Project Code Overview

The Virtual Makeup engine is a software library designed to enable virtual makeup and facial feature visualization. This tool primarily utilizes OpenCV and custom neural network components to detect faces, map facial landmarks, and apply graphical effects like makeup or watermarks.

Key Components

Core Classes

• Detection: Represents detected objects (primarily faces), containing information like confidence level, class ID, name, and bounding rectangle.
• Engine: The central class for initializing the processing engine, handling image inputs, applying facial features, and adding watermarks.
• NeuralBase: A foundational class for managing neural networks, loading network configurations, and handling inference.

Face Detection

FaceDetectorYunet: Uses OpenCV's FaceDetectorYN for face detection with DNN support, allowing integration with deep learning frameworks. It detects faces and provides bounding boxes and landmark data.

Facial Landmark and Makeup Visualization

• FaceMesh: Builds on the neural network framework to identify and process facial landmarks, enabling face mesh creation.
• FaceDrawer: Handles the drawing of facial feature triangulations and makeup masks on faces. It can load triangles from files or buffers and uses OpenCV for drawing.

Mathematics and Image Utilities

CvUtils: Contains various utilities for image transformations, such as resizing rectangles, histogram equalization, and calculating angles or transformations between points.

Function Highlights

• Engine Initialization: Engine::init(resourcePath) sets up the engine with required assets and configurations.
• Image Processing: Engine::process(input, output) takes an input image, detects faces, and applies effects before saving to an output image.
• Watermarking: Engine::addSigno(image) adds a copyright watermark to an image.
• Face Detection: FaceDetectorYunet::process(frame) detects faces in a given frame and returns a list of Detection objects with information on each detected face.
• Face Mesh Detection: FaceMesh::process(frame, detections) processes an image frame to detect face mesh structures.
• Drawing: FaceDrawer::draw(image, facialPoints) draws a triangular mesh over the provided facial points on an image, useful for visualizing facial structure.

File and Dependency Structure

The project files are organized in the /AntalEngine/Core directory. Key files include:

• CvUtils.h: Provides image processing utilities.
• Detection.h: Contains the Detection struct for object detection data.
• Engine.h/.cpp: Defines the main Engine class.
• FaceDetectorYunet.h/.cpp: Contains face detection capabilities.
• FaceDrawer.h/.cpp: Implements facial point triangulation.
• NeuralBase.h/.cpp: Core neural network functionalities and configurations.

This structure enables modularity, allowing the use of each component independently or as part of a larger virtual makeup application. The software is well-suited for applications in AR (Augmented Reality) makeup, virtual try-ons, or facial recognition-based effects.