The task of this project is to apply deep neural network to develop models for face recognition and gender classification problems.
Main steps include:
- Captured pictures from a video using FPS of 30, then applied Dlib (sliding window) to detect face within the picture
- Applied Caffe to train deep neural network to get models based on AlexNet and GoogleNet
- Feature Extraction for gallery set and probe data (CNN forwarding, using feature layer)
- Matching by computing the distances between the probe feature with all the features in gallery
- Conduct Fine-tuning on pre-trained model for face gender classification (change last layer(s))
Caffe and VGG Face Dataset are used in this project for training models.