Crop -> crop and resize

This commit is contained in:
Arnaud_Cayrol 2026-02-04 21:15:55 +01:00
parent 1afcf2de21
commit 24f7b86ab1
5 changed files with 78 additions and 177 deletions

View file

@ -92,19 +92,18 @@ impl Pipeline {
/// Create the processing pipeline based on configuration. /// Create the processing pipeline based on configuration.
/// ///
/// Only enabled steps are added to the pipeline. Core steps (decode, crop, resize) /// Only enabled steps are added to the pipeline. Core steps (decode, crop_and_resize)
/// are always included. /// are always included.
/// ///
/// Pipeline order: /// Pipeline order:
/// 1. FaceResolutionStep - Validate face size from Immich metadata (if enabled) /// 1. FaceResolutionStep - Validate face size from Immich metadata (if enabled)
/// 2. DecodeImageStep - Load and orient the image (always) /// 2. DecodeImageStep - Load and orient the image (always)
/// 3. CropFaceStep - Extract face region with padding (always) /// 3. CropAndResizeStep - Extract face region with padding and resize to output size (always)
/// 4. BrightnessStep - Filter by luminance on cropped face (if enabled) /// 4. BrightnessStep - Filter by luminance on cropped face (if enabled)
/// 5. HeadPoseStep - Filter non-frontal faces (if enabled) /// 5. HeadPoseStep - Filter non-frontal faces (if enabled)
/// 6. LandmarksStep - Detect 68 facial landmarks (always) /// 6. LandmarksStep - Detect 68 facial landmarks (always)
/// 7. EyeFilterStep - Filter closed eyes by EAR (if enabled) /// 7. EyeFilterStep - Filter closed eyes by EAR (if enabled)
/// 8. AlignmentStep - Align face based on eye positions (always) /// 8. AlignmentStep - Align face based on eye positions and resize to final output (always)
/// 9. ResizeStep - Final resize to output size (always)
pub fn with_steps_from_config(config: &Config) -> Self { pub fn with_steps_from_config(config: &Config) -> Self {
use steps::*; use steps::*;
@ -118,8 +117,8 @@ impl Pipeline {
// Core: Always decode the image // Core: Always decode the image
pipeline.add_step(Box::new(DecodeImageStep)); pipeline.add_step(Box::new(DecodeImageStep));
// Core: Always crop the face // Core: Always crop and resize the face
pipeline.add_step(Box::new(CropFaceStep)); pipeline.add_step(Box::new(CropAndResizeStep));
// Optional: Brightness validation (on cropped face region) // Optional: Brightness validation (on cropped face region)
if config.processing.brightness.enabled { if config.processing.brightness.enabled {
@ -139,12 +138,9 @@ impl Pipeline {
pipeline.add_step(Box::new(EyeFilterStep)); pipeline.add_step(Box::new(EyeFilterStep));
} }
// Core: Always align face based on eye positions // Core: Always align face based on eye positions (also performs final resize)
pipeline.add_step(Box::new(AlignmentStep)); pipeline.add_step(Box::new(AlignmentStep));
// Core: Always resize to output size
pipeline.add_step(Box::new(ResizeStep));
pipeline pipeline
} }
@ -156,13 +152,12 @@ impl Pipeline {
let mut pipeline = Self::new(); let mut pipeline = Self::new();
pipeline.add_step(Box::new(FaceResolutionStep)); pipeline.add_step(Box::new(FaceResolutionStep));
pipeline.add_step(Box::new(DecodeImageStep)); pipeline.add_step(Box::new(DecodeImageStep));
pipeline.add_step(Box::new(CropFaceStep)); pipeline.add_step(Box::new(CropAndResizeStep));
pipeline.add_step(Box::new(BrightnessStep)); pipeline.add_step(Box::new(BrightnessStep));
pipeline.add_step(Box::new(HeadPoseStep)); pipeline.add_step(Box::new(HeadPoseStep));
pipeline.add_step(Box::new(LandmarksStep)); pipeline.add_step(Box::new(LandmarksStep));
pipeline.add_step(Box::new(EyeFilterStep)); pipeline.add_step(Box::new(EyeFilterStep));
pipeline.add_step(Box::new(AlignmentStep)); pipeline.add_step(Box::new(AlignmentStep));
pipeline.add_step(Box::new(ResizeStep));
pipeline pipeline
} }
@ -359,12 +354,11 @@ mod tests {
assert!(ids.contains(&"face_resolution")); assert!(ids.contains(&"face_resolution"));
assert!(ids.contains(&"decode")); assert!(ids.contains(&"decode"));
assert!(ids.contains(&"brightness")); assert!(ids.contains(&"brightness"));
assert!(ids.contains(&"crop")); assert!(ids.contains(&"crop_and_resize"));
assert!(ids.contains(&"head_pose")); assert!(ids.contains(&"head_pose"));
assert!(ids.contains(&"landmarks")); assert!(ids.contains(&"landmarks"));
assert!(ids.contains(&"eye_filter")); assert!(ids.contains(&"eye_filter"));
assert!(ids.contains(&"alignment")); assert!(ids.contains(&"alignment"));
assert!(ids.contains(&"resize"));
} }
#[test] #[test]
@ -379,13 +373,12 @@ mod tests {
let pipeline = Pipeline::with_steps_from_config(&config); let pipeline = Pipeline::with_steps_from_config(&config);
let ids = pipeline.step_ids(); let ids = pipeline.step_ids();
// Core steps: decode, crop, landmarks, alignment, resize // Core steps: decode, crop_and_resize, landmarks, alignment
assert!(ids.contains(&"decode")); assert!(ids.contains(&"decode"));
assert!(ids.contains(&"crop")); assert!(ids.contains(&"crop_and_resize"));
assert!(ids.contains(&"landmarks")); assert!(ids.contains(&"landmarks"));
assert!(ids.contains(&"alignment")); assert!(ids.contains(&"alignment"));
assert!(ids.contains(&"resize")); assert_eq!(ids.len(), 4);
assert_eq!(ids.len(), 5);
} }
#[test] #[test]
@ -399,12 +392,11 @@ mod tests {
let pipeline = Pipeline::with_steps_from_config(&config); let pipeline = Pipeline::with_steps_from_config(&config);
let ids = pipeline.step_ids(); let ids = pipeline.step_ids();
// Core steps (decode, crop, landmarks, alignment, resize) plus eye_filter // Core steps (decode, crop_and_resize, landmarks, alignment) plus eye_filter
assert!(ids.contains(&"decode")); assert!(ids.contains(&"decode"));
assert!(ids.contains(&"crop")); assert!(ids.contains(&"crop_and_resize"));
assert!(ids.contains(&"landmarks")); assert!(ids.contains(&"landmarks"));
assert!(ids.contains(&"eye_filter")); assert!(ids.contains(&"eye_filter"));
assert!(ids.contains(&"alignment")); // Alignment is now always enabled assert!(ids.contains(&"alignment")); // Alignment is now always enabled
assert!(ids.contains(&"resize"));
} }
} }

View file

@ -1,42 +1,57 @@
//! Face cropping step. //! Face cropping and resizing step.
//! //!
//! Crops the face region from the full image using the bounding box data. //! Crops the face region from the full image using the bounding box data,
//! then resizes it to the configured output size.
use crate::config::Config; use crate::config::Config;
use crate::pipeline::crop_face_with_intermediate; use crate::pipeline::crop_face_with_intermediate;
use crate::pipeline::{computed_keys, ComputedValue, PipelineContext, ProcessingStep, StepOutcome}; use crate::pipeline::{computed_keys, BoundingBox, ComputedValue, PipelineContext, ProcessingStep, StepOutcome};
use async_trait::async_trait; use async_trait::async_trait;
/// Crops the face region from the full image. /// Crops the face region from the full image and resizes it.
/// ///
/// This transformer step extracts the face region using the bounding box /// This transformer step extracts the face region using the bounding box
/// from Immich, adding padding around the face for context. /// from Immich (with padding) and immediately resizes it to the configured
pub struct CropFaceStep; /// output size.
pub struct CropAndResizeStep;
#[async_trait] #[async_trait]
impl ProcessingStep for CropFaceStep { impl ProcessingStep for CropAndResizeStep {
fn id(&self) -> &'static str { fn id(&self) -> &'static str {
"crop" "crop_and_resize"
} }
fn name(&self) -> &'static str { fn name(&self) -> &'static str {
"Face Crop" "Crop & Resize"
} }
async fn execute(&self, mut ctx: PipelineContext, config: &Config) -> StepOutcome { async fn execute(&self, mut ctx: PipelineContext, config: &Config) -> StepOutcome {
let image = match ctx.require_image("cropping") { let image = match ctx.require_image("cropping and resizing") {
Ok(img) => img, Ok(img) => img,
Err(e) => return StepOutcome::Error { ctx, error: e }, Err(e) => return StepOutcome::Error { ctx, error: e },
}; };
// Crop returns CropResult with cropped images and face rectangle in crop coordinates. let output_size = config.processing.output.size;
// We use the full_res_crop here and let the resize step handle the final sizing.
match crop_face_with_intermediate(image, &ctx.face_data, config.processing.output.size) { // Crop returns CropResult with cropped images and face rectangle in crop coordinates
match crop_face_with_intermediate(image, &ctx.face_data, output_size) {
Ok(crop_result) => { Ok(crop_result) => {
// Use the full-resolution cropped image; resize step will handle final size // Use the pre-resized image from the crop function
ctx.image = Some(crop_result.cropped); let cropped_size = crop_result.cropped.width();
// Store the face rectangle in crop coordinates for later steps ctx.image = Some(crop_result.resized);
ctx.set_computed(computed_keys::FACE_RECT, ComputedValue::FaceRect(crop_result.face_rect));
// Scale the face rectangle to match the resized image coordinates
let scale = output_size as f32 / cropped_size as f32;
let scaled_face_rect = BoundingBox {
x1: crop_result.face_rect.x1 * scale,
y1: crop_result.face_rect.y1 * scale,
x2: crop_result.face_rect.x2 * scale,
y2: crop_result.face_rect.y2 * scale,
};
// Store the scaled face rectangle for later steps
ctx.set_computed(computed_keys::FACE_RECT, ComputedValue::FaceRect(scaled_face_rect));
StepOutcome::Continue(ctx) StepOutcome::Continue(ctx)
} }
Err(e) => StepOutcome::Skip { Err(e) => StepOutcome::Skip {
@ -46,7 +61,6 @@ impl ProcessingStep for CropFaceStep {
}, },
} }
} }
} }
#[cfg(test)] #[cfg(test)]
@ -78,26 +92,29 @@ mod tests {
} }
#[tokio::test] #[tokio::test]
async fn test_crop_success() { async fn test_crop_and_resize_success() {
let step = CropFaceStep; let step = CropAndResizeStep;
let img = create_test_image(100, 100); let img = create_test_image(100, 100);
let ctx = make_ctx_with_image(img); let ctx = make_ctx_with_image(img);
let config = Config::default();
let mut config = Config::default();
config.processing.output.size = 512;
match step.execute(ctx, &config).await { match step.execute(ctx, &config).await {
StepOutcome::Continue(new_ctx) => { StepOutcome::Continue(new_ctx) => {
assert!(new_ctx.image.is_some()); assert!(new_ctx.image.is_some());
// The cropped image should be square let resized = new_ctx.image.unwrap();
let cropped = new_ctx.image.unwrap(); // Should be resized to the configured output size
assert_eq!(cropped.width(), cropped.height()); assert_eq!(resized.width(), 512);
assert_eq!(resized.height(), 512);
} }
_ => panic!("Expected Continue"), _ => panic!("Expected Continue"),
} }
} }
#[tokio::test] #[tokio::test]
async fn test_crop_no_image() { async fn test_crop_and_resize_no_image() {
let step = CropFaceStep; let step = CropAndResizeStep;
let face_data = FaceData { let face_data = FaceData {
bounding_box_x1: 30.0, bounding_box_x1: 30.0,
bounding_box_y1: 30.0, bounding_box_y1: 30.0,
@ -116,4 +133,23 @@ mod tests {
_ => panic!("Expected Error"), _ => panic!("Expected Error"),
} }
} }
#[tokio::test]
async fn test_crop_and_resize_different_sizes() {
let step = CropAndResizeStep;
let img = create_test_image(200, 200);
let ctx = make_ctx_with_image(img);
let mut config = Config::default();
config.processing.output.size = 256;
match step.execute(ctx, &config).await {
StepOutcome::Continue(new_ctx) => {
let resized = new_ctx.image.unwrap();
assert_eq!(resized.width(), 256);
assert_eq!(resized.height(), 256);
}
_ => panic!("Expected Continue"),
}
}
} }

View file

@ -5,20 +5,18 @@
mod alignment; mod alignment;
mod brightness; mod brightness;
mod crop; mod crop_and_resize;
mod decode; mod decode;
mod eye_filter; mod eye_filter;
mod face_resolution; mod face_resolution;
mod head_pose; mod head_pose;
mod landmarks; mod landmarks;
mod resize;
pub use alignment::AlignmentStep; pub use alignment::AlignmentStep;
pub use brightness::BrightnessStep; pub use brightness::BrightnessStep;
pub use crop::CropFaceStep; pub use crop_and_resize::CropAndResizeStep;
pub use decode::DecodeImageStep; pub use decode::DecodeImageStep;
pub use eye_filter::EyeFilterStep; pub use eye_filter::EyeFilterStep;
pub use face_resolution::FaceResolutionStep; pub use face_resolution::FaceResolutionStep;
pub use head_pose::HeadPoseStep; pub use head_pose::HeadPoseStep;
pub use landmarks::LandmarksStep; pub use landmarks::LandmarksStep;
pub use resize::ResizeStep;

View file

@ -1,125 +0,0 @@
//! Image resize step.
//!
//! Resizes the cropped face image to the configured output size.
use crate::config::Config;
use crate::pipeline::{PipelineContext, ProcessingStep, StepOutcome};
use async_trait::async_trait;
use image::imageops::FilterType;
/// Resizes the image to the configured output size.
///
/// This transformer step is typically the last step in the pipeline,
/// producing the final output image at the configured resolution.
pub struct ResizeStep;
#[async_trait]
impl ProcessingStep for ResizeStep {
fn id(&self) -> &'static str {
"resize"
}
fn name(&self) -> &'static str {
"Resize"
}
async fn execute(&self, mut ctx: PipelineContext, config: &Config) -> StepOutcome {
let image = match ctx.take_image("resizing") {
Ok(img) => img,
Err(e) => return StepOutcome::Error { ctx, error: e },
};
let output_size = config.processing.output.size;
// Resize to square output
let resized = image.resize_exact(output_size, output_size, FilterType::Lanczos3);
ctx.image = Some(resized);
StepOutcome::Continue(ctx)
}
}
#[cfg(test)]
mod tests {
use super::*;
use crate::immich_api::FaceData;
use image::{DynamicImage, RgbImage, Rgb};
fn make_ctx_with_image(image: DynamicImage) -> PipelineContext {
let face_data = FaceData {
bounding_box_x1: 0.0,
bounding_box_y1: 0.0,
bounding_box_x2: 100.0,
bounding_box_y2: 100.0,
image_width: 100,
image_height: 100,
};
PipelineContext::new("test".to_string(), "2024-01-01".to_string(), face_data)
.with_image(image)
}
fn create_test_image(width: u32, height: u32) -> DynamicImage {
let img = RgbImage::from_fn(width, height, |_, _| Rgb([128, 128, 128]));
DynamicImage::ImageRgb8(img)
}
#[tokio::test]
async fn test_resize_success() {
let step = ResizeStep;
let img = create_test_image(200, 200);
let ctx = make_ctx_with_image(img);
let mut config = Config::default();
config.processing.output.size = 512;
match step.execute(ctx, &config).await {
StepOutcome::Continue(new_ctx) => {
let resized = new_ctx.image.unwrap();
assert_eq!(resized.width(), 512);
assert_eq!(resized.height(), 512);
}
_ => panic!("Expected Continue"),
}
}
#[tokio::test]
async fn test_resize_no_image() {
let step = ResizeStep;
let face_data = FaceData {
bounding_box_x1: 0.0,
bounding_box_y1: 0.0,
bounding_box_x2: 100.0,
bounding_box_y2: 100.0,
image_width: 100,
image_height: 100,
};
let ctx = PipelineContext::new("test".to_string(), "2024-01-01".to_string(), face_data);
let config = Config::default();
match step.execute(ctx, &config).await {
StepOutcome::Error { error, .. } => {
assert!(error.contains("No image"));
}
_ => panic!("Expected Error"),
}
}
#[tokio::test]
async fn test_resize_upscale() {
let step = ResizeStep;
let img = create_test_image(50, 50); // Small image
let ctx = make_ctx_with_image(img);
let mut config = Config::default();
config.processing.output.size = 256;
match step.execute(ctx, &config).await {
StepOutcome::Continue(new_ctx) => {
let resized = new_ctx.image.unwrap();
assert_eq!(resized.width(), 256);
assert_eq!(resized.height(), 256);
}
_ => panic!("Expected Continue"),
}
}
}

View file

@ -25,7 +25,7 @@ pub mod computed_keys {
pub const LANDMARKS: &str = "landmarks"; pub const LANDMARKS: &str = "landmarks";
/// Head pose (yaw, pitch, roll) computed by HeadPoseStep. /// Head pose (yaw, pitch, roll) computed by HeadPoseStep.
pub const HEAD_POSE: &str = "head_pose"; pub const HEAD_POSE: &str = "head_pose";
/// Face bounding box in current image coordinates computed by CropFaceStep. /// Face bounding box in current image coordinates computed by CropAndResizeStep.
pub const FACE_RECT: &str = "face_rect"; pub const FACE_RECT: &str = "face_rect";
} }