Face Recognition
Two different approach of face recognition will be tested and evaluated based on their accuracy and efficiency. The approaches can be divided into 3 stages, namely face detection, feature extraction and classification.
Face Detection
The face detection stage will be done using OpenCV due to time limitation. OpenCV provides 2 different method for face detection, the HAAR classifier and the local binary pattern classifier, both method will be tested to determine which will be best fitted for the application. The detected face will then be extracted and processed before moving on to the feature extraction stage.
The face detection stage will be done using OpenCV due to time limitation. OpenCV provides 2 different method for face detection, the HAAR classifier and the local binary pattern classifier, both method will be tested to determine which will be best fitted for the application. The detected face will then be extracted and processed before moving on to the feature extraction stage.
Feature Extraction
The feature extraction stage is the stage where the 2 approach differs from each other, the details on the extraction stage are explained in section 3.1.1 and 3.1.2.
The feature extraction stage is the stage where the 2 approach differs from each other, the details on the extraction stage are explained in section 3.1.1 and 3.1.2.
a) Face Recognition using machine learning to extract facial features
This approach will use a deep convolutional neural network to select facial landmark from the extracted facial image. During each training iteration, 2 photo of the same person and a photo of a different person will be used, the neural network will then be tweaked such that the feature points of the same person will be more similar and vice versa. |
b) Face Recognition using ratios between facial landmarks as features
This approach will use a face landmark estimation algorithm to locate the facial features (e.g. eye, nose, mouth). The ratios between different facial landmarks (e.g. eye to nose distance to eye to chin distance) will then be used as features for classification. |
Classification
After feature extraction, a classifier or clustering algorithm will be used to classify the person in the photo using the extracted features.
After feature extraction, a classifier or clustering algorithm will be used to classify the person in the photo using the extracted features.
This section will be updated when the face recognition feature is implemented
Datasets
|
About Us
Supervisor
Dr Tam Anthony |
Student
Kwan Hoi Shun |