feat: Added postprocessing for retinaface

This commit is contained in:
uttarayan21
2025-08-04 19:27:45 +05:30
parent f55f0ab089
commit df5584d797
8 changed files with 235 additions and 21 deletions

3
Cargo.lock generated
View File

@@ -255,6 +255,8 @@ dependencies = [
"nalgebra", "nalgebra",
"ndarray 0.16.1", "ndarray 0.16.1",
"num", "num",
"simba",
"thiserror 2.0.12",
] ]
[[package]] [[package]]
@@ -498,6 +500,7 @@ dependencies = [
"bounding-box", "bounding-box",
"clap", "clap",
"clap_complete", "clap_complete",
"color",
"error-stack", "error-stack",
"fast_image_resize", "fast_image_resize",
"image", "image",

View File

@@ -50,6 +50,7 @@ mnn = { workspace = true }
mnn-bridge = { workspace = true } mnn-bridge = { workspace = true }
mnn-sync = { workspace = true } mnn-sync = { workspace = true }
bounding-box = { version = "0.1.0", path = "bounding-box" } bounding-box = { version = "0.1.0", path = "bounding-box" }
color = "0.3.1"
[profile.release] [profile.release]
debug = true debug = true

View File

@@ -9,6 +9,8 @@ itertools = "0.14.0"
nalgebra = "0.33.2" nalgebra = "0.33.2"
ndarray = { version = "0.16.1", optional = true } ndarray = { version = "0.16.1", optional = true }
num = "0.4.3" num = "0.4.3"
simba = "0.9.0"
thiserror = "2.0.12"
[features] [features]
ndarray = ["dep:ndarray"] ndarray = ["dep:ndarray"]

View File

@@ -1,39 +1,53 @@
use crate::*; use crate::*;
pub use color::Rgba8; pub use color::Rgba8;
use ndarray::{Array3, ArrayViewMut3}; use ndarray::{Array1, Array3, ArrayViewMut3};
pub trait Draw<T> { pub trait Draw<T> {
fn draw(&mut self, item: T, color: color::Rgba8, thickness: usize); fn draw(&mut self, item: T, color: color::Rgba8, thickness: usize);
} }
// impl<T: Drawable<Self>> Draw<T> for Array3<u8> { impl Draw<Aabb2<usize>> for Array3<u8> {
// fn draw(&self, item: T, color: color::Rgba8, thickness: usize) { fn draw(&mut self, item: Aabb2<usize>, color: color::Rgba8, thickness: usize) {
// item.draw(&self, color, thickness); item.draw(self, color, thickness)
// } }
// } }
pub trait Drawable<Canvas, T> { pub trait Drawable<Canvas> {
fn draw(&self, canvas: &mut Canvas, color: color::Rgba8, thickness: T); fn draw(&self, canvas: &mut Canvas, color: color::Rgba8, thickness: usize);
} }
/// Implementing Drawable for Aabb2 with Array3<u8> as the canvas type /// Implementing Drawable for Aabb2 with Array3<u8> as the canvas type
/// Assuming Array3<u8> is a 3D array representing an image with RGB/RGBA channels /// Assuming Array3<u8> is a 3D array representing an image with RGB/RGBA channels
impl<T> Drawable<ArrayViewMut3<'_, u8>, T> for Aabb2<T> impl Drawable<ArrayViewMut3<'_, u8>> for Aabb2<usize> {
where fn draw(&self, canvas: &mut ArrayViewMut3<u8>, color: color::Rgba8, thickness: usize) {
T: Num + core::ops::SubAssign + core::ops::AddAssign + core::ops::DivAssign,
T: PartialOrd,
{
fn draw(&self, canvas: &mut ArrayViewMut3<u8>, color: color::Rgba8, thickness: T) {
use itertools::Itertools; use itertools::Itertools;
let (height, width, channels) = canvas.dim(); // let (height, width, channels) = canvas.dim();
let color = Array1::from_vec(vec![color.r, color.g, color.b, color.a]);
self.corners() self.corners()
.iter() .iter()
.zip(self.padding(thickness).corners()) .zip(self.padding(thickness).corners())
.tuple_windows() .tuple_windows()
.for_each(|((a, b), (c, d))| { .for_each(|((a, b), (c, d))| {
let bbox = Aabb2::from_vertices([*a, b, *c, d]); let bbox = Aabb2::from_vertices([*a, b, *c, d]).expect("Invalid bounding box");
todo!(); use crate::roi::RoiMut;
let mut out = canvas.roi_mut(bbox).expect("Failed to get ROI");
out.lanes_mut(ndarray::Axis(2))
.into_iter()
.for_each(|mut pixel| {
pixel.assign(&color);
});
}); });
} }
} }
impl Drawable<Array3<u8>> for Aabb2<usize> {
fn draw(&self, canvas: &mut Array3<u8>, color: color::Rgba8, thickness: usize) {
use itertools::Itertools;
// let (height, width, channels) = canvas.dim();
let color = Array1::from_vec(vec![color.r, color.g, color.b, color.a]);
let pixel_size = canvas.dim().2;
let color = color.slice(ndarray::s![..pixel_size]);
let [x1y1, x2y1, x1y2, x2y2] = self.corners();
}
}

View File

@@ -1,5 +1,6 @@
pub mod draw; pub mod draw;
pub mod nms; pub mod nms;
pub mod roi;
use nalgebra::{Point, Point2, Point3, SVector}; use nalgebra::{Point, Point2, Point3, SVector};
pub trait Num: num::Num + Copy + core::fmt::Debug + 'static {} pub trait Num: num::Num + Copy + core::fmt::Debug + 'static {}
@@ -31,6 +32,8 @@ impl<T: Num, const D: usize> AxisAlignedBoundingBox<T, D> {
Self::new(point1, SVector::from(size)) Self::new(point1, SVector::from(size))
} }
/// Only considers the points closest and furthest from origin
/// Points which are rotated along in the z axis (in 2d) are not considered
pub fn from_vertices(points: [Point<T, D>; 4]) -> Option<Self> pub fn from_vertices(points: [Point<T, D>; 4]) -> Option<Self>
where where
T: core::ops::SubAssign, T: core::ops::SubAssign,
@@ -182,9 +185,51 @@ impl<T: Num, const D: usize> AxisAlignedBoundingBox<T, D> {
Point::from(max), Point::from(max),
)) ))
} }
pub fn denormalize(&self, factor: nalgebra::SVector<T, D>) -> Self
where
T: core::ops::MulAssign,
T: core::ops::AddAssign,
// nalgebra::constraint::ShapeConstraint:
// nalgebra::constraint::DimEq<nalgebra::Const<D>, nalgebra::Const<D>>,
{
Self {
point: (self.point.coords.component_mul(&factor)).into(),
size: self.size.component_mul(&factor),
}
}
pub fn cast<T2>(&self) -> Option<Aabb<T2, D>>
where
// T: num::NumCast,
T2: Num + simba::scalar::SubsetOf<T>,
{
Some(Aabb {
point: Point::from(self.point.coords.try_cast::<T2>()?),
size: self.size.try_cast::<T2>()?,
})
}
// pub fn as_<T2>(&self) -> Option<Aabb<T2, D>>
// where
// T2: Num + simba::scalar::SubsetOf<T>,
// {
// Some(Aabb {
// point: Point::from(self.point.coords.as_()),
// size: self.size.as_(),
// })
// }
} }
impl<T: Num> Aabb2<T> { impl<T: Num> Aabb2<T> {
pub fn from_x1y1x2y2(x1: T, x2: T, y1: T, y2: T) -> Self
where
T: core::ops::SubAssign,
{
let point1 = Point2::new(x1, y1);
let point2 = Point2::new(x2, y2);
Self::from_min_max_vertices(point1, point2)
}
pub fn new_2d(point1: Point2<T>, point2: Point2<T>) -> Self pub fn new_2d(point1: Point2<T>, point2: Point2<T>) -> Self
where where
T: core::ops::SubAssign, T: core::ops::SubAssign,
@@ -217,6 +262,28 @@ impl<T: Num> Aabb2<T> {
Point2::new(self.point.x, self.point.y + self.size.y) Point2::new(self.point.x, self.point.y + self.size.y)
} }
pub fn x1(&self) -> T {
self.point.x
}
pub fn y1(&self) -> T {
self.point.y
}
pub fn x2(&self) -> T
where
T: core::ops::AddAssign,
{
self.point.x + self.size.x
}
pub fn y2(&self) -> T
where
T: core::ops::AddAssign,
{
self.point.y + self.size.y
}
pub fn corners(&self) -> [Point2<T>; 4] pub fn corners(&self) -> [Point2<T>; 4]
where where
T: core::ops::AddAssign, T: core::ops::AddAssign,

97
bounding-box/src/roi.rs Normal file
View File

@@ -0,0 +1,97 @@
use crate::*;
use ndarray::{Array3, ArrayView3, ArrayViewMut3};
/// A trait that extracts a region of interest from an image
pub trait Roi<'a, Output> {
type Error;
fn roi(&'a self, aabb: Aabb2<usize>) -> Result<Output, Self::Error>;
}
pub trait RoiMut<'a, Output> {
type Error;
fn roi_mut(&'a mut self, aabb: Aabb2<usize>) -> Result<Output, Self::Error>;
}
#[derive(thiserror::Error, Debug, Copy, Clone)]
pub enum RoiError {
#[error("Region of intereset is out of bounds")]
RoiOutOfBounds,
}
impl<'a, T: Num> Roi<'a, ArrayView3<'a, T>> for Array3<T> {
type Error = RoiError;
fn roi(&'a self, aabb: Aabb2<usize>) -> Result<ArrayView3<'a, T>, Self::Error> {
let x1 = aabb.x1();
let x2 = aabb.x2();
let y1 = aabb.y1();
let y2 = aabb.y2();
if x1 >= x2 || y1 >= y2 || x2 > self.shape()[1] || y2 > self.shape()[0] {
return Err(RoiError::RoiOutOfBounds);
}
Ok(self.slice(ndarray::s![y1..y2, x1..x2, ..]))
}
}
impl<'a, T: Num> RoiMut<'a, ArrayViewMut3<'a, T>> for Array3<T> {
type Error = RoiError;
fn roi_mut(&'a mut self, aabb: Aabb2<usize>) -> Result<ArrayViewMut3<'a, T>, Self::Error> {
let x1 = aabb.x1();
let x2 = aabb.x2();
let y1 = aabb.y1();
let y2 = aabb.y2();
if x1 >= x2 || y1 >= y2 || x2 > self.shape()[1] || y2 > self.shape()[0] {
return Err(RoiError::RoiOutOfBounds);
}
Ok(self.slice_mut(ndarray::s![y1..y2, x1..x2, ..]))
}
}
impl<'a, 'b, T: Num> Roi<'a, ArrayView3<'b, T>> for ArrayView3<'b, T> {
type Error = RoiError;
fn roi(&'a self, aabb: Aabb2<usize>) -> Result<ArrayView3<'b, T>, Self::Error> {
let x1 = aabb.x1();
let x2 = aabb.x2();
let y1 = aabb.y1();
let y2 = aabb.y2();
if x1 >= x2 || y1 >= y2 || x2 > self.shape()[1] || y2 > self.shape()[0] {
return Err(RoiError::RoiOutOfBounds);
}
Ok(self.slice_move(ndarray::s![y1..y2, x1..x2, ..]))
}
}
// impl<'a, 'b, T: Num> Roi<'a, ArrayViewMut3<'b, T>> for ArrayViewMut3<'b, T> {
// type Error = RoiError;
// fn roi(&'a self, aabb: Aabb2<usize>) -> Result<ArrayViewMut3<'b, T>, Self::Error> {
// let x1 = aabb.x1();
// let x2 = aabb.x2();
// let y1 = aabb.y1();
// let y2 = aabb.y2();
// if x1 >= x2 || y1 >= y2 || x2 > self.shape()[1] || y2 > self.shape()[0] {
// return Err(RoiError::RoiOutOfBounds);
// }
// Ok(self.slice(ndarray::s![y1..y2, x1..x2, ..]))
// }
// }
impl<'a, 'b: 'a, T: Num> RoiMut<'a, ArrayViewMut3<'a, T>> for ArrayViewMut3<'b, T> {
type Error = RoiError;
fn roi_mut(&'a mut self, aabb: Aabb2<usize>) -> Result<ArrayViewMut3<'a, T>, Self::Error> {
let x1 = aabb.x1();
let x2 = aabb.x2();
let y1 = aabb.y1();
let y2 = aabb.y2();
if x1 >= x2 || y1 >= y2 || x2 > self.shape()[1] || y2 > self.shape()[0] {
return Err(RoiError::RoiOutOfBounds);
}
let out: ArrayViewMut3<'a, T> = self.slice_mut(ndarray::s![y1..y2, x1..x2, ..]);
Ok(out)
}
}
#[test]
pub fn reborrow_test() {
let ndarray = ndarray::Array::from_shape_vec((5, 5, 5), vec![33; 5 * 5 * 5]).unwrap();
let aabb = Aabb2::from_x1y1x2y2(2, 3, 4, 5);
let y = {
let view = ndarray.view();
view.roi(aabb).unwrap()
};
dbg!(y);
}

View File

@@ -45,6 +45,8 @@ pub struct Detect {
pub model: Option<PathBuf>, pub model: Option<PathBuf>,
#[clap(short = 'M', long, default_value = "retina-face")] #[clap(short = 'M', long, default_value = "retina-face")]
pub model_type: Models, pub model_type: Models,
#[clap(short, long)]
pub output: Option<PathBuf>,
pub image: PathBuf, pub image: PathBuf,
} }

View File

@@ -19,17 +19,45 @@ pub fn main() -> Result<()> {
.attach_printable("Failed to create face detection model")?; .attach_printable("Failed to create face detection model")?;
let image = image::open(detect.image).change_context(Error)?; let image = image::open(detect.image).change_context(Error)?;
let image = image.into_rgb8(); let image = image.into_rgb8();
let array = image let mut array = image
.into_ndarray() .into_ndarray()
.change_context(errors::Error) .change_context(errors::Error)
.attach_printable("Failed to convert image to ndarray")?; .attach_printable("Failed to convert image to ndarray")?;
let output = model let output = model
.detect_faces(array) .detect_faces(array.clone())
.change_context(errors::Error) .change_context(errors::Error)
.attach_printable("Failed to detect faces")?; .attach_printable("Failed to detect faces")?;
// output.print(20); // output.print(20);
let aabbs = output.postprocess(Default::default()); let aabbs = output
dbg!(aabbs); .postprocess(Default::default())
.change_context(errors::Error)
.attach_printable("Failed to attach context")?;
for bbox in aabbs {
println!("Detected face: {:?}", bbox);
use bounding_box::draw::*;
let bbox = bbox
.denormalize(nalgebra::SVector::<f32, 2>::new(
array.shape()[1] as f32,
array.shape()[0] as f32,
))
.cast()
.ok_or(errors::Error)
.attach_printable("Failed to cast f32 to usize")?;
dbg!(bbox);
array.draw(bbox, color::palette::css::GREEN_YELLOW.to_rgba8(), 20);
break;
}
let v = array.view();
if let Some(output) = detect.output {
let image: image::RgbImage = v
.to_image()
.change_context(errors::Error)
.attach_printable("Failed to convert ndarray to image")?;
image
.save(output)
.change_context(errors::Error)
.attach_printable("Failed to save output image")?;
}
} }
cli::SubCommand::List(list) => { cli::SubCommand::List(list) => {
println!("List: {:?}", list); println!("List: {:?}", list);