Machine Learning

• Facial recognition implementation using Keras VGG face
• Important note:
• The weights and model are not in github due to large size.
• Please run the machine learning program (jupyter notebook) first to generate weight and model.
• Run face_detection.py for still images
• Run facecam_detection.py for video images

Project Summary

• This project uses Open CV library to locate a face in real time from video recording and the face will be classified as either Yining, Mohammad, or a stranger. This recognition is accomplished through fine-tuning a pre-trained deep neural network (VGG Face) in which the weights of the final layers in the network are retrained to fit this project.
• We used 600 pictures for training and 100 pictures for testing the model. Each picture is resized to 64 * 64 pixels.

Table of content for code

• Training algorithm
• Resizing images
• Training Images
• Test Images
• Real time face detection

Built with:

• Tensorflow
• Keras
• Open CV
• VGG Face
• Flickerapi

Contributors:

• Sihao Chen
• Mohammad Rafi
• Yining Wang

References

1. Cropping Faces
   1. Directory images
   2. Webcam
2. OpenCV images conversion
3. Saving Keras Model
4. CNN lab 8

Powerpoint Link

• https://docs.google.com/presentation/d/12KdVGsoL2VPiVyvsXnAxDHVP1pKVoVZlH3tx9Axc0lQ/edit?usp=sharing