README.md

metadata

license: mit
library_name: torch-pointcloud
tags:
  - point-cloud
  - 3d
  - pytorch
  - torch-pointcloud
  - kpconv
  - segmentation
datasets:
  - s3dis
model-index:
  - name: kpfcnn-base-sm.s3dis.hugues-thomas
    results:
      - task:
          type: point-cloud-segmentation
        dataset:
          name: S3DIS (6-fold)
          type: s3dis
        metrics:
          - name: mIoU
            type: mean_iou
            value: 65.39

Model card for kpfcnn-base-sm.s3dis.hugues-thomas

A KPConv point cloud segmentation model (deformable kernel point convolution). Trained on S3DIS (6-fold).

Model Details

• Model Type: Point cloud semantic segmentation
• Model Stats:
  • Params (M): 23.9
  • Input channels: 5
  • Classes: 13
  • Features: 128
• Dataset: S3DIS (6-fold)
• Metrics: mIoU 65.39 (reference 65.4)
• Paper: KPConv: Flexible and Deformable Convolution for Point Clouds
• Converted from: HuguesTHOMAS/KPConv-PyTorch (MIT)
• Library: torch-pointcloud

Install

pip install torch-pointcloud

Usage

import torch
import torch_pointcloud as tp
from torch_pointcloud.utils.data import collate

model, info = tp.create_model(
    "kpfcnn-base-sm.s3dis.hugues-thomas",
    task="segmentation",
    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,
    "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

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, 128)

Citation

@inproceedings{thomas2019kpconv,
  title   = {KPConv: Flexible and Deformable Convolution for Point Clouds},
  author  = {Hugues Thomas and Charles R. Qi and Jean-Emmanuel Deschaud and Beatriz Marcotegui and François Goulette and Leonidas J. Guibas},
  booktitle = {ICCV},
  year    = {2019}
}

@inproceedings{armeni2016s3dis,
  title     = {{3D} Semantic Parsing of Large-Scale Indoor Spaces},
  author    = {Armeni, Iro and Sener, Ozan and Zamir, Amir R. and Jiang, Helen and Brilakis, Ioannis and Fischer, Martin and Savarese, Silvio},
  booktitle = {CVPR},
  year      = {2016}
}

@software{dujardin2026pytorchpointcloud,
  author  = {Arthur Dujardin},
  title   = {PyTorch PointCloud},
  year    = {2026},
  doi     = {10.5281/zenodo.22159632},
  url     = {https://github.com/arthurdjn/pytorch-pointcloud},
}