M-Labs/nalgebra
3
0
Fork 1
Code Issues Pull Requests Packages Projects Releases Wiki Activity

simple skeleton compiling

Browse Source
This commit is contained in:
Marc Haubenstock 2019-07-27 18:18:50 +02:00
parent 0384268bd4
commit d46ffe607c
2 changed files with 99 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

64
src/linalg/convolution.rs
View File

@ -4,10 +4,11 @@ use crate::base::allocator::Allocator;
use crate::base::default_allocator::DefaultAllocator;
use crate::base::dimension::{Dim, DimAdd, DimDiff, DimSub, DimSum};
use crate::storage::Storage;
use crate::{zero, RealField, Vector, VectorN, U1};
use crate::{zero, RealField, Vector, VectorN, U1, Scalar, Matrix, DMatrix};
use crate::alga::general::Field;
impl<N: RealField, D1: Dim, S1: Storage<N, D1>> Vector<N, D1, S1> {
/// Returns the convolution of the target vector and a kernel.
/// Returns the convolution of the target vector and a kernel.s
///
/// # Arguments
///
@ -128,3 +129,62 @@ impl<N: RealField, D1: Dim, S1: Storage<N, D1>> Vector<N, D1, S1> {
conv
}
}
// TODO: @Investigate -> Only implemented for DMatrix for now as images are usually DMatrix
impl<N: RealField> DMatrix<N> {
/// Returns the convolution of the target vector and a kernel.
///
/// # Arguments
///
/// * `kernel` - A Matrix with rows > 0 and cols > 0
///
/// # Errors
/// Inputs must satisfy `self.shape() >= kernel.shape() > 0`.
///
pub fn mat_convolve_full<R1, C1, S1>(
&self,
kernel: Matrix<N, R1, C1, S1>, //TODO: Would be nice to have an IsOdd trait. As kernels could be of even size atm
) -> DMatrix<N>
where
R1: Dim,
C1: Dim,
S1: Storage<N, R1, C1>
{
let mat_rows = self.nrows();
let mat_cols = self.ncols();
let ker_rows = kernel.data.shape().0.value();
let ker_cols = kernel.data.shape().1.value();
if ker_rows == 0 || ker_rows > mat_rows || ker_cols == 0|| ker_cols > mat_cols {
panic!(
"convolve_full expects `self.nrows() >= kernel.nrows() > 0 and self.ncols() >= kernel.ncols() > 0 `, \
rows received {} and {} respectively. \
cols received {} and {} respectively.",
mat_rows, ker_rows, mat_cols, ker_cols);
}
let zero = zero::<N>();
let mut conv = DMatrix::from_diagonal_element(mat_cols, mat_rows, zero);
//
// for i in 0..(vec + ker - 1) {
// let u_i = if i > vec { i - ker } else { 0 };
// let u_f = cmp::min(i, vec - 1);
//
// if u_i == u_f {
// conv[i] += self[u_i] * kernel[(i - u_i)];
// } else {
// for u in u_i..(u_f + 1) {
// if i - u < ker {
// conv[i] += self[u] * kernel[(i - u)];
// }
// }
// }
// }
conv
}
//TODO: rest
}

37
tests/linalg/convolution.rs
View File

@ -43,6 +43,43 @@ fn convolve_same_check(){
);
}
//// >>> convolve([1,2,3,4],[1,2],"same")
//// array([ 1, 4, 7, 10])
//#[test]
//fn convolve_same_integers_check(){
// // Static Tests
// let actual_s = Vector4::new(1, 4, 7, 10);
// let expected_s = Vector4::new(1, 2, 3, 4).convolve_same(Vector2::new(1, 2));
//
// assert!(relative_eq!(actual_s, expected_s, epsilon = 1.0e-7));
//
// // Dynamic Tests
// let actual_d = DVector::from_vec(vec![1, 4, 7, 10]);
// let expected_d = DVector::from_vec(vec![1, 2, 3, 4]).convolve_same(DVector::from_vec(vec![1, 2]));
//
// assert!(relative_eq!(actual_d, expected_d, epsilon = 1.0e-7));
//
// // Panic Tests
// // These really only apply to dynamic sized vectors
// assert!(
// panic::catch_unwind(|| {
// DVector::from_vec(vec![1, 2]).convolve_same(DVector::from_vec(vec![1, 2, 3, 4]));
// }).is_err()
// );
//
// assert!(
// panic::catch_unwind(|| {
// DVector::<usize>::from_vec(vec![]).convolve_same(DVector::from_vec(vec![1, 2, 3, 4]));
// }).is_err()
// );
//
// assert!(
// panic::catch_unwind(|| {
// DVector::from_vec(vec![1, 2, 3, 4]).convolve_same(DVector::<usize>::from_vec(vec![]));
// }).is_err()
// );
//}
// >>> convolve([1,2,3,4],[1,2],"full")
// array([ 1, 4, 7, 10, 8])
#[test]