OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Point Transformer Pytorch

Point Transformer Pytorch is a Python library implementing a Point Transformer self-attention layer for PyTorch point-cloud models. The layer processes point features together with their 3D positions, accepts an optional mask, and can restrict attention to a configured number of nearest neighbors.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Point Transformer Pytorch — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when a PyTorch point-cloud model needs the documented Point Transformer layer and its feature, position, mask or nearest-neighbor options.

Who it is for

  • PyTorch developers adding attention to point-cloud models.
  • Teams building point-cloud classification or segmentation pipelines that need a reusable layer.

Documented uses

  • Add Point Transformer self-attention to a point-cloud model.
  • Process point features in relation to their 3D positions.
  • Limit attention to a configured number of nearest neighbors.
Before you choose

Requirements & limits

The repository is public and MIT-licensed; installation is documented through pip.

  • The layer expects point features and 3D positions, with dimensions configured when constructing it.
  • A boolean mask is supported in the documented call.
  • num_neighbors provides a nearest-neighbor attention limit.
  • The README warns that this vector attention is more expensive than traditional attention.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install and call the Point Transformer layer

Add the documented layer to a Python point-cloud model and pass it feature, position and mask tensors.

Before you start

  • Python environment with PyTorch available; the example imports torch.
  • Install the package with pip.

Steps

  1. Install with the command below.

    Documented command
    pip install point-transformer-pytorch

  2. Import PointTransformerLayer from point_transformer_pytorch.

  3. Configure the layer with feature dimension and the documented positional and attention MLP settings.

  4. Pass batched feature and 3D position tensors, with an optional boolean mask, to the layer.

Expected outcome

The layer returns a transformed feature tensor, as shown in the README example.

Guide limitations

  • The README documents a layer call, not a complete model, training loop or classification/segmentation pipeline.
  • The documentation notes that this vector attention is more expensive than traditional attention.
Documentation 1 ↗
Source-based practical guide

Limit attention to nearest points

Configure the layer to attend only to a fixed number of nearest points for each point.

Before you start

  • The point-transformer-pytorch package and PyTorch.
  • Batched point features and 3D positions.

Steps

  1. Create PointTransformerLayer with the documented model dimensions.

  2. Set num_neighbors to the desired neighbor count; the README example uses 16.

  3. Pass the feature and position tensors, and an optional mask, to the layer.

Expected outcome

The README describes attention restricted to the configured number of nearest neighbors per point.

Guide limitations

  • The documented example does not provide benchmarking or a broader performance guarantee.
  • The project notes that vector attention is more expensive than traditional attention.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
lucidrains/point-transformer-pytorch · Official homepage title · retained in its source language
How we review sourcesSuggest a correction