def load_model():
"""
加载本地模型
:return:
"""
model = CNN3()
model.load_weights('./models/cnn3_best_weights.h5')
return model
def load_model():
"""
加载本地模型
:return:
"""
from model import CNN3
model = CNN3()
model.load_weights('cnn3_best_weights.h5')
return model
def load_cnn_model():
"""
载入CNN模型
:return:
"""
from model import CNN3
model = CNN3()
model.load_weights('./models/cnn3_best_weights.h5')
return model
def predict_no_gen():
"""
无增广预测
:return:
"""
from model import CNN3
from data import Jaffe
import numpy as np
model = CNN3()
model.load_weights('../models/cnn3_best_weights.h5')
expression, x_test, y_test = Jaffe().gen_train()
pred = model.predict(x_test)
pred = np.argmax(pred, axis=1)
print(np.sum(pred.reshape(-1) == y_test.reshape(-1)))
def predict_data_gen():
"""
有增广预测
:return:
"""
from model import CNN3
from data import Jaffe
import numpy as np
model = CNN3()
model.load_weights('../models/cnn3_best_weights.h5')
expression, x_test, y_test = Jaffe().gen_train()
pred = []
x_test = np.squeeze(x_test, axis=-1)
for index in range(x_test.shape[0]):
faces = generate_faces(x_test[index])
results = model.predict(faces)
result_sum = np.sum(results, axis=0)
label_index = np.argmax(result_sum, axis=0)
pred.append(label_index)
pred = np.array(pred)
print(np.sum(pred.reshape(-1) == y_test.reshape(-1)))
def plot_feature_map():
from model import CNN3
model = CNN3()
model.load_weights('../models/cnn3_best_weights.h5')
def get_feature_map(model, layer_index, channels, input_img):
from tensorflow.keras import backend as K
layer = K.function([model.layers[0].input], [model.layers[layer_index].output])
feature_map = layer([input_img])[0]
import matplotlib.pyplot as plt
plt.figure(figsize=(20, 8))
for i in range(channels):
img = feature_map[:, :, :, i]
plt.subplot(4, 8, i + 1)
plt.imshow(img[0], cmap='gray')
plt.savefig('rst.png')
plt.show()
import cv2
img = cv2.cvtColor(cv2.imread('../data/demo.jpg'), cv2.cv2.COLOR_BGR2GRAY)
img.shape = (1, 48, 48, 1)
get_feature_map(model, 4, 32, img)
def get_feature_map(model, layer_index, channels, input_img=None):
"""
可视化每个卷积层学到的特征图
:param model:
:param layer_index:
:param channels:
:param input_img:
:return:
"""
if not input_img:
input_img = load_test_image('../data/demo.jpg')
input_img.shape = (1, 48, 48, 1)
from keras import backend as K
layer = K.function([model.layers[0].input], [model.layers[layer_index+1].output])
feature_map = layer([input_img])[0]
import matplotlib.pyplot as plt
plt.figure(figsize=(20, 8))
for i in range(channels):
img = feature_map[:, :, :, i]
plt.subplot(4, 8, i+1)
plt.imshow(img[0], cmap='gray')
plt.show()
if __name__ == '__main__':
from model import CNN3
model = CNN3()
model.load_weights('../models/cnn3_best_weights.h5')
get_feature_map(model, 1, 32)
if opt.dataset == "fer2013":
expressions, x_train, y_train = Fer2013().gen_train()
_, x_valid, y_valid = Fer2013().gen_valid()
_, x_test, y_test = Fer2013().gen_test()
# target编码
y_train = to_categorical(y_train).reshape(y_train.shape[0], -1)
y_valid = to_categorical(y_valid).reshape(y_valid.shape[0], -1)
# 为了统一几个数据集，必须增加一列为0的
y_train = np.hstack((y_train, np.zeros((y_train.shape[0], 1))))
y_valid = np.hstack((y_valid, np.zeros((y_valid.shape[0], 1))))
print(
"load fer2013 dataset successfully, it has {} train images and {} valid iamges"
.format(y_train.shape[0], y_valid.shape[0]))
model = CNN3(input_shape=(48, 48, 1), n_classes=8)
sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(optimizer=sgd,
loss='categorical_crossentropy',
metrics=['accuracy'])
callback = [
# EarlyStopping(monitor='val_loss', patience=50, verbose=True),
# ReduceLROnPlateau(monitor='lr', factor=0.1, patience=20, verbose=True),
ModelCheckpoint('../models/cnn2_best_weights.h5',
monitor='val_acc',
verbose=True,
save_best_only=True,
save_weights_only=True)
]
train_generator = ImageDataGenerator(rotation_range=10,
def load_cnn_model():
from model import CNN3
model = CNN3()
model.load_weights('../models/cnn3_best_weights.h5')
return model
