Build a face recognition system
- Implement the triplet loss function
- Use a pretrained model to map face images into 128-dimensional encodings
- Use these encodings to perform face verification and face recognition
- keras.models - Sequential-
- keras.layers - Conv2D, ZeroPadding2D, Activation, Input, concatenate -
- keras.models - Model -
- keras.layers.normalization - BatchNormalization -
- keras.layers.pooling - MaxPooling2D, AveragePooling2D -
- keras.layers.merge - Concatenate -
- keras.layers.core - Lambda, Flatten, Dense -
- keras.initializers - glorot_uniform -
- keras.engine.topology - Layer -
- keras - backend - K.set_image_data_format('channels_first')
- cv2
- os
- numpy
- pandas
- tensorflow
- fr_utils import *
- inception_blocks_v2 import *
details inside the notebook