Machine learning feature vector generator using depth image foreground attributes
Certain aspects provide a method, including: receiving a depth image from a depth sensor; receiving a segmentation mask corresponding to the depth image and segmenting the depth image into a set of foreground pixels and a set of background pixels; determining a set of seed pixels in the depth image; for each respective seed pixel of the set of seed pixels: determining a sampling line in the depth image that starts at the respective seed pixel and passes through a portion of the depth image; for each respective sampling line pixel in the sampling line having a value in the segmentation mask indicating a foreground object in the depth image: determining one or more data attribute values based on a depth value for the respective sampling line pixel in the depth image; and adding the one or more data attribute values to a feature vector.
1. A computer-implemented method of foreground object analysis of image data, comprising, via at least one processor of at least one computing device:
accessing a depth image comprising a plurality of image pixels;
accessing a segmentation mask comprising a plurality of segmentation pixels that correspond to the depth image;
determining whether one of the plurality of segmentation pixels corresponds to a background object or a foreground object; and
determining feature data of one of the plurality of image pixels corresponding to the one of the plurality of segmentation pixels responsive to the one of the plurality of segmentation pixels corresponding to the foreground object.
2. The computer-implemented method of claim 1 , comprising not determining feature data of the one of the plurality of image pixels responsive to the one of the plurality of segmentation pixels corresponding to the background object.
3. The computer-implemented method of claim 1 , wherein the feature data comprises a pixel coordinate and a depth value.
4. The computer-implemented method of claim 1 , comprising adding the feature data to a unidimensional feature vector.
5. The computer-implemented method of claim 4 , comprising providing the unidimensional feature vector to a machine learning model configured to generate at least one inference based on the input, the inference associated with at least one characteristic of the at least one foreground object.
6. The computer-implemented method of claim 5 , wherein the machine learning model is a fully-connected neural network.
7. The computer-implemented method of claim 1 , comprising determining a sequence of seed pixels for traversing the depth image, the sequence of seed pixels comprising a set of image pixels starting at an initial coordinate and traversing at least one perimeter of the depth image.
8. A computer-implemented method of generating an inference based on feature data of a depth image, comprising, via at least one processor of at least one computing device:
accessing input data comprising a unidimensional feature vector associated with a foreground object of a depth image;
providing the input data to a machine learning model trained to generate the inference; and
receiving output from the machine learning model comprising the inference.
9. The computer-implemented method of claim 8 , wherein the unidimensional feature vector comprises a vector of feature data of at least one image pixel of the depth image corresponding to a foreground object in a segmentation mask associated with the depth image.
10. The computer-implemented method of claim 8 , wherein the unidimensional feature vector is determined via:
accessing the depth image comprising a plurality of image pixels;
accessing a segmentation mask comprising a plurality of segmentation pixels that correspond to the depth image;
determining whether one of the plurality of segmentation pixels corresponds to a background object or a foreground object; and
adding feature data of one of the plurality of image pixels corresponding to the one of the plurality of segmentation pixels responsive to the one of the plurality of segmentation pixels corresponding to the foreground object to the unidimensional feature vector.
11. The computer-implemented method of claim 8 , further comprising training the machine learning model using the unidimensional feature vector and ground truth data associated with the inference.
12. The computer-implemented method of claim 8 , wherein the inference is associated with at least one characteristic of the at least one foreground object.
13. The computer-implemented method of claim 8 , wherein the machine learning model is a fully-connected neural network.
14. An apparatus, comprising:
at least one processor; and
at least one memory coupled to the at least one processor, the at least one memory comprising instructions that, when executed by the at least one processor, cause the at least one processor to:
access a depth image comprising a plurality of image pixels;
access a segmentation mask comprising a plurality of segmentation pixels that correspond to the depth image;
determine whether one of the plurality of segmentation pixels corresponds to a background object or a foreground object; and
determine feature data of one of the plurality of image pixels corresponding to the one of the plurality of segmentation pixels responsive to the one of the plurality of segmentation pixels corresponding to the foreground object.
15. The apparatus of claim 14 , the instructions, when executed by the at least one processor, to cause the at least one processor to not determine feature data of the one of the plurality of image pixels responsive to the one of the plurality of segmentation pixels corresponding to the background object.
16. The apparatus of claim 14 , wherein the feature data comprises a pixel coordinate and a depth value.
17. The apparatus of claim 14 , the instructions, when executed by the at least one processor, to cause the at least one processor to add the feature data to a unidimensional feature vector.
18. The apparatus of claim 17 , the instructions, when executed by the at least one processor, to cause the at least one processor to provide the unidimensional feature vector to a machine learning model configured to generate at least one inference based on the input, the inference associated with at least one characteristic of the at least one foreground object.
19. The apparatus of claim 18 , wherein the machine learning model is a fully-connected neural network.
20. The apparatus of claim 14 , the instructions, when executed by the at least one processor, to cause the at least one processor to determine a sequence of seed pixels for traversing the depth image, the sequence of seed pixels comprising a set of image pixels starting at an initial coordinate and traversing at least one perimeter of the depth image.