face recognition for Indian Celebrity

Models Used:

• pretrained inception_v2 as model which takes (3,96,96) dimension input images and gives 128 dimension output embeddings
• pretrained inception_resnet_v1 as model1 which takes (160,160,3) dimensional input images and gives 128 dimensional output vectors.

Files Description

• face_images: Contains cropped and uncropped images of celebrity arranged according to name of the celebrity.

• images: Contains celebrity images arranged in number in order to compare the test images from database.

• images1: Contains 6 images to fintune the threshhold and debug code fastly.

• test_images: Contain just one image to test the output of the model

• weights: contains the weight of pretrained inception_v2 in csv files

• celebrity_face_recognition.ipynb: notebook containing the codes with slight explanation.

• celebrity_face_recognition.py python file for the ipynb file provided above.

• facenet_keras_weight.h5: pretrained weight of inception_resnet_v1.

• image_to_face.py : Converts uncropped imges to the cropped image.

• video_to_face.py: Convert the uncropped images from webcam to cropped and same them to folder images.