Similarly, the system 100 may process a second point cloud 540 to generate second mesh data 550, with a second comparison 560 illustrating an accuracy of the second mesh data 550 relative to the second point cloud 540. For example, the comparison 560 illustrates the second point cloud 540 (e.g., represented as specific data points) interleaved with the second mesh data 550 (e.g., represented as an outline), indicating that there are only minor differences between the two. Thus, the second mesh data 550 accurately approximates the second point cloud 540 such that there only small portions of the second point cloud 540 distinguishable from the second mesh data 550.

FIG. 6 illustrates an example of a classification network according to embodiments of the present disclosure. In some examples, the system 100 performs point cloud classification and may generate object classification 119, which indicates an object label inferred from the point cloud data. Thus, the object classification 119 may correspond to an object label, classification label, point cloud classification, and/or the like and may indicate a classification associated with the point cloud data (e.g., human, cat, box, piano, etc.).