Face Recognition System using OpenCV and Deep Learning

This is an implementation of Face recognition and identification to implement an attendance system using a live video feed.

Setup

1. Install and activate a virtualenv

pip install virtualenv
python3 -m virtualenv env
source env/bin/activate

2. Install Requirements

pip3 install -r requirements.txt

3. Generate Training Data [OPTIONAL]

The training data is the pictures of all the people whom you want to recognize. This script uses your webcam to capture 50 images continuously and store it in the train_img folder.

python3 generate_data.py

3. Preprocess data

The training data is now preprocessed with the faces being cropped and the faces are labelled here.

python3 data_preprocess.py

4. Model Training

The model is trained to extract the features from the face by using Haar Cascaders and the classes are labelled here. `classifier.py' contains the model which we are currently using. I currently use Linear SVM to predict the probability of a particular class

python3 train_model.py

5. Identify Face

This is the last part, where we try to identify a particular face by drawing a bounding box around it. The class with the highest probability for the image is shown as the label of the image. The current confidence threshhold is 70%, but this can be changed according to the preference of the user. If the confidence is less than 70% then the face is not identified.

python3 indentify_face_video.py

Credits:

1. FaceNet: A Unified Embedding for Face Recognition and Clustering, David Sandberg Repo
2. A Discriminative Feature Learning Approach for Deep Face Recognition
3. Deep Face Recognition