The environment of Point-NN

[whuhxb]

Hi @ZrrSkywalker

What's the specific environment of Point-NN? Which cuda version and pytorch version? Thanks a lot.

[ZrrSkywalker]

Hi. Thanks for your interest. Our environment is:

torch             1.7.1+cu110
torchvision       0.8.2+cu110
cudatoolkit       10.2.89

Our nvcc -V with RTX 3090 is

cuda_11.2

[whuhxb]

Hi. Thanks for your interest. Our environment is:

torch             1.7.1+cu110
torchvision       0.8.2+cu110
cudatoolkit       10.2.89

Our nvcc -V with RTX 3090 is

cuda_11.2

Could other environment work? For example, cuda10.1 and pytorch 1.7.0. Thanks a lot.

[ZrrSkywalker]

Yes it can work.

[whuhxb]

Yes it can work.

OK. Will have a try. Thanks.

[whuhxb]

Yes it can work.

Have you tested this backbone for semantic segmentation task? Thanks.

[whuhxb]

Yes it can work.

The code runs so quick. How to visualize the classification result? And, how to output the classification result? Thanks.

[ZrrSkywalker]

Yes it can work.

The code runs so quick. How to visualize the classification result? And, how to output the classification result? Thanks.

The classification logits of the test set are here: https://github.com/ZrrSkywalker/Point-NN/blob/9707b2ed5afe79be79b9a19d70aec9b04d8fb648/run_nn_cls.py#L120

but you need to use the best_gamma after the loop to obtain final logits of Point-NN.

[whuhxb]

Yes it can work.

The code runs so quick. How to visualize the classification result? And, how to output the classification result? Thanks.

The classification logits of the test set are here:

https://github.com/ZrrSkywalker/Point-NN/blob/9707b2ed5afe79be79b9a19d70aec9b04d8fb648/run_nn_cls.py#L120

but you need to use the best_gamma after the loop to obtain final logits of Point-NN.

OK. Thanks.