Systems and methods for detecting waste receptacles
A system can include a camera, a single-stage object detector, and one or more processors. The camera can capture image data that includes a waste receptacle. The single-stage object detector can detect waste receptables. The one or more processors can communicate with a waste-collection vehicle, the camera, and the single-stage object detector. The one or more processors can receive, from the camera, the image data. The one or more processors can provide, as an input, the image data to the single-stage object detector. The one or more processors can identify, based on an output of the single-stage object detector, the waste receptacle.
1. A system, comprising:
a camera configured to capture image data that includes a waste receptacle;
a single-stage object detector configured to detect waste receptables; and
one or more processors in communication with a waste-collection vehicle, the camera, and the single-stage object detector, the one or more processors configured to:
receive, from the camera, the image data;
provide, as an input, the image data to the single-stage object detector; and
identify, based on an output of the single-stage object detector, the waste receptacle.
2. The system of claim 1 , wherein the one or more processors are configured to:
select, responsive to identification of the waste receptacle, an action to implement with respect to the waste-collection vehicle.
3. The system of claim 1 , wherein the single-stage object detector is configured to:
perform, using a convolution filter, bounding box regression on the image data; and
predict, based on one or more bounding boxes, an object classification for the waste receptacle.
4. The system of claim 3 , wherein the object classification comprises at least one of garbage, recycling, compost, or background.
5. The system of claim 1 , wherein the single-stage object detector is configured to:
receive, from a feature extractor, a feature map associated with the image data; and
predict, responsive to application of a convolution filter to the feature map, at least one of a class label, a class confidence score, or a bounding box.
6. The system of claim 5 , wherein the class confidence score includes a first confidence score and a second confidence score, wherein the bounding box includes a first bounding box and a second bounding box, the first confidence score corresponding to the first bounding box, the second confidence score corresponding to the second bounding box, and wherein the single-stage object detector is configured to:
retain, based on the first confidence score, the first bounding box for subsequent evaluation; and
reject, based on the second confidence score, the second bounding box.
7. The system of claim 1 , wherein the single-stage object detector is integrated with a convolutional neural network, and wherein the convolutional neural network comprises a MobileNet architecture.
8. A waste-collection vehicle, comprising:
a camera configured to capture image data that includes a waste receptacle; and
one or more processors in communication with the camera and a single-stage object detector, the one or more processors configured to:
receive, from the camera, the image data;
provide, as an input, the image data to the single-stage object detector; and
identify, based on an output of the single-stage object detector, the waste receptacle.
9. The waste-collection vehicle of claim 8 , wherein the one or more processors are configured to:
select, responsive to identification of the waste receptacle, an action to implement with respect to the waste-collection vehicle.
10. The waste-collection vehicle of claim 8 , comprising the single-stage object detector, and wherein the single-stage object detector is configured to:
perform, using a convolution filter, bounding box regression on the image data; and
predict, based on one or more bounding boxes, an object classification for the waste receptacle.
11. The waste-collection vehicle of claim 10 , wherein the object classification comprises at least one of garbage, recycling, compost, or background.
12. The waste-collection vehicle of claim 8 , comprising the single-stage object detector, and wherein the single-stage object detector is configured to:
receive, from a feature extractor, a feature map associated with the image data; and
predict, responsive to application of a convolution filter to the feature map, at least one of a class label, a class confidence score, or a bounding box.
13. The waste-collection vehicle of claim 12 , wherein the class confidence score includes a first confidence score and a second confidence score, wherein the bounding box includes a first bounding box and a second bounding box, the first confidence score corresponding to the first bounding box, the second confidence score corresponding to the second bounding box, and wherein the single-stage object detector is configured to:
retain, based on the first confidence score, the first bounding box for subsequent evaluation; and
reject, based on the second confidence score, the second bounding box.
14. The waste-collection vehicle of claim 8 , wherein the single-stage object detector is integrated with a convolutional neural network, and wherein the convolutional neural network comprises a MobileNet architecture.
15. A system for a waste-collection vehicle, the system comprising:
one or more processors in communication with a camera of the waste-collection vehicle and a single-stage object detector, the one or more processors configured to:
receive data captured by the camera;
provide, as an input, the data to the single-stage object detector; and
identify, based on an output of the single-stage object detector, a waste receptacle.
16. The system of claim 15 , wherein the one or more processors are configured to:
select, responsive to identification of the waste receptacle, an action to implement with respect to the waste-collection vehicle.
17. The system of claim 15 , comprising the single-stage object detector, and wherein the single-stage object detector is configured to:
perform, using a convolution filter, bounding box regression on the data; and
predict, based on one or more bounding boxes, an object classification for the waste receptacle.
18. The system of claim 17 , wherein the object classification comprises at least one of garbage, recycling, compost, or background.
19. The system of claim 15 , comprising the single-stage object detector, and wherein the single-stage object detector is configured to:
receive, from a feature extractor, a feature map associated with the data; and
predict, responsive to application of a convolution filter to the feature map, at least one of a class label, a class confidence score, or a bounding box.
20. The system of claim 19 , wherein the class confidence score includes a first confidence score and a second confidence score, wherein the bounding box includes a first bounding box and a second bounding box, the first confidence score corresponding to the first bounding box, the second confidence score corresponding to the second bounding box, and wherein the single-stage object detector is configured to:
retain, based on the first confidence score, the first bounding box for subsequent evaluation; and
reject, based on the second confidence score, the second bounding box.