| --- |
| license: mit |
| library_name: torch-pointcloud |
| tags: |
| - point-cloud |
| - 3d |
| - pytorch |
| - torch-pointcloud |
| - randlanet |
| - segmentation |
| datasets: |
| - semantic3d |
| model-index: |
| - name: randlanet.semantic3d.open3d |
| results: |
| - task: |
| type: point-cloud-segmentation |
| dataset: |
| name: Semantic3D |
| type: semantic3d |
| metrics: |
| - name: mIoU |
| type: mean_iou |
| value: 70.97 |
| - name: OA |
| type: accuracy |
| value: 83.39 |
| --- |
| |
| # Model card for randlanet.semantic3d.open3d |
|
|
| A RandLA-Net point cloud semantic segmentation model (random sampling with local feature aggregation). Trained on Semantic3D. |
|
|
| ## Model Details |
|
|
| - **Model Type:** Point cloud semantic segmentation |
| - **Model Stats:** |
| - Params (M): 5.0 |
| - Input channels: 6 |
| - Classes: 8 |
| - Features: 32 |
| - **Dataset:** Semantic3D |
| - **Metrics:** mIoU 70.97, OA 83.39 |
| - **Paper:** [RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds](https://arxiv.org/abs/1911.11236) |
| - **Converted from:** [isl-org/Open3D-ML](https://github.com/isl-org/Open3D-ML) (MIT) |
| - **Library:** [torch-pointcloud](https://github.com/arthurdjn/pytorch-pointcloud) |
|
|
| ## Install |
|
|
| ```bash |
| pip install torch-pointcloud |
| ``` |
|
|
| ## Usage |
|
|
| ```python |
| import torch |
| import torch_pointcloud as tp |
| from torch_pointcloud.utils.data import collate |
| |
| model, info = tp.create_model( |
| "randlanet.semantic3d.open3d", |
| pretrained=True, |
| return_info=True, |
| ) |
| model = model.eval() |
| |
| # synthetic sample with the keys a dataset provides |
| num_points = 8192 |
| sample = { |
| "pos": torch.randn(num_points, 3), |
| "color": torch.rand(num_points, 3) * 255, |
| "intensity": torch.rand(num_points, 1), |
| "segment": torch.zeros(num_points, dtype=torch.long), |
| } |
| data = info["transform"](sample) |
| data = collate([data]) |
| |
| with torch.no_grad(): |
| logits = model(data.get("x"), data["pos"], data["batch"]) |
| ``` |
|
|
| ## Feature extraction |
|
|
| ```python |
| with torch.no_grad(): |
| features = model.forward_features(data.get("x"), data["pos"], data["batch"]) |
| |
| model.reset_classifier(num_classes=0) |
| with torch.no_grad(): |
| features = model(data.get("x"), data["pos"], data["batch"]) # (N, 32) |
| ``` |
|
|
| ## Citation |
|
|
| ```bibtex |
| @inproceedings{hu2020randlanet, |
| title = {RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds}, |
| author = {Qingyong Hu and Bo Yang and Linhai Xie and Stefano Rosa and Yulan Guo and Zhihua Wang and Niki Trigoni and Andrew Markham}, |
| booktitle = {CVPR}, |
| year = {2020} |
| } |
| |
| @article{hackel2017semantic3d, |
| title = {Semantic3D.net: A new Large-scale Point Cloud Classification Benchmark}, |
| author = {Timo Hackel and Nikolay Savinov and Lubor Ladicky and Jan D. Wegner and Konrad Schindler and Marc Pollefeys}, |
| journal = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences}, |
| volume = {IV-1/W1}, |
| year = {2017} |
| } |
| |
| @software{dujardin2026pytorchpointcloud, |
| author = {Arthur Dujardin}, |
| title = {PyTorch PointCloud}, |
| year = {2026}, |
| doi = {10.5281/zenodo.22159632}, |
| url = {https://github.com/arthurdjn/pytorch-pointcloud}, |
| } |
| ``` |
|
|
