Performing semantic segmentation of 3D data using deep learning
A computer-implemented method of training a deep artificial neural network includes receiving a three-dimensional point cloud and training the deep artificial neural network by subdividing the three-dimensional point cloud, and updating weights of the deep artificial neural network. A computing system includes a processor; and a memory having stored thereon computer-executable instructions that, when executed by the processor, cause the computing system to receive a three-dimensional point cloud and train the deep artificial neural network by subdividing the three-dimensional point cloud, and updating weights of the deep artificial neural network. In yet another aspect, a non-transitory computer-readable medium includes computer-executable instructions that when executed, cause a computer to receive a three-dimensional point cloud and train the deep artificial neural network by subdividing the three-dimensional point cloud, and updating weights of the deep artificial neural network.
1 . A computer-implemented method of training a deep artificial neural network to generate a semantically-segmented three-dimensional point cloud, the computer-implemented method comprising:
receiving a three-dimensional point cloud having at least one labeled feature corresponding to an outdoor structure; and
training the deep artificial neural network to output a set of point labels corresponding to an outdoor scene by:
subdividing the three-dimensional point cloud into a plurality of subdivisions, wherein at least one subdivision of the plurality of subdivisions overlaps with another subdivision of the plurality of subdivisions;
pre-processing the plurality of subdivisions by:
determining a corresponding number of samples included in each of the plurality of subdivisions; and
downsampling or upsampling a subdivision of the plurality of subdivisions in response to determining that the corresponding number of samples included in the subdivision does not equal a particular number of samples;
normalizing Z coordinate values such that a ground layer of the outdoor scene is at an origin Z-value; and
processing the at least one labeled feature to update one or more weights of the deep artificial neural network.
2 . The computer-implemented method of claim 1 , wherein pre-processing the plurality of subdivisions further includes:
collapsing a set of points within the plurality of subdivisions onto a unit sphere.
3 . The computer-implemented method of claim 2 , further comprising:
generating a semantically-segmented three-dimensional point cloud by using the trained deep artificial neural network to analyze a three-dimensional point cloud not used for training the deep artificial neural network; and
storing the semantically-segmented three-dimensional point cloud on a computer readable storage medium, wherein the three-dimensional point cloud comprises the set of point labels.
4 . The computer-implemented method of claim 1 , wherein one or more of the point labels is determined using a pairwise distance function.
5 . The computer-implemented method of claim 1 , wherein one or more of the point labels is determined by selecting an arg max of a vector of labels corresponding to the one or more point labels.
6 . The computer-implemented method of claim 1 , wherein each of the plurality of subdivisions is columnar in shape.
7 . The computer-implemented method of claim 1 , wherein the set of point labels includes a type and an elevation.
8 . A computing system for training a deep artificial neural network to generate a semantically-segmented three-dimensional point cloud, the computing system comprising:
one or more processors; and
one or more memories having stored thereon computer-executable instructions that, when executed by the one or more processors, cause the computing system to:
receive a three-dimensional point cloud having at least one labeled feature corresponding to an outdoor structure; and
train the deep artificial neural network to output a set of point labels corresponding to an outdoor scene by:
subdivide the three-dimensional point cloud into a plurality of subdivisions, wherein at least one subdivision of the plurality of subdivisions overlaps with another subdivision of the plurality of subdivisions;
pre-process the plurality of subdivisions by:
determining a corresponding number of samples included in each of the plurality of subdivisions; and
downsampling or upsampling a subdivision of the plurality of subdivisions in response to determining that the corresponding number of samples included in the subdivision does not equal a particular number of samples;
normalize Z coordinate values such that a ground layer of the outdoor scene is at an origin Z-value; and
process the at least one labeled feature to update a set of weights of the deep artificial neural network.
9 . The computing system of claim 8 , wherein to pre-process the plurality of subdivisions, the processor is further configured to:
collapse a set of points within the plurality of subdivisions onto a unit sphere.
10 . The computing system of claim 9 , the one or more memories having stored thereon computer-executable instructions that, when executed by the one or more processors, cause the computing system to:
generate the semantically-segmented three-dimensional point cloud by using the trained deep artificial neural network to process a three-dimensional point cloud not used for training the deep artificial neural network, and
store the semantically-segmented three-dimensional point cloud on the one or more memories, wherein the three-dimensional point cloud comprises the set of point labels.
11 . The computing system of claim 9 , wherein one or more of the point labels is determined using a pairwise distance function.
12 . The computing system of claim 9 , wherein one or more of the point labels is determined by selecting an arg max of a vector of labels corresponding to the one or more point labels.
13 . The computing system of claim 8 , wherein each of the plurality of subdivisions is columnar in shape.
14 . The computing system of claim 8 , wherein the set of point labels includes a type and an elevation.
15 . A non-transitory computer-readable medium having stored thereon computer-executable instructions that when executed by one or more processors, cause a computer to:
receive a three-dimensional point cloud having at least one labeled feature corresponding to an outdoor structure; and
train a deep artificial neural network to output a set of point labels corresponding to an outdoor scene by:
subdivide the three-dimensional point cloud into a plurality of subdivisions, wherein at least one subdivision of the plurality of subdivisions overlaps with another subdivision of the plurality of subdivisions;
normalize Z coordinate values such that a ground layer of the outdoor scene is at an origin Z-value; and
process the at least one labeled feature to update a set of weights of the deep artificial neural network.
16 . The non-transitory computer-readable medium of claim 15 , having stored thereon further instructions that, when executed by the one or more processors, cause the computer to collapse a set of points within the plurality of subdivisions onto a unit sphere.
17 . The non-transitory computer-readable medium of claim 16 , having stored thereon further instructions that, when executed by the one or more processors, cause the computer to:
generate a semantically-segmented three-dimensional point cloud by using the trained deep artificial neural network to analyze a three-dimensional point cloud not used for training the deep artificial neural network; and
store the semantically-segmented three-dimensional point cloud on the computer readable medium, wherein the three-dimensional point cloud comprises the set of point labels.
18 . The non-transitory computer-readable medium of claim 16 , wherein one or more of the point labels is determined using a pairwise distance function.
19 . The non-transitory computer-readable medium of claim 16 , wherein one or more of the point labels is determined by selecting an arg max of a vector of labels corresponding to the one or more point labels.
20 . The non-transitory computer-readable medium of claim 16 , wherein the set of point labels includes a type and an elevation.