Systems and methods for detecting objects in an image of an environment

1. A method, comprising:

receiving, by a device, an image that depicts an environment associated with a vehicle;

partitioning, by the device, the image into a plurality of subsections;

analyzing, by the device, the plurality of subsections to determine respective subsection information of the plurality of subsections,

wherein subsection information, for an individual subsection, indicates:

a probability score that the subsection includes a line segment associated with an object class,

a position of a representative point of the line segment, and

a direction of the line segment;

determining, by the device and based on the respective subsection information of the plurality of subsections, respective affinity levels of the plurality of subsections;

identifying, by the device based on aggregating the respective affinity levels of the plurality of subsections, a set of subsections of the plurality of subsections;

identifying, by the device and based on positions of points of the set of subsections, a line associated with the object class that is associated with the set of subsections of the plurality of subsections; and

performing, by the device, one or more actions based on identifying the line associated with the object class.

2. The method of claim 1 , wherein the plurality of subsections are analyzed using a machine learning model that is trained based on images that depict various vehicle environments.

3. The method of claim 1 , wherein the object class is associated with at least one of the following objects:

a lane marking;

a road edge;

a road curb; or

a highway shoulder.

4. The method of claim 1 , wherein identifying the line associated with the object class comprises:

determining, based on the respective subsection information of the plurality of subsections, that the set of subsections include respective line segments associated with the object class that have directions that match;

identifying, based on the respective subsection information of the set of subsections, positions of the representative points of the respective line segments of the set of subsections; and

processing the positions of the representative points of the respective line segments of the set of subsections to identify the line associated with the object class.

5. The method of claim 1 , wherein performing the one or more actions comprises at least one of:

rendering the line of the object class over the image on a display of the vehicle;

providing, to a navigation data structure, information associated with the environment and the line for use with a navigation system; or

providing, to a vehicle management data structure, the information for analysis associated with a driving behavior of a driver of the vehicle.

6. The method of claim 1 , wherein performing the one or more actions comprises:

identifying at least two additional lines associated with the object class that are associated with respective sets of subsections of the plurality of subsections;

identifying a bounding polygon associated with the object class based on the line associated with the object class and the at least two additional lines associated with the object class; and

causing the bounding polygon associated with the object class to be displayed on a display of the vehicle.

7. The method of claim 6 , wherein identifying the bounding polygon associated with the object class comprises:

identifying a first line, a second line, and a third line of the line associated with the object class and the at least two additional lines associated with the object class;

determining that a starting point of the first line coincides with an ending point of the second line; and

determining that an ending point of the first line coincides with a starting point of the third line.

8. A device, comprising:

one or more processors configured to:

receive an image that depicts an environment associated with a vehicle;

partition the image into a plurality of subsections;

analyze, using a machine learning model, the plurality of subsections to determine respective subsection information of the plurality of subsections;

determine, based on the respective subsection information, respective affinity levels of the plurality of subsections;

identify, based on aggregating the respective affinity levels of the plurality of subsections, a set of subsections of the plurality of subsections;

identify, based on positions of points of the set of subsections, a line associated with an object class that is associated with the set of subsections of the plurality of subsections; and

perform one or more actions based on identifying the line associated with the object class.

9. The device of claim 8 , wherein the object class is associated with at least one of the following objects:

another vehicle;

a road sign;

a street sign;

an exit sign; or

a traffic light.

10. The device of claim 8 , wherein the one or more processors, when analyzing the plurality of subsections to determine the respective subsection information, are configured to:

process a subsection, of the plurality of subsections, using the machine learning model to determine a probability score that the subsection includes a line segment associated with the object class.

11. The device of claim 8 , wherein the one or more processors, when analyzing the plurality of subsections to determine the respective subsection information, are configured to:

process a subsection using the machine learning model to determine a probability score that the subsection includes a line segment associated with the object class;

identify, based on the probability score satisfying a threshold, a representative point of the line segment in the subsection; and

determine a position of the representative point of the line segment in relation to a reference point of the subsection.

12. The device of claim 8 , wherein the one or more processors, when analyzing the plurality of subsections to determine the respective subsection information, are configured to:

process a subsection using the machine learning model to determine that a line segment associated with the object class exists in the subsection;

determine a position of a representative point of the line segment within the subsection; and

determine a direction associated with the line segment.

13. The device of claim 12 , wherein the one or more processors, when determining the direction associated with the line segment, are configured to:

determine a position of an additional point of the line segment within the subsection; and

determine the direction associated with the line segment based on a difference between the position of the representative point and the position of the additional point.

14. The device of claim 8 , wherein the line associated with the object class is a best fit line associated with representative points of respective line segments associated with the object class that are included in the set of subsections.

15. The device of claim 8 , wherein the one or more processors, when performing the one or more actions, are configured to:

identify, based on the line associated with the object class, a bounding polygon associated with the object class; and

render the bounding polygon over the image on a display of the vehicle.

16. A non-transitory computer-readable medium storing instructions, the instructions comprising:

one or more instructions that, when executed by one or more processors of a device, cause the device to:

receive an image that depicts an environment;

partition the image into a plurality of subsections;

analyze, using a machine learning model, the plurality of subsections to determine respective subsection information of the plurality of subsections;

determine, based on the respective subsection information, respective affinity levels of the plurality of subsections;

identify, based on aggregating the respective affinity levels of the plurality of subsections, a set of subsections of the plurality of subsections;

identify, based on positions of points of the set of subsections, at least one line associated with an object class that is associated with the set of subsections of the plurality of subsections; and

perform one or more actions based on identifying the at least one line associated with the object class.

17. The non-transitory computer-readable medium of claim 16 , wherein the one or more instructions that cause the device to analyze the plurality of subsections to determine the respective subsection information, cause the device to:

process a subsection, of the plurality of subsections, to determine for the subsection at least one of:

a probability score that the subsection includes a line segment associated with the object class,

a position of a representative point of the line segment, or

a direction of the line segment.

18. The non-transitory computer-readable medium of claim 16 , wherein the one or more instructions that cause the device to perform the one or more actions, cause the device to:

identifying a bounding polygon associated with the object class based on the at least one line associated with the object class; and

causing the bounding polygon associated with the object class and the image to be displayed on a display of another device.

19. The non-transitory computer-readable medium of claim 16 , wherein the one or more instructions that cause the device to perform the one or more actions, cause the device to:

provide the at least one line associated with the object class and the image for display on another device.

20. The method of claim 1 , wherein the object class is associated with at least one of the following objects:

another vehicle;

a road sign;

a street sign;

an exit sign; or

a traffic light.