transmit signals to the plurality of target sampling points; and causing a receiver of the low-density depth sensor to receive reflected signals from the plurality of target sampling points.
6. The method of claim 1 , wherein generating the computed depth map of the runtime scene based includes: generating, using the neural network, the computed depth map of the runtime scene based on the camera image of the runtime scene and the low-density depth map of the runtime scene.
7. A system for computing depth maps, the system comprising: a camera; a low-density depth sensor; and one or more processors configured to perform operations including: capturing, using the camera, a camera image of a runtime scene; analyzing the camera image of the runtime scene to determine a plurality of target sampling points at which to capture depth of the runtime scene, wherein the plurality of target sampling points are determined using a neural network having been previously trained using a camera image of a training scene captured by the camera and a high-density depth map of the training scene captured by a high-density depth sensor; adjusting a setting associated with the low-density depth sensor based on the plurality of target sampling points such that the low-density depth sensor is configured to capture depths of the runtime scene at each of the plurality of target sampling points; capturing, using the low-density depth sensor, a low-density depth map of the runtime scene at the plurality of target sampling points, wherein adjusting the setting includes causing a first subset of a plurality of pixel receivers of the low-density depth sensor to be powered and a second subset of the plurality of pixel receivers to be unpowered, wherein the first subset corresponds to the plurality of target sampling points, and wherein powering the first subset causes the first subset to receive reflected signals from the plurality of target sampling points; and generating a computed depth map of the runtime scene based on the camera image of the runtime scene and the low-density depth map of the runtime scene.
8. The system of claim 7 , wherein analyzing the camera image of the runtime scene to determine the plurality of target sampling points includes: detecting one or more interest points in the camera image of the runtime scene; and determining the plurality of target sampling points based on the one or more interest points.
9. The system of claim 7 , wherein analyzing the camera image of the runtime scene to determine the plurality of target sampling points includes: generating, using the neural network, the plurality of target sampling points based on the camera image of the runtime scene.
10. The system of claim 7 , wherein adjusting the setting associated with the low-density depth sensor based on the plurality of target sampling points includes: causing a transmitter of the low-density depth sensor to transmit signals to the plurality of target sampling points.
11. The system of claim 7 , wherein adjusting the setting associated with the low-density depth sensor based on the plurality of target sampling points includes: causing a transmitter of the low-density depth sensor to sequentially transmit signals to the plurality of target sampling points; and causing a receiver of the low-density depth sensor to receive reflected signals from the plurality of target sampling points.
12. The system of claim 7 , wherein generating the computed depth map of the runtime scene based includes: generating, using the neural network, the computed depth map of the runtime scene based on the camera image of the runtime scene and the low-density depth map of the runtime scene.
13. A non-transitory computer-readable medium comprising instructions that, when executed by a processor, cause the processor to perform operations comprising: capturing, using a camera, a camera image of a runtime scene; analyzing the camera image of the runtime scene to determine a plurality of target sampling points at which to capture depth of the runtime scene, wherein the plurality of target sampling points are determined using a neural network having been previously trained using a camera image of a training scene captured by the camera and a high-density depth map of the training scene captured by a high-density depth sensor; adjusting a setting associated with a low-density depth sensor based on the plurality of target sampling points such that the low-density depth sensor is configured to capture depths of the runtime scene at each of the plurality of target sampling points; capturing, using the low-density depth sensor, a low-density depth map of the runtime scene at the plurality of target sampling points, wherein adjusting the setting includes causing a first subset of a plurality of pixel receivers of the low-density depth sensor to be powered and a second subset of the plurality of pixel receivers to be unpowered, wherein the first subset corresponds to the plurality of target sampling points, and wherein powering the first subset causes the first subset to receive reflected signals from the plurality of target sampling points; and generating a computed depth map of the runtime scene based on the camera image of the runtime scene and the low-density depth map of the runtime scene.
14. The non-transitory computer-readable medium of claim 13 , wherein analyzing the camera image of the runtime scene to determine the plurality of target sampling points includes: detecting one or more interest points in the camera image of the runtime scene; and determining the plurality of target sampling points based on the one or more interest points.
15. The non-transitory computer-readable medium of claim 13 , wherein analyzing the camera image of the runtime scene to determine the plurality of target sampling points includes: generating, using the neural network, the plurality of target sampling points based on the camera image of the runtime scene.
16. The non-transitory computer-readable medium of claim 13 , wherein adjusting the setting associated with the low-density depth sensor based on the plurality of target sampling points includes: causing a transmitter of the low-density depth sensor to transmit signals to the plurality of target sampling points.
17. The non-transitory computer-readable medium of claim 13 , wherein adjusting the setting associated with the low-density depth sensor based on the plurality of target sampling points includes: causing a transmitter of the low-density depth sensor to sequentially transmit signals to the plurality of target sampling points; and causing a receiver of the low-density depth sensor to receive reflected signals from the plurality of target sampling points.