trying facenet

facenettest.py

import os
import cv2
import time
import glob
import facenet
import fr_utils
import numpy as np
import tensorflow as tf
from numpy import genfromtxt
from keras import backend as K
from keras.optimizers import Adam
from multiprocessing.dummy import Pool

K.set_image_data_format('channels_first')
PADDING = 50
epsilon = 0.2
imgsize = 96
trainpath = "../ssd/gaurav_train/gaurav/*"
testpath = "../ssd/gaurav_test/gaurav/"


def cosine_similarity(src_rep, dst_rep):
    a = np.matmul(np.transpose(src_rep), dst_rep)
    b = np.sum(np.multiply(src_rep, src_rep))
    c = np.sum(np.multiply(dst_rep, dst_rep))
    return 1 - (a / (np.sqrt(b) * np.sqrt(c)))


def triplet_loss(ytrue, ypred, alpha=0.3):
    anchor, positive, negative = ypred[0], ypred[1], ypred[2]

    sub = tf.subtract(anchor, positive)
    pos_dist = tf.reduce_sum(tf.square(sub), axis=-1)
    neg_dist = tf.reduce_sum(tf.square(tf.subtract(anchor, negative)), axis=-1)

    basic_loss = tf.add(tf.subtract(pos_dist, neg_dist), alpha)
    loss = tf.reduce_sum(tf.maximum(basic_loss, 0.0))

    return loss


model = facenet.FaceNet(input_shape=(3, imgsize, imgsize))
model.compile(optimizer='adam', loss=triplet_loss, metrics=['accuracy'])
fr_utils.load_weights_from_FaceNet(model)
# fr_utils.load_weights()


def prep_db():
    db = {}
    for file in glob.glob(trainpath):
        identify = os.path.splitext(os.path.basename(file))[0]
        db[identify] = fr_utils.img_path_to_encoding(file, model)

    return db


def verify(image, db, model):
    min_dist = 100
    identity = None
    distlist = []
    encoding = fr_utils.img_to_encoding(image, model)

    for (name, dbenc) in db.items():
        dist = np.linalg.norm(dbenc - encoding)
        #dist = cosine_similarity(dbenc[0, :], encoding[0, :])
        if dist < min_dist:
            print('Distance for %s is: %s' % (name, dist))
            min_dist = dist
            identity = name
        if dist not in distlist:
            distlist.append(dist)

    if min_dist > epsilon:
        return None
    else:
        best_dist = np.prod(np.array(distlist))
        return str(identity), best_dist


db = prep_db()

print('Testing for: ', testpath+"gaurav.jpg")
img = cv2.imread(testpath+"gaurav.jpg")
print(verify(img, db, model))

print()
print('------------------------------------------------------------------------------------------------')
print()

print('Testing for: ', testpath+"gav3.jpg")
img = cv2.imread(testpath+"gav3.jpg")
print(verify(img, db, model))

print()
print('------------------------------------------------------------------------------------------------')
print()

print('Testing for: ', testpath+"gav4.jpg")
img = cv2.imread(testpath+"gav4.jpg")
print(verify(img, db, model))

print()
print('##############################################################################################################')
print()

print('Testing for: ', testpath+"roma.jpg")
img = cv2.imread(testpath+"roma.jpg")
print(verify(img, db, model))

print()
print('##############################################################################################################')
print()

print('Testing for: ', testpath+"tar2.jpeg")
img = cv2.imread(testpath+"tar2.jpeg")
print(verify(img, db, model))

print()
print('##############################################################################################################')
print()

print('Testing for: ', testpath+"tarang.jpg")
img = cv2.imread(testpath+"tarang.jpg")
print(verify(img, db, model))