Models

torch_points3d.applications.sparseconv3d.SparseConv3d(architecture: str = None, input_nc: int = None, num_layers: int = None, config: omegaconf.DictConfig = None, backend: str = 'minkowski', *args, **kwargs)[source]
Create a Sparse Conv backbone model based on architecture proposed in

https://arxiv.org/abs/1904.08755

Two backends are available at the moment:
architecturestr, optional

Architecture of the model, choose from unet, encoder and decoder

input_ncint, optional

Number of channels for the input

output_ncint, optional

If specified, then we add a fully connected head at the end of the network to provide the requested dimension

num_layersint, optional

Depth of the network

configDictConfig, optional

Custom config, overrides the num_layers and architecture parameters

block:

Type of resnet block, ResBlock by default but can be any of the blocks in modules/SparseConv3d/modules.py

backend:

torchsparse or minkowski

torch_points3d.applications.kpconv.KPConv(architecture: str = None, input_nc: int = None, num_layers: int = None, config: omegaconf.DictConfig = None, *args, **kwargs)[source]

Create a KPConv backbone model based on the architecture proposed in https://arxiv.org/abs/1904.08889

Parameters
  • architecture (str, optional) – Architecture of the model, choose from unet, encoder and decoder

  • input_nc (int, optional) – Number of channels for the input

  • output_nc (int, optional) – If specified, then we add a fully connected head at the end of the network to provide the requested dimension

  • num_layers (int, optional) – Depth of the network

  • in_grid_size (float, optional) – Size of the grid at the entry of the network. It is divided by two at each layer

  • in_feat (int, optional) – Number of channels after the first convolution. Doubles at each layer

  • config (DictConfig, optional) – Custom config, overrides the num_layers and architecture parameters

torch_points3d.applications.pointnet2.PointNet2(architecture: str = None, input_nc: int = None, num_layers: int = None, config: omegaconf.DictConfig = None, multiscale=False, *args, **kwargs)[source]
Create a PointNet2 backbone model based on the architecture proposed in

https://arxiv.org/abs/1706.02413

architecturestr, optional

Architecture of the model, choose from unet, encoder and decoder

input_ncint, optional

Number of channels for the input

output_ncint, optional

If specified, then we add a fully connected head at the end of the network to provide the requested dimension

num_layersint, optional

Depth of the network

configDictConfig, optional

Custom config, overrides the num_layers and architecture parameters

torch_points3d.applications.rsconv.RSConv(architecture: str = None, input_nc: int = None, num_layers: int = None, config: omegaconf.DictConfig = None, *args, **kwargs)[source]

Create a RSConv backbone model based on the architecture proposed in https://arxiv.org/abs/1904.07601

Parameters
  • architecture (str, optional) – Architecture of the model, choose from unet, encoder and decoder

  • input_nc (int, optional) – Number of channels for the input

  • output_nc (int, optional) – If specified, then we add a fully connected head at the end of the network to provide the requested dimension

  • num_layers (int, optional) – Depth of the network

  • config (DictConfig, optional) – Custom config, overrides the num_layers and architecture parameters