This repository contains a complete pipline for real time face recognition. It includes:
- gathering images from your webcam/video source
- detection,alignment,resize of face images using Multi-task Cascaded Convolutional networks
- train models using your own aligned images
- computing embedings from your images
- real-time facial recognition through webcam
Usage:
- To add new data, put python add_new_data.py database_name subject_name <number_of_photos> <size_factor> in command line.
- To allign data, put python align_data_mtccn.py input_dir output_dir <image_size> <random_order> <gpu_memory_fraction> in command line.
- To train your model, put
- To activate webcam/video for facial recognition: python face_rec.py model_dir classifier_dir
Modification possibility:
- build raspberry Pi video pipeline(stream data from server) and add voice module for greeting purposes (Alert when no one is home? greet friends and family?)
- rescale images with better technique to obtain a more accurate representation of faces.