Project Code Overview

The Makeup Engine is a software library designed to apply virtual makeup and other visual enhancements to faces detected in an image or video stream. The engine utilizes advanced face detection and landmark tracking to accurately map facial features, allowing for the realistic application of digital cosmetics. It is built upon a modular base that handles the core computer vision tasks, enabling easy extension with different visual effects.

Key Components

Core Classes

• MakeupEngine: The main engine class responsible for processing images and applying the virtual makeup effects. It orchestrates the underlying detection and rendering components.
• EngineBase: A foundational class that provides the main framework for image processing. It handles initialization, resource loading, and the overall processing pipeline that the MakeupEngine extends.
• core::Detection: A structure that holds all the necessary information about a detected face, including its bounding box and the precise coordinates of its facial landmarks.
• core::NeuralBase: This class serves as the base for all neural network models used in the system, primarily for detecting the facial mesh.

Face Detection and Landmark Mapping

core::face::FaceDetectorYunet: This class performs the initial step of robustly detecting faces in the input image using a DNN-based model from OpenCV.

core::face::FaceMesh: After a face is located, this class generates a detailed mesh of facial landmarks. These landmarks are crucial as they define the exact regions where makeup should be applied, such as the lips, eyes, and cheeks.

Makeup Application and Rendering

visualization::FaceDrawer: This utility class is central to the makeup application. Its drawMakeupMask method uses the detected facial landmarks to overlay makeup textures and colors onto the face. It processes a reference face with makeup and transfers the look to the current face by warping the makeup mask according to the user's specific facial geometry.

Temporal Smoothing

core::filter::LandmarkFlowFilter: In video applications, this filter tracks facial landmarks smoothly across consecutive frames. This ensures that the applied makeup remains stable and does not jitter during head movements, leading to a more realistic effect.

Function Highlights

• Engine Initialization: MakeupEngine::init(resourcePath) initializes the engine, loading all necessary resources such as face detection models and makeup templates.
• Image Processing: MakeupEngine::process(input, output) executes the full pipeline on an input image: it detects faces, generates landmarks, applies the virtual makeup, and writes the result to the output image.
• Makeup Application: FaceDrawer::drawMakeupMask(...) is the core function that applies the makeup. It uses the landmarks of both a reference and the current face to accurately warp and blend the makeup effect.
• Face Detection: FaceDetectorYunet::process(frame) returns a list of all faces found in the current frame.
• Landmark Generation: FaceMesh::process(frame, detections) enriches the detected face data with a detailed set of landmark points.

File and Dependency Structure

The project files are organized in the C:/Projects/Engine/AntalEngine/Engine directory. Key header files include:

• MakeupEngine.h: Contains the declaration of the main engine class.
• FaceDrawer.h: Provides the rendering functions, including the core makeup application logic.
• FaceMesh.h: Defines the class responsible for detailed facial landmark detection.
• FaceDetectorYunet.h: Defines the class for initial face detection.
• EngineBase.h: Contains the definition for the base engine class.
• LandmarkFlowFilter.h: Declares the filter for smoothing landmark tracking in video.