Point cloud semantic segmentation model with general multi-feature fusion and extraction.
The link to the paper is: https://doi.org/10.1080/01431161.2025.2603691
- OS: Windwos 10 or 11
- RAM: 64GB or above
- GPU: NVIDIA Quadro RTX6000 or similar
- IDE: Spyder (in Anaconda)
- Environment: Python 3.7.16, Keras 2.6.0(with Tensorflow 2.6.0 as the backend), cudatoolkit 11.3.1, cudnn 8.2.1, plyfile 0.7.2, scikit-learn 1.0.2, pandas 1.1.5, Open3D 0.17.0
Take the SensatUrban dataset as an example. Start by activating the environment and opening Spyder.
Then open the file "step1_GMFE-Net_pre_data_random_subsample_SensatUrban.py".
Click "Run" to start the data preprocessing.
After running the program "step2_GMFE-Net_train_test_visulization_SensatUrban.py", the following options are displayed.
The final visualization is shown in the figure below.
Note: The path in the code needs to be modified to where you store the dataset.
This project wouldn't have been possible without the support and contributions of several individuals and resources.
Thanks to (in no particular order):
- https://github.com/charlesq34/pointnet
- https://github.com/charlesq34/pointnet2
- https://github.com/HuguesTHOMAS/KPConv
- https://github.com/hkust-vgd/shellnet
- https://github.com/QingyongHu/RandLA-Net
- https://github.com/leofansq/SCF-Net
- https://github.com/WangYueFt/dgcnn
- https://keras.io/examples/vision/pointnet/
Due to GitHub's limitation on uploading large files, we save the weights files at the following URL: https://drive.google.com/drive/folders/1CDEY-h0tkaENEV_f9XltuCIB7-LRvClj?usp=drive_link
Dataset download URL: