search

pyg-team / pytorch_geometric

Graph Neural Network Library for PyTorch
https://pyg.org
MIT License
21.09k stars 3.63k forks source link

Point Cloud Learning #2717

Open asiamanman opened 3 years ago

asiamanman commented 3 years ago

Do you know how to train the point cloud data measured by LiDAR converted into a PCD file as training data?

rusty1s commented 3 years ago

You mean something like this?

asiamanman commented 3 years ago

I want to train point cloud data

rusty1s commented 3 years ago

Can you tell me a bit more? Other-wise this Colab and this example should be useful to you :)

asiamanman commented 3 years ago

I think the data (ModelNet10) used in pointnet2_classification.py is the data of the cad system with the extension .off. I want to train data with a .pcd extension. In that case, what part of the source code of pointnet2_classification.py should be changed?

rusty1s commented 3 years ago

You can use the original ModelNet dataset and pass in the SamplePoints transform.

asiamanman commented 3 years ago

OK, thank you. I think that the output label is calculated by pred = model (data) .max (1) [1] in the function test in pointnet2_classfication. How is this result calculated?

rusty1s commented 3 years ago

model(data) returns the output probabilities for each class. We simply take the class with the highest probability as prediction, e.g., via max(dim=-1)[1] or argmax(dim=-1).

asiamanman commented 3 years ago

OK, Thank you