Struct FlannMatcher 

pub struct FlannMatcher { /* private fields */ }

FLANN-inspired approximate nearest-neighbour matcher built on a set of KD-trees.

Each tree is built by recursively partitioning the descriptor space along alternating axes. At query time the trees are searched in parallel with a limited number of checks per tree to trade accuracy for speed.

Implementations

impl FlannMatcher

pub fn new() -> Self

Creates a FLANN matcher with reasonable defaults (k=2, trees=5, checks=32).

pub fn with_k(self, k: usize) -> Self

Sets the number of nearest neighbours.

pub fn with_trees(self, trees: usize) -> Self

Sets the number of trees in the forest.

pub fn with_checks(self, checks: usize) -> Self

Sets the maximum number of leaf checks during search.

pub fn find_matches<B: Backend>( &self, desc1: &Tensor<B, 2>, desc2: &Tensor<B, 2>, ) -> Result<Vec<DMatch>>

Finds approximate nearest-neighbour matches between two descriptor matrices.

desc1 is the query matrix of shape [N1, D] and desc2 is the train matrix of shape [N2, D]. The function returns one DMatch per query descriptor.

Trait Implementations

impl Default for FlannMatcher

fn default() -> Self

Returns the “default value” for a type.