System Implementation and Architecture

---

[figure1]

[figure2]

---

Suggested System and Implementation

Figure 1) shows the system and implementation structure. The proposed makeup guide system recognizes facial features in the incoming video through a webcam and analyzes facial features through a neural network. In addition, based on this analysis, augmented reality technology was used to propose a makeup guideline suitable for the user, and a beauty mirror was manufactured by attaching a half mirror to the LCD display to show it in real time. Details of the proposed system are as follows.

---

Software Architecture

Figure 2) shows the software architecture of the proposed system. The proposed system is divided into three parts. 1) face recognition, 2) face type analysis and classification 3) 3D makeup guideline proposal. Face Detection module detects many feature points such as jaw, mouth, nose, eyes, eyebrows, etc. from the user's face through WebCamTexture by using a webcam and displays it in real time . Specific points associated with each other, such as facial contours, eyes, and lips, are linked.

---

The module for face shape analysis has developed a facial analysis algorithm based on machine learning of neural network. First, the face type analysis algorithm extracts the feature value by calculating the ratio of the angle of the jaw and the horizontal and vertical lengths of the face using the coordinate value of the facial contour point. Using these calculated feature values as inputs, we classify facial features into four types: angular, round, long, and other facial types through neural network.

---

The module for the cosmetic guideline suggests guidelines for eyebrows, lips, blushers and shading on the LCD to suggest makeup that complements the disadvantages of the user's face type. It floats in real time. When a user selects a desired color among the colors displayed on a 3D object, the corresponding color is applied to the inside of the guideline. Thus, the combination of the color of the user's skin and the overall color can be confirmed in advance .