Multi-task learning for real-time semantic and/or depth aware instance segmentation and/or three-dimensional object bounding

1. A system comprising:

one or more processors; and

a memory storing processor-executable instructions that, when executed by the one or more processors, cause the system to perform operations comprising:

receiving an image from an image sensor associated with an autonomous vehicle;

inputting at least a portion of the image into a machine learned (ML) model;

determining, by the ML model and based on the image, a set of outputs, the set of outputs comprising:

a region of interest (ROI) associated with an object that appears in the image;

a semantic segmentation associated with the object, the semantic segmentation indicative of a classification of the object;

directional data that indicates a center of the object, wherein a portion of the directional data indicates a direction toward the center of the object from the portion;

depth data associated with at least the portion of the image, wherein determining the depth data comprises:

determining, a depth bin from among a set of depth bins, the depth bin associated with a discrete portion of an environment; and

determining a depth residual associated with the depth bin, the depth residual indicating a deviation of a surface associated with the discrete portion from a position associated with the depth bin; and

an instance segmentation associated with the object; and

controlling the autonomous vehicle based at least in part on at least one of the ROI, the semantic segmentation, the instance segmentation, or the depth data.

2. The system of claim 1 , wherein determining the set of outputs comprises:

determining a first set of features associated with a first resolution;

determining a second set of features associated with a second resolution, the first resolution differing from the second resolution;

upsampling, as upsampled features, the second set of features to have a same resolution as the first resolution; and

combining, as combined features, the upsampled features with the first set of features,

wherein at least one of the semantic segmentation, depth data, directional data, or instance segmentation is based at least in part on the combined features.

3. The system of claim 1 , wherein the set of outputs further comprises a three-dimensional ROI.

4. The system of claim 1 , wherein determining the depth bin comprises:

determining a smoothed set of logits based at least in part on determining an average or probability distribution of logits of other discrete portions in a region surrounding the discrete portion; and

selecting the depth bin from among the set of depth bins based at least in part on determining the depth bin is associated with a maximum smoothed logit value among the smoothed set of logits.

5. A method comprising:

receiving an image from an image sensor associated with an autonomous vehicle;

inputting at least a portion of the image into a machine learned (ML) model;

determining, by the ML model and based on the image, a set of outputs, the set of outputs comprising:

a semantic segmentation associated with an object;

depth data associated with at least the portion of the image; and

an instance segmentation associated with the object,

wherein determining the depth data comprises:

determining, a depth bin from among a set of depth bins, the depth bin associated with a discrete portion of an environment; and

determining a depth residual associated with the depth bin, the depth residual indicating a deviation of a surface associated with the discrete portion from a position associated with the depth bin; and

controlling the autonomous vehicle based at least in part on at least one of the semantic segmentation, the instance segmentation, or the depth data.

6. The method of claim 5 , wherein determining the set of outputs comprises:

determining a first set of features associated with a first resolution;

determining a second set of features associated with a second resolution, the first resolution differing from the second resolution;

upsampling, as an upsampled features, the second set of features to have a same resolution as the first resolution; and

combining, as combined features, the upsampled features with the first set of features,

wherein at least one of the semantic segmentation, depth data, or instance segmentation is based at least in part on the combined features.

7. The method of claim 5 , wherein the set of outputs further comprises a three-dimensional region of interest (ROI).

8. The method of claim 7 , wherein:

the set of outputs further comprises directional data that indicates a center of the object; and

determining the three-dimensional ROI is based at least in part on the semantic segmentation, the depth data, the directional data, and the instance segmentation.

9. The method of claim 5 , wherein determining the depth bin comprises:

determining a smoothed set of logits based at least in part on determining an average or probability distribution of logits of other discrete portions in a region surrounding the discrete portion; and

selecting the depth bin from among the set of depth bins based at least in part on determining the depth bin is associated with a maximum smoothed logit value among the smoothed set of logits.

10. The method of claim 5 , wherein:

the set of outputs further comprises directional data that indicates a center of the object; and

determining the instance segmentation is based at least in part on the semantic segmentation, the depth data, and the directional data.

11. A non-transitory computer-readable medium storing computer-executable instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

receiving an image from an image sensor associated with an autonomous vehicle;

inputting at least a portion of the image into a machine learned (ML) model;

determining, by the ML model and based on the image, a set of outputs, the set of outputs comprising:

a semantic segmentation associated with an object;

depth data associated with at least the portion of the image, wherein determining the depth data comprises:

determining, a depth bin from among a set of depth bins, the depth bin associated with a discrete portion of an environment; and

determining a depth residual associated with the depth bin, the depth residual indicating a deviation of a surface associated with the discrete portion from a position associated with the depth bin;

directional data that indicates a center of the object, wherein a portion of the directional data indicates a direction toward the center of the object from the portion; and

an instance segmentation associated with the object; and

controlling the autonomous vehicle based at least in part on at least one of the semantic segmentation, the instance segmentation, or the depth data.

12. The non-transitory computer-readable medium of claim 11 , wherein determining the set of outputs comprises:

determining a first set of features associated with a first resolution;

determining a second set of features associated with a second resolution, the first resolution differing from the second resolution;

upsampling, as upsampled features, the second set of features to have a same resolution as the first resolution; and

combining, as combined features, the upsampled features with the first set of features,

wherein at least one of the semantic segmentation, depth data, or instance segmentation is based at least in part on the combined features.

13. The non-transitory computer-readable medium of claim 12 , wherein determining the set of outputs further comprises:

downsampling, as downsampled features, the combined features to reduce a number of channels associated with the combined features;

convolving, as convolved features, the downsampled features two or more times according to different dilation rates;

upsampling, as a feature data structure, the convolved features, and

wherein at least one of the semantic segmentation, depth data, or instance segmentation is based at least in part on the feature data structure.

14. The non-transitory computer-readable medium of claim 11 , wherein the set of outputs further comprises a three-dimensional region of interest (ROI).

15. The non-transitory computer-readable medium of claim 14 , wherein:

the set of outputs further comprises directional data that indicates a center of the object; and

determining the three-dimensional ROI is based at least in part on the semantic segmentation, the depth data, the directional data, and the instance segmentation.

16. The non-transitory computer-readable medium of claim 11 , wherein determining the depth bin comprises:

determining a smoothed set of logits based at least in part on determining an average or probability distribution of logits of other discrete portions in a region surrounding the discrete portion; and

selecting the depth bin from among the set of depth bins based at least in part on determining the depth bin is associated with a maximum smoothed logit value among the smoothed set of logits.

17. The non-transitory computer-readable medium of claim 11 , wherein:

the set of outputs further comprises directional data that indicates a center of the object; and

determining the instance segmentation is based at least in part on the semantic segmentation, the depth data, and the directional data.