def predict_ethnic(self, name):
means = np.load('means_ethnic.npy')
model = create_face_network(nb_class=4,
hidden_dim=512,
shape=(224, 224, 3))
model.load_weights('weights_ethnic.hdf5')
im = cv2.imread(name, cv2.IMREAD_COLOR)
im = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
im = cv2.cvtColor(im, cv2.COLOR_GRAY2RGB)
im = cv2.resize(im, (224, 224))
im = np.float64(im)
im /= 255.0
im = im - means
#return model.predict(np.array([im]))
return model.predict(np.array([im]))
def predict_region(im):
means = np.load('means_region.npy')
model = create_face_network(nb_class=4,
hidden_dim=256,
shape=(224, 224, 3))
model.load_weights('weights_region.hdf5')
im = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
im = cv2.cvtColor(im, cv2.COLOR_GRAY2RGB)
im = cv2.resize(im, (224, 224))
# cv2.imshow("fsfs",im)
# cv2.waitKey(0)
im = np.float64(im)
im /= 255.0
im = im - means
return model.predict(np.array([im]))
def predict_ethnic(image_path):
means = np.load('means_ethnic.npy')
model = create_face_network(nb_class=4,
hidden_dim=512,
shape=(224, 224, 3))
model.load_weights('weights_ethnic.hdf5')
im = cv2.imread(image_path, cv2.IMREAD_COLOR)
im = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
im = cv2.cvtColor(im, cv2.COLOR_GRAY2RGB)
im = cv2.resize(im, (224, 224))
im = np.float64(im)
im /= 255.0
im = im - means
#print("NP: ", np)
#print("np.array([im]): ", np.array([im]))
#print("RESULT: ", model.predict(np.array([im])))
result = model.predict(np.array([im]))
race = ETHNIC[np.argmax(result)]
return race
X_data = X_data[p]
y_data = y_data[p]
X_train = X_data[0:int(round(train_split*num_images))]
y_train = y_data[0:int(round(train_split*num_images))]
X_test = X_data[int(round(train_split*num_images))+1:-1]
y_test = y_data[int(round(train_split*num_images))+1:-1]
means = np.mean(X_train, axis = 0)
X_train -= means
X_test -= means
np.save('means_ethnic.npy',means)
opt = SGD(lr=1e-3, decay=1e-6, momentum=0.9, nesterov=True)
checkpoint = ModelCheckpoint('weights_ethnic.hdf5', monitor='val_acc', verbose=1, save_best_only=True,
save_weights_only=True, mode='max')
model = create_face_network(nb_class=4, hidden_dim=512, shape=(224, 224, 3))
model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])
model.fit(X_train, y_train,
batch_size=32,
epochs=10,
verbose=1,
callbacks=[checkpoint],
validation_data=(X_test, y_test),
shuffle=True,
class_weight=None,
sample_weight=None,
initial_epoch=0)