How to process my point cloud file with the network about samplenet

[itailang]

Hi @barrym1,

To see how the data is loaded, follow the variable self.data_dir in modelnet_loader_torch.

You can see the input and output data in the function compute_pcrnet_loss. The sampled point clouds are p0 and p1 and the rotated point cloud is p1_est. You can find a point cloud visualization example here.

Good luck!

[barrym1]

Sorry, I still can't solve this problem, can you tell me the specific steps?I am very interested in your research.Thank you so much!

[itailang]

Can you explain what exactly the problem is?

[barrym1]

My problem is how to process my point cloud file(my_cloud.ply) with the network after training. Can you tell me the specific steps?

[itailang]

I assume that you already have the template (original) and source (rotated) point clouds in your file. They are denoted as $p_0$ and $p_1$, respectively.

1. Load your data. See the function _load_data_file for reference.
2. Feed the point clouds to SampleNet to sample them. See the function compute_samplenet_loss for reference.
3. Feed the sampled point clouds to the task network to compute the rotation for registration. See the function compute_pcrnet_loss for reference.