Does the data preparation process is not right?

[What-ifs]

trafficstars

I visualized some CAD modules in which some strange vertices exists, the corresponding consequence is that one normalize vertices at first will results in impertinent scaling. For example, the file "./ModelNet40/guitar/train/guitar_0109.off", i got this: vmax is array([[322.10598755, 328.86300659, 0.539612 ]]) vmin is array([[10.35239983, 9.14276981, -4. ]]) obviously, it is not right size of guitar.

[chrischoy]

I don't understand the question, but shapes are normalized. https://github.com/StanfordVL/MinkowskiEngine/blob/master/examples/modelnet40.py#L302

[What-ifs]

Let me clarify this question clearly. I think some CAD modules in ModelNet40.tgz exists some noisy points. For the example of file "./ModelNet40/guitar/train/guitar_0109.off", the vmax is (322.10598755, 328.86300659, 0.539612) and vmin is (10.35239983, 9.14276981, -4. ), but the range of x of this module is about from 10 to 25, y is about from 9 to 48, and z is about from -4 to 0.5. so the normalization is not right according to vmax and vmin which are affected by noisy points.

[chrischoy]

I see. This is the original ModelNet40 dataset and I fixed some headers in the .off files to load them correctly. If there's another version of this dataset that cleaned up the model net, please let me know.