Working with pointcloud datasets

Nowadays it's not that unusual having to work with scanned files (point clouds), but most DCCs don't provide native support for those file formats and we ended up whether using external converters or parsing ASCII files.

This entry is about the latter.

There's the quick and dirty approach, where we simply read the file and create points (nulls, spheres, cubes, whatever) using those coordinates via scripting, it works (sort of) in very simple scenarios, but it's not reliable when dealing with large point clouds (and that's usually the case).

Plan B: recreating a native icecache file!

Cache files can be created using python... but it's a bit low-level and write such code kinda sucks!

Fortunatelly there's a not-so-known python module by Bradley Gabe that provides a higher level API and free us from dealing with bytes, data blocks, gzip and other boring stuff.

Standford Bunny

So, parsing a xyz file (like this stanford bunny) and write an icecache file to disk should look like this:

#!/usr/bin/env python
# -*- coding: utf-8 -*-
import os
import sys
import re
import icecache  # Brad's API

def parse(xyz_file):
    with open(xyz_file) as f:
        coords = re.findall(r"(-?[0-9\.]+)[\s\t\n]+",
    pos = [[float(coords[i]), float(coords[i + 1]), float(coords[i + 2])]
           for i in range(len(coords))[::3]]
    ic = icecache.icecache(len(pos))

if __name__ == "__main__":
    if len(sys.argv) == 2 and sys.argv[1].endswith(".xyz"):
        raise Exception("Invalid file type")

Now, how cool is that? a simple regex plus 3 calls to Brad's API and we're done.


Questions? Comments?

Please feel free to ping me on twitter or send me an email, I would love to hear from you!