Seafloor_mapping_PointNet2

A seafloor mapping model based on PointNet++ using Pytorch

Usage

Model training

Firstly, make sure you have at least one ICESat-2 h5 file in you data directory, then change mode to 'train', and run:

./preprocessing_script.sh

Then, annotate the data in the 'split_data' folder. Remember to put your annotated data in a folder called 'input_data'.

Lastly, when your data is ready, run:

./train_script.sh
./test_script.sh

Model prediction

Make sure you have at least one ICESat-2 h5 file in you data directory, then change mode to 'test', and run:

./preprocessing_script.sh
./predict_script.sh

Data

We have provided the training data as "data_8192.zip".

And we've also provided the trained model "model.pth" in the "trained_model" folder.

Acknowlegements

@article{Pytorch_Pointnet_Pointnet2, Author = {Xu Yan}, Title = {Pointnet/Pointnet++ Pytorch}, Journal = { https://github.com/yanx27/Pointnet_Pointnet2_pytorch }, Year = {2019} }