Skip to content

PointCNN

PointCNN classification and segmentation models.

First page of PointCNN: Convolution On X-Transformed Points

1801.07791 · January 2018

Classes:

  • PointCNNIntermediate –

    Per-stage encoder features and the point cloud they live on, consumed as decoder skips.

  • PointCNNEncoderBlock –

    Optional FPS downsampling followed by an XConv over the kNN graph of the sampled points.

  • PointCNNDecoderBlock –

    Upsamples features to the skip resolution with an XConv, then fuses them with the skip features via an MLP.

  • PointCNNEncoder –

    Stack of PointCNNEncoderBlock units that progressively decimate the cloud with FPS.

  • PointCNNDecoder –

    Stack of PointCNNDecoderBlock units that walk the encoder intermediates back to full resolution.

  • PointCNNClassification –

    Classification model as described in the paper

  • PointCNNSegmentation –

    Segmentation model as described in the paper

PointCNNIntermediate

Bases: NamedTuple

Per-stage encoder features and the point cloud they live on, consumed as decoder skips.

PointCNNEncoderBlock 

PointCNNEncoderBlock(
    in_channels: int,
    out_channels: int,
    spatial_dim: int,
    kernel_size: int,
    hidden_channels: Optional[int] = None,
    dilation: int = 1,
    bias: bool = True,
    act: Union[str, Callable, None] = "relu",
    act_kwargs: Optional[Dict[str, Any]] = None,
    downsample: Optional[
        Callable[[Tensor, Tensor], Tensor]
    ] = None,
)

Bases: Module

Optional FPS downsampling followed by an XConv over the kNN graph of the sampled points.

PointCNNDecoderBlock 

PointCNNDecoderBlock(
    in_channels: int,
    skip_channels: int,
    out_channels: int,
    spatial_dim: int,
    kernel_size: int,
    hidden_channels: Optional[int] = None,
    dilation: int = 1,
    bias: bool = True,
    act: Union[str, Callable, None] = "relu",
    act_kwargs: Optional[Dict[str, Any]] = None,
    act_first: bool = False,
    norm: Union[str, Callable, None] = "batch_norm",
    norm_kwargs: Optional[Dict[str, Any]] = None,
)

Bases: Module

Upsamples features to the skip resolution with an XConv, then fuses them with the skip features via an MLP.

PointCNNEncoder 

PointCNNEncoder(
    channels: Sequence[int],
    kernel_sizes: Sequence[int],
    spatial_dim: int,
    ratios: Sequence[float],
    hidden_channels: Optional[
        Union[int, Sequence[int]]
    ] = None,
    dilations: Sequence[int] = (1, 1, 1, 1),
    bias: bool = True,
    act: Union[str, Callable, None] = "relu",
    act_kwargs: Optional[Dict[str, Any]] = None,
)

Bases: Module

Stack of PointCNNEncoderBlock units that progressively decimate the cloud with FPS.

A stage with a ratio of 0 keeps every point and only transforms features. When return_intermediates=True is passed to forward, the pre-downsampling features of each decimating stage are returned in coarse-to-fine order for PointCNNDecoder.

PointCNNDecoder 

PointCNNDecoder(
    channels: Sequence[int],
    skip_channels: Sequence[int],
    kernel_sizes: Sequence[int],
    spatial_dim: int,
    hidden_channels: Optional[
        Union[int, Sequence[int]]
    ] = None,
    dilations: Union[int, Sequence[int]] = 1,
    bias: bool = True,
    act: Union[str, Callable, None] = "relu",
    act_kwargs: Optional[Dict[str, Any]] = None,
)

Bases: Module

Stack of PointCNNDecoderBlock units that walk the encoder intermediates back to full resolution.

PointCNNClassification 

PointCNNClassification(
    in_channels: int,
    num_classes: int,
    *,
    spatial_dim: int = 3,
    channels: Sequence[int],
    kernel_sizes: Sequence[int],
    ratios: Sequence[float],
    hidden_channels: Optional[
        Union[int, Sequence[int]]
    ] = None,
    dilations: Sequence[int] = (1, 1, 1, 1),
    bias: bool = True,
    act: Union[str, Callable, None] = "relu",
    act_kwargs: Optional[Dict[str, Any]] = None,
    act_first: bool = False,
    norm: Union[str, Callable, None] = "batch_norm",
    norm_kwargs: Optional[Dict[str, Any]] = None,
    dropout: float = 0.0,
    head_channels: Optional[
        Union[int, Sequence[int]]
    ] = None,
    global_pool: PoolLike = "max",
)

Bases: ClassificationModel

Classification model as described in the paper "PointCNN: Convolution On X-Transformed Points" by Yangyan Li, Rui Bu, Mingchao Sun, Wei Wu, Xinhan Di, Baoquan Chen.

This classification model consists of a encoder of XConv layers and FPS downsampling layers, and a MLP classification head.

Methods:

Attributes:

  • num_features (int) –

    Feature dimension \(C\) of the features entering the head.

num_features property 

num_features: int

Feature dimension \(C\) of the features entering the head.

configure_encoder 

configure_encoder() -> Module

Build the PointCNNEncoder backbone.

PointCNNSegmentation 

PointCNNSegmentation(
    in_channels: int,
    num_classes: int,
    *,
    spatial_dim: int = 3,
    channels: Sequence[int],
    hidden_channels: Optional[
        Union[int, Sequence[int]]
    ] = None,
    kernel_sizes: Sequence[int],
    dilations: Sequence[int] = (1, 1, 1, 1),
    ratios: Sequence[float],
    act: Union[str, Callable, None] = "relu",
    act_kwargs: Optional[Dict[str, Any]] = None,
    act_first: bool = False,
    norm: Union[str, Callable, None] = "batch_norm",
    norm_kwargs: Optional[Dict[str, Any]] = None,
    bias: bool = True,
    dropout: float = 0.0,
    head_channels: Optional[
        Union[int, Sequence[int]]
    ] = None,
)

Bases: SegmentationModel

Segmentation model as described in the paper "PointCNN: Convolution On X-Transformed Points" by Yangyan Li, Rui Bu, Mingchao Sun, Wei Wu, Xinhan Di, Baoquan Chen.

An encoder of XConv layers with FPS downsampling, a decoder of XConv layers upsampling back to the skip resolutions, and a per-point MLP head.

Methods:

  • configure_encoder –

    Build the PointCNNEncoder backbone.

  • configure_decoder –

    Build the PointCNNDecoder, mirroring the decimating encoder stages in reverse.

Attributes:

  • num_features (int) –

    Feature dimension \(C\) of the decoder output.

num_features property 

num_features: int

Feature dimension \(C\) of the decoder output.

configure_encoder 

configure_encoder() -> PointCNNEncoder

Build the PointCNNEncoder backbone.

configure_decoder 

configure_decoder() -> PointCNNDecoder

Build the PointCNNDecoder, mirroring the decimating encoder stages in reverse.