def test_densenet(modelpath, batch_size):
dataLoader = DataLoader()
net = DenseNet.build_densenet()
net.compile(loss='sparse_categorical_crossentropy', metrics=['accuracy'])
net.build((1,64,64,3))
net.load_weights(modelpath)
test_images, test_labels = dataLoader.get_batch_test(batch_size)
net.evaluate(test_images, test_labels, verbose=2)
def train_densenet(batch_size, epoch):
dataLoader = DataLoader()
# build callbacks
checkpoint = tf.keras.callbacks.ModelCheckpoint('{epoch}_epoch_densenet_weight.h5',
save_weights_only=True,
verbose=1,
save_freq='epoch')
# build model
net = DenseNet.build_densenet()
net.compile(optimizer=tf.keras.optimizers.Adam(lr=0.001, decay=1e-6),loss='sparse_categorical_crossentropy', metrics=['accuracy'])
# 详细参数见官方文档：https://tensorflow.google.cn/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator?hl=en
data_generate = ImageDataGenerator(
featurewise_center=False,# 将输入数据的均值设置为0
samplewise_center=False, # 将每个样本的均值设置为0
featurewise_std_normalization=False, # 将输入除以数据标准差，逐特征进行
samplewise_std_normalization=False, # 将每个输出除以其标准差
zca_epsilon=1e-6, # ZCA白化的epsilon值，默认为1e-6
zca_whitening=False, # 是否应用ZCA白化
rotation_range=10, # 随机旋转的度数范围，输入为整数
width_shift_range=0.1, # 左右平移，输入为浮点数，大于1时输出为像素值
height_shift_range=0.1, # 上下平移，输入为浮点数，大于1时输出为像素值
shear_range=0., # 剪切强度，输入为浮点数
zoom_range=0.1, # 随机缩放，输入为浮点数
channel_shift_range=0., # 随机通道转换范围，输入为浮点数
fill_mode='nearest', # 输入边界以外点的填充方式，还有constant,reflect,wrap三种填充方式
cval=0., # 用于填充的值，当fill_mode='constant'时生效
horizontal_flip=True, # 随机水平翻转
vertical_flip=False, # 随机垂直翻转
rescale=None, # 重缩放因子，为None或0时不进行缩放
preprocessing_function=None, # 应用于每个输入的函数
data_format='channels_last', # 图像数据格式，默认为channels_last
validation_split=0.0
)
# 引用自：https://www.jianshu.com/p/1576da1abd71
train_images,train_labels = dataLoader.get_batch_train(60000)
net.fit(
data_generate.flow(train_images, train_labels,
batch_size=batch_size,
shuffle=True,
#save_to_dir='resource/images'
),
steps_per_epoch=len(train_images) // batch_size,
epochs=epoch,
callbacks=[checkpoint],
shuffle=True)
def train_densenet(batch_size, epoch):
dataLoader = DataLoader()
# build callbacks
checkpoint = tf.keras.callbacks.ModelCheckpoint(f'./weight/{epoch}_epoch_densenet_weight.h5', save_best_only=True, save_weights_only=True, verbose=1, save_freq='epoch')
# build model
net = DenseNet.build_densenet()
net.compile(tf.keras.optimizers.Adam(lr=0.001, decay=1e-6), loss='sparse_categorical_crossentropy', metrics=['accuracy'])
# num_iter = dataLoader.num_train//batch_size
# for e in range(epoch):
# for i in range(num_iter):
# train_images, train_labels = dataLoader.get_batch_train(batch_size)
# net.fit(train_images, train_labels, shuffle=False, batch_size=batch_size, validation_split=0.1, callbacks=[checkpoint])
# net.save_weights("./weight/"+str(e+1)+"epoch_iter"+str(i)+"_resnet_weight.h5")
data_generate = ImageDataGenerator(
featurewise_center=False,
samplewise_center=False,
featurewise_std_normalization=False,
samplewise_std_normalization=False,
zca_epsilon=1e-6,
zca_whitening=False,
rotation_range=10,
width_shift_range=0.1,
height_shift_range=0.1,
shear_range=0.,
zoom_range=0.1,
channel_shift_range=0,
fill_mode='nearest',
cval=0.,
horizontal_flip=True,
vertical_flip=False,
rescale=None,
preprocessing_function=None,
data_format='channels_last',
validation_split=0.0)
train_images, train_labels = dataLoader.get_batch_train(60000)
net.fit(data_generate.flow(train_images, train_labels, batch_size=batch_size, shuffle=True,),
steps_per_epoch=len(train_images)//batch_size,
epochs=epoch,
callbacks=[checkpoint],
shuffle=True)