Tensorflow Implementation of Depp-Face Recognition
Requirements - tensorflow==1.2 scipy scikit-learn opencv-python h5py matplotlib Pillow requests psutil
Pre Trained Models used - Inception Resnet V1.0 Dataset Used - Casia WebFace https://drive.google.com/open?id=1wYFljcEeT-GgCw7C93syEXG9n5v-nOAh
Face Aignment using MTCNN Face Detection.
How to -
Data Perparation
- Prepare a dataset of images of each person and place them in the Dataset folder in the following format
Deepface Recognition
----Dataset
------------Person1
------------Person2
...................
...................
...................
------------PersonN
- Create a folder aligned_faces
- Run python alignImages.py to crop and align the faces.
- The file aligned_images/bounding_boxes_28997.txt will cotain the x-axis,y-axis, width, height of the cropped face from the original image. (Not Needed)
Train the classifier
- Train a classifier, run python trainClassifier.py
- This will create a classifier at "Deepface Recognition/classifier" directory along with a text file consisting of the key,value pair (ID,Name)
Realtime Recognition - To recognize faces run "python RecognizeFaces.py"