407 lines
14 KiB
Rust
407 lines
14 KiB
Rust
use crate::errors::*;
|
|
use bounding_box::{Aabb2, nms::nms};
|
|
use error_stack::ResultExt;
|
|
use mnn_bridge::ndarray::*;
|
|
use nalgebra::{Point2, Vector2};
|
|
use ndarray_resize::NdFir;
|
|
use std::path::Path;
|
|
|
|
/// Configuration for face detection postprocessing
|
|
#[derive(Debug, Clone, PartialEq)]
|
|
pub struct FaceDetectionConfig {
|
|
/// Minimum confidence to keep a detection
|
|
pub threshold: f32,
|
|
/// NMS threshold for suppressing overlapping boxes
|
|
pub nms_threshold: f32,
|
|
/// Variances for bounding box decoding
|
|
pub variances: [f32; 2],
|
|
/// The step size (stride) for each feature map
|
|
pub steps: Vec<usize>,
|
|
/// The minimum anchor sizes for each feature map
|
|
pub min_sizes: Vec<Vec<usize>>,
|
|
/// Whether to clip bounding boxes to the image dimensions
|
|
pub clamp: bool,
|
|
/// Input image width (used for anchor generation)
|
|
pub input_width: usize,
|
|
/// Input image height (used for anchor generation)
|
|
pub input_height: usize,
|
|
}
|
|
|
|
impl FaceDetectionConfig {
|
|
pub fn with_threshold(mut self, threshold: f32) -> Self {
|
|
self.threshold = threshold;
|
|
self
|
|
}
|
|
pub fn with_nms_threshold(mut self, nms_threshold: f32) -> Self {
|
|
self.nms_threshold = nms_threshold;
|
|
self
|
|
}
|
|
pub fn with_variances(mut self, variances: [f32; 2]) -> Self {
|
|
self.variances = variances;
|
|
self
|
|
}
|
|
pub fn with_steps(mut self, steps: Vec<usize>) -> Self {
|
|
self.steps = steps;
|
|
self
|
|
}
|
|
pub fn with_min_sizes(mut self, min_sizes: Vec<Vec<usize>>) -> Self {
|
|
self.min_sizes = min_sizes;
|
|
self
|
|
}
|
|
pub fn with_clip(mut self, clip: bool) -> Self {
|
|
self.clamp = clip;
|
|
self
|
|
}
|
|
pub fn with_input_width(mut self, input_width: usize) -> Self {
|
|
self.input_width = input_width;
|
|
self
|
|
}
|
|
pub fn with_input_height(mut self, input_height: usize) -> Self {
|
|
self.input_height = input_height;
|
|
self
|
|
}
|
|
}
|
|
|
|
impl Default for FaceDetectionConfig {
|
|
fn default() -> Self {
|
|
Self {
|
|
threshold: 0.5,
|
|
nms_threshold: 0.4,
|
|
variances: [0.1, 0.2],
|
|
steps: vec![8, 16, 32],
|
|
min_sizes: vec![vec![16, 32], vec![64, 128], vec![256, 512]],
|
|
clamp: true,
|
|
input_width: 1024,
|
|
input_height: 1024,
|
|
}
|
|
}
|
|
}
|
|
|
|
#[derive(Debug)]
|
|
pub struct FaceDetection {
|
|
handle: mnn_sync::SessionHandle,
|
|
}
|
|
|
|
#[derive(Debug, Clone, PartialEq)]
|
|
pub struct FaceDetectionModelOutput {
|
|
pub bbox: ndarray::Array3<f32>,
|
|
pub confidence: ndarray::Array3<f32>,
|
|
pub landmark: ndarray::Array3<f32>,
|
|
}
|
|
|
|
/// Represents the 5 facial landmarks detected by RetinaFace
|
|
#[derive(Debug, Copy, Clone, PartialEq)]
|
|
pub struct FaceLandmarks {
|
|
pub left_eye: Point2<f32>,
|
|
pub right_eye: Point2<f32>,
|
|
pub nose: Point2<f32>,
|
|
pub left_mouth: Point2<f32>,
|
|
pub right_mouth: Point2<f32>,
|
|
}
|
|
|
|
#[derive(Debug, Clone, PartialEq)]
|
|
pub struct FaceDetectionProcessedOutput {
|
|
pub bbox: Vec<Aabb2<f32>>,
|
|
pub confidence: Vec<f32>,
|
|
pub landmarks: Vec<FaceLandmarks>,
|
|
}
|
|
|
|
#[derive(Debug, Clone, PartialEq)]
|
|
pub struct FaceDetectionOutput {
|
|
pub bbox: Vec<Aabb2<usize>>,
|
|
pub confidence: Vec<f32>,
|
|
pub landmark: Vec<FaceLandmarks>,
|
|
}
|
|
|
|
fn generate_anchors(config: &FaceDetectionConfig) -> ndarray::Array2<f32> {
|
|
let mut anchors = Vec::new();
|
|
let feature_maps: Vec<(usize, usize)> = config
|
|
.steps
|
|
.iter()
|
|
.map(|&step| {
|
|
(
|
|
(config.input_height as f32 / step as f32).ceil() as usize,
|
|
(config.input_width as f32 / step as f32).ceil() as usize,
|
|
)
|
|
})
|
|
.collect();
|
|
|
|
for (k, f) in feature_maps.iter().enumerate() {
|
|
let min_sizes = &config.min_sizes[k];
|
|
for i in 0..f.0 {
|
|
for j in 0..f.1 {
|
|
for &min_size in min_sizes {
|
|
let s_kx = min_size as f32 / config.input_width as f32;
|
|
let s_ky = min_size as f32 / config.input_height as f32;
|
|
let dense_cx =
|
|
(j as f32 + 0.5) * config.steps[k] as f32 / config.input_width as f32;
|
|
let dense_cy =
|
|
(i as f32 + 0.5) * config.steps[k] as f32 / config.input_height as f32;
|
|
anchors.push([
|
|
dense_cx - s_kx / 2.,
|
|
dense_cy - s_ky / 2.,
|
|
dense_cx + s_kx / 2.,
|
|
dense_cy + s_ky / 2.,
|
|
]);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
ndarray::Array2::from_shape_vec((anchors.len(), 4), anchors.into_iter().flatten().collect())
|
|
.unwrap()
|
|
}
|
|
|
|
impl FaceDetectionModelOutput {
|
|
pub fn postprocess(self, config: &FaceDetectionConfig) -> Result<FaceDetectionProcessedOutput> {
|
|
use ndarray::s;
|
|
|
|
let priors = generate_anchors(config);
|
|
|
|
let scores = self.confidence.slice(s![0, .., 1]);
|
|
let boxes = self.bbox.slice(s![0, .., ..]);
|
|
let landmarks_raw = self.landmark.slice(s![0, .., ..]);
|
|
|
|
let mut decoded_boxes = Vec::new();
|
|
let mut decoded_landmarks = Vec::new();
|
|
let mut confidences = Vec::new();
|
|
|
|
for i in 0..priors.shape()[0] {
|
|
if scores[i] > config.threshold {
|
|
let prior = priors.row(i);
|
|
let loc = boxes.row(i);
|
|
let landm = landmarks_raw.row(i);
|
|
|
|
// Decode bounding box
|
|
let prior_cx = (prior[0] + prior[2]) / 2.0;
|
|
let prior_cy = (prior[1] + prior[3]) / 2.0;
|
|
let prior_w = prior[2] - prior[0];
|
|
let prior_h = prior[3] - prior[1];
|
|
|
|
let var = config.variances;
|
|
let cx = prior_cx + loc[0] * var[0] * prior_w;
|
|
let cy = prior_cy + loc[1] * var[0] * prior_h;
|
|
let w = prior_w * (loc[2] * var[1]).exp();
|
|
let h = prior_h * (loc[3] * var[1]).exp();
|
|
|
|
let xmin = cx - w / 2.0;
|
|
let ymin = cy - h / 2.0;
|
|
let xmax = cx + w / 2.0;
|
|
let ymax = cy + h / 2.0;
|
|
|
|
let mut bbox =
|
|
Aabb2::from_min_max_vertices(Point2::new(xmin, ymin), Point2::new(xmax, ymax));
|
|
if config.clamp {
|
|
bbox.component_clamp(0.0, 1.0);
|
|
}
|
|
decoded_boxes.push(bbox);
|
|
|
|
// Decode landmarks
|
|
let mut points = [Point2::new(0.0, 0.0); 5];
|
|
for j in 0..5 {
|
|
points[j].x = prior_cx + landm[j * 2] * var[0] * prior_w;
|
|
points[j].y = prior_cy + landm[j * 2 + 1] * var[0] * prior_h;
|
|
}
|
|
let landmarks = FaceLandmarks {
|
|
left_eye: points[0],
|
|
right_eye: points[1],
|
|
nose: points[2],
|
|
left_mouth: points[3],
|
|
right_mouth: points[4],
|
|
};
|
|
decoded_landmarks.push(landmarks);
|
|
confidences.push(scores[i]);
|
|
}
|
|
}
|
|
|
|
Ok(FaceDetectionProcessedOutput {
|
|
bbox: decoded_boxes,
|
|
confidence: confidences,
|
|
landmarks: decoded_landmarks,
|
|
})
|
|
}
|
|
}
|
|
|
|
impl FaceDetectionModelOutput {
|
|
pub fn print(&self, limit: usize) {
|
|
tracing::info!("Detected {} faces", self.bbox.shape()[1]);
|
|
|
|
for (bbox, confidence) in self
|
|
.bbox
|
|
.clone()
|
|
.remove_axis(ndarray::Axis(0))
|
|
.axis_iter(ndarray::Axis(0))
|
|
.zip(
|
|
self.confidence
|
|
.clone()
|
|
.remove_axis(ndarray::Axis(0))
|
|
.axis_iter(ndarray::Axis(0))
|
|
.map(|c| c[1]),
|
|
)
|
|
.filter(|(_, c)| *c > 0.1)
|
|
.take(limit)
|
|
{
|
|
tracing::info!("Face BBox: {:?}, Confidence: {:.2}", bbox, confidence);
|
|
}
|
|
}
|
|
}
|
|
|
|
impl FaceDetection {
|
|
pub fn new(path: impl AsRef<Path>) -> Result<Self> {
|
|
let model = std::fs::read(path)
|
|
.change_context(Error)
|
|
.attach_printable("Failed to read model file")?;
|
|
Self::new_from_bytes(&model)
|
|
}
|
|
|
|
pub fn new_from_bytes(model: &[u8]) -> Result<Self> {
|
|
tracing::info!("Loading face detection model from bytes");
|
|
let mut model = mnn::Interpreter::from_bytes(model)
|
|
.map_err(|e| e.into_inner())
|
|
.change_context(Error)
|
|
.attach_printable("Failed to load model from bytes")?;
|
|
model.set_session_mode(mnn::SessionMode::Release);
|
|
let bc = mnn::BackendConfig::default().with_memory_mode(mnn::MemoryMode::High);
|
|
let sc = mnn::ScheduleConfig::new()
|
|
.with_type(mnn::ForwardType::CPU)
|
|
.with_backend_config(bc);
|
|
tracing::info!("Creating session handle for face detection model");
|
|
let handle = mnn_sync::SessionHandle::new(model, sc)
|
|
.change_context(Error)
|
|
.attach_printable("Failed to create session handle")?;
|
|
Ok(FaceDetection { handle })
|
|
}
|
|
|
|
pub fn detect_faces(
|
|
&self,
|
|
image: ndarray::ArrayView3<u8>,
|
|
config: FaceDetectionConfig,
|
|
) -> Result<FaceDetectionOutput> {
|
|
let (height, width, _channels) = image.dim();
|
|
let output = self
|
|
.run_models(image)
|
|
.change_context(Error)
|
|
.attach_printable("Failed to detect faces")?;
|
|
// denormalize the bounding boxes
|
|
let factor = Vector2::new(width as f32, height as f32);
|
|
let mut processed = output
|
|
.postprocess(&config)
|
|
.attach_printable("Failed to postprocess")?;
|
|
|
|
use itertools::Itertools;
|
|
let (boxes, scores, landmarks): (Vec<_>, Vec<_>, Vec<_>) = processed
|
|
.bbox
|
|
.iter()
|
|
.cloned()
|
|
.zip(processed.confidence.iter().cloned())
|
|
.zip(processed.landmarks.iter().cloned())
|
|
.sorted_by_key(|((_, score), _)| ordered_float::OrderedFloat(*score))
|
|
.map(|((b, s), l)| (b, s, l))
|
|
.multiunzip();
|
|
|
|
let keep_indices =
|
|
nms(&boxes, &scores, config.threshold, config.nms_threshold).change_context(Error)?;
|
|
|
|
let bboxes = boxes
|
|
.into_iter()
|
|
.enumerate()
|
|
.filter(|(i, _)| keep_indices.contains(i))
|
|
.flat_map(|(_, x)| x.denormalize(factor).try_cast::<usize>())
|
|
.collect();
|
|
let confidence = scores
|
|
.into_iter()
|
|
.enumerate()
|
|
.filter(|(i, _)| keep_indices.contains(i))
|
|
.map(|(_, score)| score)
|
|
.collect();
|
|
let landmark = landmarks
|
|
.into_iter()
|
|
.enumerate()
|
|
.filter(|(i, _)| keep_indices.contains(i))
|
|
.map(|(_, score)| score)
|
|
.collect();
|
|
|
|
Ok(FaceDetectionOutput {
|
|
bbox: bboxes,
|
|
confidence,
|
|
landmark,
|
|
})
|
|
}
|
|
|
|
pub fn run_models(&self, image: ndarray::ArrayView3<u8>) -> Result<FaceDetectionModelOutput> {
|
|
#[rustfmt::skip]
|
|
use ::tap::*;
|
|
let output = self
|
|
.handle
|
|
.run(move |sr| {
|
|
let mut resized = image
|
|
.fast_resize(1024, 1024, None)
|
|
.change_context(mnn::ErrorKind::TensorError)?
|
|
.mapv(|f| f as f32)
|
|
.tap_mut(|arr| {
|
|
arr.axis_iter_mut(ndarray::Axis(2))
|
|
.zip([104, 117, 123])
|
|
.for_each(|(mut array, pixel)| {
|
|
let pixel = pixel as f32;
|
|
array.map_inplace(|v| *v -= pixel);
|
|
});
|
|
})
|
|
.permuted_axes((2, 0, 1))
|
|
.insert_axis(ndarray::Axis(0))
|
|
.as_standard_layout()
|
|
.into_owned();
|
|
let tensor = resized
|
|
.as_mnn_tensor_mut()
|
|
.attach_printable("Failed to convert ndarray to mnn tensor")
|
|
.change_context(mnn::error::ErrorKind::TensorError)?;
|
|
tracing::trace!("Image Tensor shape: {:?}", tensor.shape());
|
|
let (intptr, session) = sr.both_mut();
|
|
tracing::trace!("Copying input tensor to host");
|
|
unsafe {
|
|
let mut input = intptr.input_unresized::<f32>(session, "input")?;
|
|
tracing::trace!("Input shape: {:?}", input.shape());
|
|
intptr.resize_tensor_by_nchw::<mnn::View<&mut f32>, _>(
|
|
input.view_mut(),
|
|
1,
|
|
3,
|
|
1024,
|
|
1024,
|
|
);
|
|
}
|
|
intptr.resize_session(session);
|
|
let mut input = intptr.input::<f32>(session, "input")?;
|
|
tracing::trace!("Input shape: {:?}", input.shape());
|
|
input.copy_from_host_tensor(tensor.view())?;
|
|
|
|
tracing::info!("Running face detection session");
|
|
intptr.run_session(&session)?;
|
|
let output_tensor = intptr
|
|
.output::<f32>(&session, "bbox")?
|
|
.create_host_tensor_from_device(true)
|
|
.as_ndarray()
|
|
.to_owned();
|
|
tracing::trace!("Output Bbox: \t\t{:?}", output_tensor.shape());
|
|
let output_confidence = intptr
|
|
.output::<f32>(&session, "confidence")?
|
|
.create_host_tensor_from_device(true)
|
|
.as_ndarray::<ndarray::Ix3>()
|
|
.to_owned();
|
|
tracing::trace!("Output Confidence: \t{:?}", output_confidence.shape());
|
|
let output_landmark = intptr
|
|
.output::<f32>(&session, "landmark")?
|
|
.create_host_tensor_from_device(true)
|
|
.as_ndarray::<ndarray::Ix3>()
|
|
.to_owned();
|
|
tracing::trace!("Output Landmark: \t{:?}", output_landmark.shape());
|
|
Ok(FaceDetectionModelOutput {
|
|
bbox: output_tensor,
|
|
confidence: output_confidence,
|
|
landmark: output_landmark,
|
|
})
|
|
})
|
|
.map_err(|e| e.into_inner())
|
|
.change_context(Error)?;
|
|
Ok(output)
|
|
}
|
|
}
|