This structure is designed to evaluate different kinds of face recognition algorithms in real-time application on platform of Windows.

The structure consists of five components, which are connected together by MS SQL Database:
A. Training Samples Gather
This module detects the position of face in the video sequence captured by camera, and locates the facial landmarks, then segments the facial image. After a series of preprocessing, including rotation, resizing, histogram equalization, masking and so on, the face samples for training are derived finally.                                                [Download: FaceGather@RCIR_v1.0.rar, 2MB]
B. Classifier Training
This module is employed for training face classifier. Through this module the parameters for training would be set easily and the training samples would be organized efficiently. By modifying some parts of this module, different classifier based on different algorithm would be obtained, that is very convenient for the comparison between different algorithms.
C. Real-Time Face Recognition
This module is designed for testing FR algorithms in real-time application.
D. Static Picture Face Recognition
This module is designed for testing FR algorithm in static pictures, for example, the standard face database, such as FERET.
E. Database Management
By using this module, the database would be managed effectively and efficiently.

The structure is shown in the figure above.

The data flows of this frame is described in the following figure.

In order to help unprofessional users to use the evaluation system, we developed friendly GUI for each module.

Face Samples Gathering

Database Management

Feature Extraction & Classifier Training

Real-Time Face Recognition

Static Picture Face Recognition

[1] J. Dai and J. Su, "An Open Framework for Face Recognition Algorithm Evaluation," (CN Patent No: 200910199356.1)