We provide an example of using both ExactVecKM and FastVecKM to compute the per-point local point cloud encoding, given a (100000,3) point cloud:
python main.py
It shows how the API is used, and how to check the encoding quality with the visualization tool. The command outputs are as followed:
The shape of pts is (n, 3): (100000, 3)
Preprocessing the data... Normalize the points into a unit ball.
[[ 0.00622046 0.62610329 0.07458451]
[-0.06785286 -0.46164555 0.09950156]
[ 0.0838672 -0.56487705 -0.12544057]
...
[-0.07780755 -0.05594401 0.10040794]
[ 0.11728748 0.19354883 -0.15473476]
[-0.13075289 0.20784576 0.02455094]]
ExactVecKM: The shape of per-point local geometry encoding is (n, d): torch.Size([100000, 384])
FastVecKM: The shape of per-point local geometry encoding is (n, d): torch.Size([100000, 384])
Generating images...: 100%|██████████| 36/36 [00:17<00:00, 2.02it/s]
GIF saved at exact_vkm_quality_check.gif
Generating images...: 100%|██████████| 36/36 [00:17<00:00, 2.00it/s]
GIF saved at fast_vkm_quality_check.gif
The quality checking gif are generated as followed. Left is ExactVecKM, Right is FastVecKM.
