oxicuda_vision/pointcloud/mod.rs
1//! Point-cloud neural network primitives.
2//!
3//! Provides:
4//! - **`point_transformer`**: the Point Transformer vector self-attention layer
5//! (Zhao et al. 2021) — subtraction-relation attention over kNN
6//! neighbourhoods with a learned relative-position encoding.
7
8pub mod point_transformer;
9
10pub use point_transformer::{PointAttention, PointTransformerConfig, PointTransformerLayer};