日萌社
人工智能AI：Keras PyTorch MXNet TensorFlow PaddlePaddle 深度學習實戰（不定時更新）
EfficientNet中的每個模型要求的輸入形狀大小
每個網絡要求的輸入形狀大小：
EfficientNetB0 - (224, 224, 3)
EfficientNetB1 - (240, 240, 3)
EfficientNetB2 - (260, 260, 3)
EfficientNetB3 - (300, 300, 3)
EfficientNetB4 - (380, 380, 3)
EfficientNetB5 - (456, 456, 3)
EfficientNetB6 - (528, 528, 3)
EfficientNetB7 - (600, 600, 3)
EfficientNet模型遷移的使用注意事項：
1.因爲該模型的源碼是在tensorflow 1.x的版本，並非是tensorflow 2.0的版本，因此在tensorflow 2.0環境中使用的話，
需要用到tf.compat.v1.disable_eager_execution()，表示關閉默認的eager模式，但要注意的是，如果關閉默認的eager模式了的話，
那麼同時還使用tf.keras.callbacks.TensorBoard的話會報錯，tf.keras.callbacks.ModelCheckpoint不會報錯，
那麼解決的方式要麼此時不使用TensorBoard，或者不關閉默認的eager模式。
2.layers.py中的class Swish中的call()函數返回值的修改建議。
如果使用了tf.compat.v1.disable_eager_execution()之後，報錯No registered 'swish_f32' OpKernel for GPU devices compatible with node的話，
把 layers.py中的class Swish中的call()函數返回值 return tf.nn.swish(inputs) 修改爲 return inputs * tf.math.sigmoid(inputs) 即可解決，
實際上底層是 tf.nn.swish(x) 封裝了 x* tf.math.sigmoid(x)，不使用tf.nn.swish之後，即可也把tf.compat.v1.disable_eager_execution()給註釋掉，
即不需要關閉默認的eager模式了，那麼此時也可以正常同時使用TensorBoard。
使用EfficientNet模型遷移 第一個版本
import multiprocessing
import numpy as np
import tensorflow as tf
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.callbacks import TensorBoard, Callback
from tensorflow.keras.layers import Dense, GlobalAveragePooling2D
from tensorflow.keras.models import Model
from tensorflow.keras.optimizers import Adam, RMSprop
from efficientnet import model as EfficientNet
import os
from sklearn.model_selection import train_test_split
from tensorflow.keras.applications.vgg16 import VGG16, preprocess_input
# 注意關閉默認的eager模式
# tf.compat.v1.disable_eager_execution()
num_epochs = 10
batch_size = 10
learning_rate = 0.001
Dataset_dir = './Sign-Language-Digits-Dataset-master/Dataset/'
Dataset_test_dir = './Sign-Language-Digits-Dataset-master/Examples/'
Dataset_dir_list = []
Dataset_test_dir_list = []
lable_list = []
train_img_fileNames = []
train_img_lables = []
test_img_fileNames = []
test_img_lables = []
input_size = 100
optimizer = tf.keras.optimizers.Adam(learning_rate=1e-3)
losses = tf.keras.losses.sparse_categorical_crossentropy
metrics = [tf.keras.metrics.sparse_categorical_accuracy]
"""
metrics 是 sparse_categorical_crossentropy，則ModelCheckpoint中的第一個路勁字符串信息中和monitor參數中都需要用val_sparse_categorical_crossentropy
metrics 是 acc，則ModelCheckpoint中的第一個路勁字符串信息中和monitor參數中都需要用val_acc
metrics 是 accuracy，則ModelCheckpoint中的第一個路勁字符串信息中和monitor參數中都需要用val_accuracy
"""
def decode_and_resize(filename, label):
image_string = tf.io.read_file(filename)
image_decoded = tf.image.decode_jpeg(image_string)
image_resized = tf.image.resize(image_decoded, [100, 100]) / 255.0
return image_resized, label
def VGG_224x244_decode_and_resize(filename, label):
image_string = tf.io.read_file(filename)
image_decoded = tf.image.decode_jpeg(image_string)
image_resized = tf.image.resize(image_decoded, [224, 224]) / 255.0
return image_resized, label
def EfficientNet_300x300_decode_and_resize(filename, label):
image_string = tf.io.read_file(filename)
image_decoded = tf.image.decode_jpeg(image_string)
image_resized = tf.image.resize(image_decoded, [300, 300]) / 255.0
return image_resized, label
def get_model():
# inputs = tf.keras.Input(shape=(100, 100, 3))
# conv1 = tf.keras.layers.Conv2D(
# filters=32, # 卷積核數量
# kernel_size=[3, 3], # 卷積核大小
# padding='same', # 領填充方式
# activation=tf.nn.relu, # 激活函數
# kernel_regularizer=tf.keras.regularizers.l2(0.001)
# )(inputs)
# pool1 = tf.keras.layers.MaxPool2D(pool_size=[2, 2], strides=2)(conv1)
# conv2 = tf.keras.layers.Conv2D(
# filters=64, # 卷積核數量
# kernel_size=[3, 3], # 卷積核大小
# padding='same', # 領填充方式
# activation=tf.nn.relu, # 激活函數
# kernel_regularizer=tf.keras.regularizers.l2(0.001)
# )(pool1)
# pool2 = tf.keras.layers.MaxPool2D(pool_size=[2, 2], strides=2)(conv2)
# flatten = tf.keras.layers.Flatten()(pool2)
# dense1 = tf.keras.layers.Dense(units=1024, activation=tf.nn.relu,
# kernel_regularizer=tf.keras.regularizers.l2(0.001))(flatten)
# # dropout = tf.keras.layers.Dropout(rate=0.4)(dense1)
# dense2 = tf.keras.layers.Dense(units=10)(dense1)
# outputs = tf.keras.layers.Softmax()(dense2)
# model = tf.keras.Model(inputs=inputs, outputs=outputs)
# ================================================================
model = tf.keras.Sequential([
tf.keras.layers.Conv2D(32, 3, activation='relu', input_shape=(100, 100, 3)),
tf.keras.layers.MaxPooling2D(),
tf.keras.layers.Conv2D(64, 3, activation='relu'),
tf.keras.layers.MaxPooling2D(),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(1024, activation='relu'),
tf.keras.layers.Dense(10, activation='softmax')
])
return model
def train_model_1(model):
# lable = os.listdir(Dataset_dir)
# print(lable) #['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
lable = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
for i in lable:
temp_list = []
for filename in os.listdir(Dataset_dir + i):
temp_list.append(Dataset_dir + str(i) + "/" + filename)
Dataset_dir_list.append(temp_list)
lable_list.append(np.full(len(temp_list), int(i))) # np.full(5,10)
# print(len(lable_list)) #10
# print(len(Dataset_dir_list)) #10
# ================================================================
for filename in os.listdir(Dataset_test_dir):
Dataset_test_dir_list.append(Dataset_test_dir + filename)
Dataset_dir_lists = tf.constant([f for filenames in Dataset_dir_list for f in filenames])
lable_lists = tf.constant([int(s) for lables in lable_list for s in lables])
Dataset_dir_lists_test = tf.constant([f for f in Dataset_test_dir_list])
lable_lists_test = tf.constant([int(s) for s in lable])
# ================================================================
train_dataset = tf.data.Dataset.from_tensor_slices((Dataset_dir_lists, lable_lists))
train_dataset = train_dataset.map(decode_and_resize)
train_dataset = train_dataset.batch(batch_size)
train_dataset = train_dataset.shuffle(buffer_size=5000)
train_dataset = train_dataset.prefetch(tf.data.experimental.AUTOTUNE)
test_dataset = tf.data.Dataset.from_tensor_slices((Dataset_dir_lists_test, lable_lists_test))
test_dataset = test_dataset.map(decode_and_resize)
test_dataset = test_dataset.batch(batch_size)
# ================================================================
#還要預先建立 ./graph/train/plugins/profile 的目錄
#要在graph的同級目錄下 使用命令 tensorboard --logdir=graph 或者 tensorboard --logdir='./graph'
#http://localhost:6006/ 訪問
tensorboard = tf.keras.callbacks.TensorBoard(log_dir='./graph', histogram_freq=1, write_graph=True, write_images=True)
#預先建立./ckpt目錄
check = tf.keras.callbacks.ModelCheckpoint('./ckpt/weights_{epoch:02d}-{val_loss:.2f}.h5',
monitor='val_loss',
save_best_only=True,
save_weights_only=False,
mode='auto',
period=1)
# ================================================================
model.compile(loss=losses, optimizer=optimizer, metrics=metrics)
model.fit(train_dataset, validation_data=test_dataset, epochs=num_epochs, callbacks=[check, tensorboard]) #
y_pred = model.predict(test_dataset)
# 定義評估函數
sparse_categorical_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()
# 定義測試數據集一共批次的大小
sparse_categorical_accuracy.update_state(y_true=lable_lists_test, y_pred=y_pred)
print("測試準確率: %f" % sparse_categorical_accuracy.result())
def train_model_2(model):
# lable = os.listdir(Dataset_dir)
# print(lable) #['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
lable = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
for i in lable:
temp_list = []
for filename in os.listdir(Dataset_dir+i):
temp_list.append(Dataset_dir+str(i)+"/"+filename)
Dataset_dir_list.append(temp_list)
lable_list.append(np.full(len(temp_list),int(i))) #np.full(5,10)
print(len(lable_list)) #10
print(len(Dataset_dir_list)) #10
# ================================================================
for i in lable:
x_train, x_test, y_train, y_test = train_test_split(Dataset_dir_list[int(i)], lable_list[int(i)], test_size=0.1,
random_state=22)
train_img_fileNames.append(x_train)
train_img_lables.append(y_train)
test_img_fileNames.append(x_test)
test_img_lables.append(y_test)
print(len(train_img_fileNames)) #10
print(len(train_img_lables)) #10
print(len(test_img_fileNames)) #10
print(len(test_img_lables)) #10
Dataset_dir_lists = [f for train_img in train_img_fileNames for f in train_img]
lable_lists = [f for train_lable in train_img_lables for f in train_lable]
Dataset_dir_lists_test = [f for test_img in test_img_fileNames for f in test_img]
lable_lists_test = [f for test_lables in test_img_lables for f in test_lables]
print(len(Dataset_dir_lists)) #1852
print(len(lable_lists)) #1852
print(len(Dataset_dir_lists_test)) #210
print(len(lable_lists_test)) #210
# ================================================================
train_dataset = tf.data.Dataset.from_tensor_slices((Dataset_dir_lists, lable_lists))
train_dataset = train_dataset.map(decode_and_resize)
train_dataset = train_dataset.batch(batch_size)
train_dataset = train_dataset.shuffle(buffer_size=5000)
train_dataset = train_dataset.prefetch(tf.data.experimental.AUTOTUNE)
test_dataset = tf.data.Dataset.from_tensor_slices((Dataset_dir_lists_test, lable_lists_test))
test_dataset = test_dataset.map(decode_and_resize)
test_dataset = test_dataset.batch(batch_size)
# ================================================================
#還要預先建立 ./graph/train/plugins/profile 的目錄
#要在graph的同級目錄下 使用命令 tensorboard --logdir=graph 或者 tensorboard --logdir='./graph'
#http://localhost:6006/ 訪問
tensorboard = tf.keras.callbacks.TensorBoard(log_dir='./graph', histogram_freq=1, write_graph=True, write_images=True)
#預先建立./ckpt目錄
check = tf.keras.callbacks.ModelCheckpoint('./ckpt/weights_{epoch:02d}-{val_loss:.2f}.h5',
monitor='val_loss',
save_best_only=True,
save_weights_only=False,
mode='auto',
period=1)
# ================================================================
model.compile(loss=losses, optimizer=optimizer, metrics=metrics)
model.fit(train_dataset, validation_data=test_dataset, epochs=num_epochs, callbacks=[check, tensorboard]) #
y_pred = model.predict(test_dataset)
# 定義評估函數
sparse_categorical_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()
# 定義測試數據集一共批次的大小
sparse_categorical_accuracy.update_state(y_true=lable_lists_test, y_pred=y_pred)
print("測試準確率: %f" % sparse_categorical_accuracy.result())
def train_model_3(model):
train_datagen = ImageDataGenerator(
rescale=1. / 255,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True)
train_generator = train_datagen.flow_from_directory(
'./Sign-Language-Digits-Dataset-master/Dataset/',
target_size=(100, 100),
batch_size=10,
class_mode='binary')
for filename in os.listdir(Dataset_test_dir):
Dataset_test_dir_list.append(Dataset_test_dir + filename)
Dataset_dir_lists_test = [f for f in Dataset_test_dir_list]
lable = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
test_dataset = tf.data.Dataset.from_tensor_slices((Dataset_dir_lists_test, lable))
test_dataset = test_dataset.map(decode_and_resize)
test_dataset = test_dataset.batch(batch_size)
# ================================================================
#還要預先建立 ./graph/train/plugins/profile 的目錄
#要在graph的同級目錄下 使用命令 tensorboard --logdir=graph 或者 tensorboard --logdir='./graph'
#http://localhost:6006/ 訪問
tensorboard = tf.keras.callbacks.TensorBoard(log_dir='./graph', histogram_freq=1, write_graph=True, write_images=True)
modelckpt = tf.keras.callbacks.ModelCheckpoint('./ckpt/transfer_{epoch:02d}-{val_loss:.2f}.h5',
monitor='val_loss',
save_best_only=True,
save_weights_only=False,
mode='auto',
period=1)
model.compile(loss=losses, optimizer=optimizer, metrics=metrics)
model.fit_generator(
train_generator,
epochs=10,
validation_steps=800, validation_data=train_generator,
callbacks=[modelckpt, tensorboard])
y_pred = model.predict(test_dataset)
# 定義評估函數
sparse_categorical_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()
# 定義測試數據集一共批次的大小
sparse_categorical_accuracy.update_state(y_true=lable, y_pred=y_pred)
print("測試準確率: %f" % sparse_categorical_accuracy.result())
def train_model_4():
train_datagen = ImageDataGenerator(
rescale=1. / 255,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True)
train_generator = train_datagen.flow_from_directory(
'./Sign-Language-Digits-Dataset-master/Dataset/',
target_size=(224, 224),
batch_size=10,
class_mode='binary')
for filename in os.listdir(Dataset_test_dir):
Dataset_test_dir_list.append(Dataset_test_dir + filename)
Dataset_dir_lists_test = [f for f in Dataset_test_dir_list]
lable = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
test_dataset = tf.data.Dataset.from_tensor_slices((Dataset_dir_lists_test, lable))
test_dataset = test_dataset.map(VGG_224x244_decode_and_resize)
test_dataset = test_dataset.batch(10)
# ================================================================
base_model = VGG16(weights='imagenet', include_top=False)
#修改VGG的模型，在VGG.outputs[0]輸出的爲[None, ?, ?, 512]--->全局平均池化-->兩個全連接層1024, 10
# 1、獲取VGG模型的輸出，不包含原有模型的top結構
x = base_model.outputs[0]
# 2、在VGG的輸出之後定義自己的模型
x = tf.keras.layers.GlobalAveragePooling2D()(x)
# 兩個全連接層
x = tf.keras.layers.Dense(1024, activation=tf.nn.relu)(x)
y_predict = tf.keras.layers.Dense(10, activation=tf.nn.softmax)(x)
#凍結VGG的前面卷積結構，不參與訓練
# 循環獲取base_model當中的層
for layer in base_model.layers:
layer.trainable = False
# 3、使用Model封裝新的模型返回
transfer_model = tf.keras.models.Model(inputs=base_model.inputs, outputs=y_predict)
# ================================================================
#還要預先建立 ./graph/train/plugins/profile 的目錄
#要在graph的同級目錄下 使用命令 tensorboard --logdir=graph 或者 tensorboard --logdir='./graph'
#http://localhost:6006/ 訪問
tensorboard = tf.keras.callbacks.TensorBoard(log_dir='./graph', histogram_freq=1, write_graph=True, write_images=True)
modelckpt = tf.keras.callbacks.ModelCheckpoint('./ckpt/transfer_{epoch:02d}-{val_loss:.2f}.h5',
monitor='val_loss',
save_best_only=True,
save_weights_only=False,
mode='auto',
period=1)
transfer_model.compile(loss=losses, optimizer=optimizer, metrics=metrics)
transfer_model.summary()
transfer_model.fit_generator(
train_generator,
steps_per_epoch=207,
epochs=2,
validation_data=train_generator,
callbacks=[modelckpt, tensorboard])
y_pred = transfer_model.predict(test_dataset)
# 定義評估函數
sparse_categorical_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()
# 定義測試數據集一共批次的大小
sparse_categorical_accuracy.update_state(y_true=lable, y_pred=y_pred)
print(sparse_categorical_accuracy.result())
def train_model_5():
# lable = os.listdir(Dataset_dir)
# print(lable) #['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
lable = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
for i in lable:
temp_list = []
for filename in os.listdir(Dataset_dir+i):
temp_list.append(Dataset_dir+str(i)+"/"+filename)
Dataset_dir_list.append(temp_list)
lable_list.append(np.full(len(temp_list),int(i))) #np.full(5,10)
print(len(lable_list)) #10
print(len(Dataset_dir_list)) #10
# ================================================================
for filename in os.listdir(Dataset_test_dir):
Dataset_test_dir_list.append(Dataset_test_dir+filename)
train_dataset_len = len(Dataset_dir_list)
print(train_dataset_len)
Dataset_dir_lists = tf.constant([f for filenames in Dataset_dir_list for f in filenames])
lable_lists = tf.constant([int(s) for lables in lable_list for s in lables])
Dataset_dir_lists_test = tf.constant([f for f in Dataset_test_dir_list])
lable_lists_test = tf.constant([int(s) for s in lable])
# ================================================================
"""
1.使用repeat() 解決如下報錯：
WARNING:tensorflow:Your input ran out of data; interrupting training.
Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs`
batches (in this case, 414 batches). You may need to use the repeat() function
when building your dataset.
警告：tensorflow：輸入的數據用完；中斷訓練。請確保您的數據集或生成器至少可以生成
“每個steps_per_epoch*epoch”批（在本例中爲414 batches）。在構建數據集時，可能需要使用repeat()函數。
2.例子：batch、repeat、steps_per_epoch、epochs的使用
1.先用 batch(批次大小) 然後才用 repeat(重複次數)，
比如：repeat(2)重複數據集2次，即複製一份數據集，最終即有兩份數據集。
2.steps_per_epoch 即表示 一個epoch 裏面遍歷批量數據的次數，即遍歷多少個批量數據完成一個epoch。
3.應保證要輸入到模型的數據集(包括訓練集/驗證集)的批量個數都均爲steps_per_epoch*epoch。
3.使用順序：from_tensor_slices -> map -> shuffle -> batch -> repeat -> prefetch
"""
train_dataset = tf.data.Dataset.from_tensor_slices((Dataset_dir_lists, lable_lists))
train_dataset = train_dataset.map(EfficientNet_300x300_decode_and_resize, num_parallel_calls=tf.data.experimental.AUTOTUNE)
train_dataset = train_dataset.shuffle(buffer_size=5000)
train_dataset = train_dataset.batch(batch_size)
epochs_repeats = 10
train_dataset = train_dataset.repeat(epochs_repeats)
train_dataset = train_dataset.prefetch(tf.data.experimental.AUTOTUNE)
test_dataset = tf.data.Dataset.from_tensor_slices((Dataset_dir_lists_test, lable_lists_test))
test_dataset = test_dataset.map(EfficientNet_300x300_decode_and_resize, num_parallel_calls=tf.data.experimental.AUTOTUNE)
test_dataset = test_dataset.batch(batch_size)
test_dataset = test_dataset.repeat(epochs_repeats)
# ================================================================
#還要預先建立 ./graph/train/plugins/profile 的目錄
#要在graph的同級目錄下 使用命令 tensorboard --logdir=graph 或者 tensorboard --logdir='./graph'
#http://localhost:6006/ 訪問
tensorboard = tf.keras.callbacks.TensorBoard(log_dir='./graph', histogram_freq=1, write_graph=True, write_images=True)
#預先建立./ckpt目錄
check = tf.keras.callbacks.ModelCheckpoint('./ckpt/weights_{epoch:02d}-{val_loss:.2f}.h5',
monitor='val_loss',
save_best_only=True,
save_weights_only=False,
mode='auto',
period=1)
# ================================================================
num_classes = 10
base_model = EfficientNet.EfficientNetB3(include_top=False, input_shape=(300, 300, 3), classes=num_classes)
x = base_model.output
x = GlobalAveragePooling2D(name='avg_pool')(x)
predictions = Dense(num_classes, activation='softmax')(x)
model = Model(inputs=base_model.input, outputs=predictions)
model.compile(loss=losses, optimizer=optimizer, metrics=metrics)
model.summary()
model.fit(
train_dataset,
# 一個epoch需要多少步
#Train on 207 steps, validate on 1 steps
steps_per_epoch=207,
epochs=epochs_repeats,
callbacks=[check, tensorboard],
validation_data=test_dataset
)
return None
def predict_model_5():
for filename in os.listdir(Dataset_test_dir):
Dataset_test_dir_list.append(Dataset_test_dir+filename)
lable = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
lable_list = []
epochs_repeats = 10
for _ in range(epochs_repeats):
lable_list.extend(lable)
Dataset_dir_lists_test = tf.constant([f for f in Dataset_test_dir_list])
lable_lists_test = tf.constant([int(s) for s in lable])
test_dataset = tf.data.Dataset.from_tensor_slices((Dataset_dir_lists_test, lable_lists_test))
test_dataset = test_dataset.map(EfficientNet_300x300_decode_and_resize, num_parallel_calls=tf.data.experimental.AUTOTUNE)
test_dataset = test_dataset.batch(batch_size)
test_dataset = test_dataset.repeat(epochs_repeats)
num_classes = 10
base_model = EfficientNet.EfficientNetB3(include_top=False, input_shape=(300, 300, 3), classes=num_classes)
x = base_model.output
x = GlobalAveragePooling2D(name='avg_pool')(x)
predictions = Dense(num_classes, activation='softmax')(x)
model = Model(inputs=base_model.input, outputs=predictions)
model.load_weights("./ckpt/weights_09-0.00.h5")
y_predict = model.predict(test_dataset)
sparse_categorical_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()
sparse_categorical_accuracy.update_state(y_true=lable_list, y_pred=y_predict)
print("測試集的準確率爲: %f" % sparse_categorical_accuracy.result())
return
if __name__ == '__main__':
# model = get_model()
#常規的用 examples文件夾中的圖片來測試
# train_model_1(model)
#使用 sklearn的 train_test_split 來進行數據集分割
# train_model_2(model)
#使用 ImageDataGenerator 來直接讀取各個類別文件中的圖片
# train_model_3(model)
#使用 vgg16 進行遷移學習
# train_model_4()
"""
1.tf.nn.swish(x) 等同於把 x * tf.sigmoid(beta * x) 封裝了。
如果使用了tf.nn.swish(x) 則需要同時使用tf.compat.v1.disable_eager_execution()。
如果使用x * tf.sigmoid(beta * x)來代替tf.nn.swish(x)的話，則可以不使用tf.compat.v1.disable_eager_execution()。
2.但注意此處可能環境問題使用tf.nn.swish(x)的話會報錯，所以此處使用x * tf.sigmoid(beta * x)來代替tf.nn.swish(x)
報錯信息如下：
tensorflow/core/grappler/utils/graph_view.cc:830] No registered 'swish_f32' OpKernel for GPU devices compatible with node
{{node swish_75/swish_f32}} Registered: <no registered kernels>
3.使用 inputs * tf.math.sigmoid(inputs) 之後 就不用 tf.compat.v1.disable_eager_execution() 了，
TensorBoard 那就也可以正常使用了，因爲此處同時使用 tf.compat.v1.disable_eager_execution() 和 TensorBoard
的時候，TensorBoard會報錯：tensorflow.python.eager.core._FallbackException: This function does not handle the case of the path where all inputs are not already EagerTensors.
"""
# EfficientNet 必須注意關閉默認的eager模式
# tf.compat.v1.disable_eager_execution()
#使用 EfficientNet模型 進行遷移學習
train_model_5()
# 使用 EfficientNet模型 進行 預測
# predict_model_5()
使用EfficientNet模型遷移 第二個版本
from day05.homework.efficientnet import model as EfficientNet
import tensorflow as tf
from tensorflow.keras.layers import Dense, GlobalAveragePooling2D
from tensorflow.keras.models import Model
from day05.homework.data_gen import data_from_sequence
from day05.homework.utils.lr_scheduler import WarmUpCosineDecayScheduler
num_epochs = 10
batch_size = 10
learning_rate = 0.001
Dataset_dir = './Sign-Language-Digits-Dataset-master/Dataset/'
Dataset_test_dir = './Sign-Language-Digits-Dataset-master/Examples/'
num_classes = 10
input_size = 300
Dataset_dir_list = []
Dataset_test_dir_list = []
lable_list = []
train_img_fileNames = []
train_img_lables = []
test_img_fileNames = []
test_img_lables = []
lr=1e-3
optimizer = tf.keras.optimizers.Adam(lr) #0.001
sparse_categorical_crossentropy = tf.keras.losses.sparse_categorical_crossentropy
#['accuracy']/['acc'] 可以代替 [tf.keras.metrics.sparse_categorical_accuracy]，
# 因爲['accuracy']/['acc']底層都會默認適配爲[tf.keras.metrics.sparse_categorical_accuracy]
sparse_categorical_accuracy = [tf.keras.metrics.sparse_categorical_accuracy]
"""
metrics 是 sparse_categorical_crossentropy，則ModelCheckpoint中的第一個路勁字符串信息中和monitor參數中都需要用val_sparse_categorical_crossentropy
metrics 是 acc，則ModelCheckpoint中的第一個路勁字符串信息中和monitor參數中都需要用val_acc
metrics 是 accuracy，則ModelCheckpoint中的第一個路勁字符串信息中和monitor參數中都需要用val_accuracy
#['accuracy']/['acc'] 可以代替 [tf.keras.metrics.sparse_categorical_accuracy]，
#因爲['accuracy']/['acc']底層都會默認適配爲[tf.keras.metrics.sparse_categorical_accuracy]
"""
def train_model():
# 1、讀取sequence數據
train_sequence, validation_sequence = data_from_sequence(Dataset_dir, batch_size, num_classes, input_size)
# ================================================================
# 還要預先建立 ./graph/train/plugins/profile 的目錄
# 要在graph的同級目錄下 使用命令 tensorboard --logdir=graph 或者 tensorboard --logdir='./graph'
# http://localhost:6006/ 訪問
tensorboard = tf.keras.callbacks.TensorBoard(log_dir='./graph', histogram_freq=1, write_graph=True,
write_images=True)
# 預先建立./ckpt目錄
check = tf.keras.callbacks.ModelCheckpoint('./ckpt/weights_{epoch:02d}-{val_loss:.2f}.h5',
monitor='val_loss',
save_best_only=True,
save_weights_only=False,
mode='auto',
period=1)
# ================================================================
base_model = EfficientNet.EfficientNetB3(include_top=False, input_shape=(300, 300, 3), classes=num_classes)
x = base_model.output
x = GlobalAveragePooling2D(name='avg_pool')(x)
predictions = Dense(num_classes, activation='softmax')(x)
model = Model(inputs=base_model.input, outputs=predictions)
# model.compile(loss=sparse_categorical_crossentropy, optimizer=optimizer, metrics=sparse_categorical_accuracy)
model.compile(loss='categorical_crossentropy', optimizer=optimizer, metrics= ['accuracy'])
model.summary()
# ================================================================
# 餘弦退回warmup
# 得到總樣本數
sample_count = len(train_sequence)
# 第二階段學習率以及總步數
learning_rate_base = lr
max_epochs = 10
total_steps = int(max_epochs * sample_count) / batch_size
# 計算第一階段的步數需要多少 warmup_steps
warmup_epoch = 5
warmup_steps = int(warmup_epoch * sample_count) / batch_size
print("sample_count:",sample_count) #185
print("total_steps:",total_steps) #185.0
print("warmup_steps:",warmup_steps) #92.5
warm_lr = WarmUpCosineDecayScheduler(learning_rate_base=learning_rate_base,
total_steps=total_steps,
warmup_learning_rate=0,
warmup_steps=warmup_steps,
hold_base_rate_steps=0)
# ================================================================
"""
要求訓練樣本數/標籤數和驗證樣本/標籤數數都必須是可以整除batch_size的，
否則在mixup中X = X1 * X_l + X2 * (1 - X_l)代碼再進行矩陣廣播運算的時候會報錯，因爲形狀不一致，
報錯信息：ValueError: operands could not be broadcast together with shapes (不滿batch_size大小,300,300,3) (batch_size,1,1,1)。
正因爲如此，所以要求訓練樣本數和驗證樣本數都必須是可以整除batch_size的，因爲在一個epoch中遍歷到最後一個step時，
剩餘的訓練樣本數不足以整除batch size，那麼就會報錯。
"""
"""
因爲手勢識別的數據集過小，因此不要使用warm_lr(warmup以及餘弦退火學習)、數據增強(隨機擦處/ImageDataGenerator)、mixup，
這樣loss才能快速降低，準確率才能上升。
"""
# 4、訓練步驟
model.fit_generator(
train_sequence,
steps_per_epoch=int(sample_count / batch_size), # 一個epoch需要多少步 ， 1epoch sample_out 140000多樣本， 140000 / 16 = 步數
epochs=max_epochs,
verbose=1,
callbacks=[check, tensorboard],
# callbacks=[check, tensorboard, warm_lr],
validation_data=validation_sequence,
# max_queue_size=10,
# workers=int(multiprocessing.cpu_count() * 0.7),
# use_multiprocessing=True,
# shuffle=True
)
return None
if __name__ == '__main__':
"""
1.tf.nn.swish(x) 等同於把 x * tf.sigmoid(beta * x) 封裝了。
如果使用了tf.nn.swish(x) 則需要同時使用tf.compat.v1.disable_eager_execution()。
如果使用x * tf.sigmoid(beta * x)來代替tf.nn.swish(x)的話，則可以不使用tf.compat.v1.disable_eager_execution()。
2.但注意此處可能環境問題使用tf.nn.swish(x)的話會報錯，所以此處使用x * tf.sigmoid(beta * x)來代替tf.nn.swish(x)
報錯信息如下：
tensorflow/core/grappler/utils/graph_view.cc:830] No registered 'swish_f32' OpKernel for GPU devices compatible with node
{{node swish_75/swish_f32}} Registered: <no registered kernels>
3.使用 inputs * tf.math.sigmoid(inputs) 之後 就不用 tf.compat.v1.disable_eager_execution() 了，
TensorBoard 那就也可以正常使用了，因爲此處同時使用 tf.compat.v1.disable_eager_execution() 和 TensorBoard
的時候，TensorBoard會報錯：tensorflow.python.eager.core._FallbackException: This function does not handle the case of the path where all inputs are not already EagerTensors.
"""
# EfficientNet 必須注意關閉默認的eager模式
# tf.compat.v1.disable_eager_execution()
# 使用 EfficientNet模型 進行遷移學習
train_model()
data_gen文件夾 processing_data.py
import math
import os
import random
import numpy as np
from PIL import Image
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.utils import to_categorical, Sequence
from sklearn.model_selection import train_test_split
from data_gen.random_eraser import get_random_eraser
class GarbageDataSequence(Sequence):
"""垃圾分類數據流，每次batch返回batch_size大小數據
model.fit_generator使用
"""
def __init__(self, img_paths, labels, batch_size, img_size, use_aug):
print("20:",np.array(img_paths).shape) #batch_size
print("21:",np.array(labels).shape) #[batch_size, 10]
# 1、獲取訓練特徵與目標值的合併結果 np.hstack( [batch_size, 1], [batch_size, 40]) 結果爲 [batch_size, 41]
self.x_y = np.hstack((np.array(img_paths).reshape(len(img_paths), 1), np.array(labels)))
print("22:",self.x_y.shape)
print("23:",self.x_y)
self.batch_size = batch_size
self.img_size = img_size # (300, 300)
self.use_aug = use_aug
self.alpha = 0.2
# 隨機擦出方法
self.eraser = get_random_eraser(s_h=0.3, pixel_level=True)
def __len__(self):
return math.ceil(len(self.x_y) / self.batch_size)
@staticmethod
def center_img(img, size=None, fill_value=255):
"""改變圖片尺寸到300x300，並且做填充使得圖像處於中間位置
"""
h, w = img.shape[:2]
if size is None:
size = max(h, w)
shape = (size, size) + img.shape[2:]
background = np.full(shape, fill_value, np.uint8)
center_x = (size - w) // 2
center_y = (size - h) // 2
background[center_y:center_y + h, center_x:center_x + w] = img
return background
def preprocess_img(self, img_path):
"""處理每張圖片，大小， 數據增強
:param img_path:
:return:
"""
# 1、讀取圖片對應內容，做形狀，內容處理, (h, w)
img = Image.open(img_path)
# [180, 200, 3]
scale = self.img_size[0] / max(img.size[:2])
img = img.resize((int(img.size[0] * scale), int(img.size[1] * scale)))
img = img.convert('RGB')
img = np.array(img)
# 2、數據增強：如果是訓練集進行數據增強操作
# if self.use_aug:
#
# # 1、隨機擦處
# img = self.eraser(img)
#
# # 2、翻轉
# datagen = ImageDataGenerator(
# width_shift_range=0.05,
# height_shift_range=0.05,
# horizontal_flip=True,
# vertical_flip=True,
# )
# img = datagen.random_transform(img)
# 4、處理一下形狀 【300， 300， 3】
# 改變到[300, 300] 建議不要進行裁剪操作，變形操作，保留數據增強之後的效果，填充到300x300
img = self.center_img(img, self.img_size[0])
return img
def __getitem__(self, idx):
# 1、獲取當前批次idx對應的特徵值和目標值
batch_x = self.x_y[idx * self.batch_size: self.batch_size * (idx + 1), 0]
batch_y = self.x_y[idx * self.batch_size: self.batch_size * (idx + 1), 1:]
# print("batch_x:",batch_x)
# print("batch_y:",batch_y)
batch_x = np.array([self.preprocess_img(img_path) for img_path in batch_x])
batch_y = np.array(batch_y).astype(np.float32)
# 2、mixup
# batch_x, batch_y = self.mixup(batch_x, batch_y)
# 3、歸一化處理
batch_x = self.preprocess_input(batch_x)
return batch_x, batch_y
def on_epoch_end(self):
np.random.shuffle(self.x_y)
def mixup(self, batch_x, batch_y):
"""
數據混合mixup
:param batch_x: 要mixup的batch_X
:param batch_y: 要mixup的batch_y
:return: mixup後的數據
"""
size = self.batch_size
l = np.random.beta(self.alpha, self.alpha, size)
X_l = l.reshape(size, 1, 1, 1)
y_l = l.reshape(size, 1)
X1 = batch_x
Y1 = batch_y
X2 = batch_x[::-1]
Y2 = batch_y[::-1]
X = X1 * X_l + X2 * (1 - X_l)
Y = Y1 * y_l + Y2 * (1 - y_l)
return X, Y
def preprocess_input(self, x):
"""歸一化處理樣本特徵值
:param x:
:return:
"""
assert x.ndim in (3, 4)
assert x.shape[-1] == 3
MEAN_RGB = [0.485 * 255, 0.456 * 255, 0.406 * 255]
STDDEV_RGB = [0.229 * 255, 0.224 * 255, 0.225 * 255]
x = x - np.array(MEAN_RGB)
x = x / np.array(STDDEV_RGB)
return x
def smooth_labels(y, smooth_factor=0.1):
assert len(y.shape) == 2
if 0 <= smooth_factor <= 1:
y *= 1 - smooth_factor
y += smooth_factor / y.shape[1]
else:
raise Exception(
'Invalid label smoothing factor: ' + str(smooth_factor))
return y
def data_from_sequence(train_data_dir, batch_size, num_classes, input_size):
"""讀取本地數據到sequence
:param train_data_dir: 訓練數據目錄
:param batch_size: 批次大小
:param num_classes: 總類別書40
:param input_size: 輸入圖片大小(300, 300)
:return:
"""
# 1、讀取txt文件，打亂文件順序, .jpg, .txt
# label_files = [os.path.join(train_data_dir, filename) for filename
# in os.listdir(train_data_dir) if filename.endswith('.txt')]
# random.shuffle(label_files)
# # 2、解析txt文件當中 特徵值以及目標值（標籤）
# img_paths = []
# labels = []
#
# for index, file_path in enumerate(label_files):
# with open(file_path, 'r') as f:
# line = f.readline()
#
# line_split = line.strip().split(', ')
# # line '*.jpg, 0'
# if len(line_split) != 2:
# print("% 文件格式出錯", (file_path))
# continue
#
# img_name = line_split[0]
# label = int(line_split[1])
#
# # 最後保存到所有的列表當中
# img_paths.append(os.path.join(train_data_dir, img_name))
# labels.append(label)
# print(img_paths, labels)
print("batch_size:",batch_size)
print("num_classes:",num_classes)
print("input_size:",input_size)
lable = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
lable_list = []
Dataset_dir_list = []
for i in lable:
temp_list = []
for filename in os.listdir(train_data_dir + i):
temp_list.append(train_data_dir + str(i) + "/" + filename)
Dataset_dir_list.append(temp_list)
lable_list.append(np.full(len(temp_list), int(i))) # np.full(5,10)
print(len(lable_list)) # 10
print(len(Dataset_dir_list)) # 10
train_img_fileNames = []
train_img_lables = []
test_img_fileNames = []
test_img_lables = []
for i in lable:
# 3、目標標籤類別ont_hot編碼轉換， 平滑處理
labels = to_categorical(lable_list[int(i)], num_classes)
print("labels:", labels.shape)
labels = smooth_labels(labels)
print("smooth_labels:", labels.shape)
"""
必須使用expand_dims在第一維，即每個路徑字符串都封裝多一層作爲一個數組，那麼第一維每個數組中都只有一個路徑字符串。
這樣做才能配合二維的labels在train_test_split函數中進行切割使用
"""
img = np.expand_dims(Dataset_dir_list[int(i)], axis=1)
print("img.shape:",img.shape)
# print("img:",img)
x_train, x_test, y_train, y_test = train_test_split(img, labels, test_size=0.1, random_state=22)
# x_train, x_test, y_train, y_test = train_test_split(Dataset_dir_list[int(i)], lable_list[int(i)], test_size=0.1, random_state=22)
train_img_fileNames.append(x_train)
train_img_lables.append(y_train)
test_img_fileNames.append(x_test)
test_img_lables.append(y_test)
print("1:",len(train_img_fileNames)) #10
print("2:",len(train_img_lables)) #10
print("3:",len(test_img_fileNames)) #10
print("4:",len(test_img_lables)) #10
# print("5:",train_img_fileNames[0]) # 整個是 三維數組，取出來的[0] 是二維數組
# print("6:",train_img_lables[0]) #整個是 三維數組，取出來的[0] 是二維數組
# print("7:",test_img_fileNames[0]) #整個是 三維數組，取出來的[0] 是二維數組
# print("8:",test_img_lables[0]) #整個是 三維數組，取出來的[0] 是二維數組
Dataset_dir_lists = [f1 for train_img in train_img_fileNames for f in train_img for f1 in f]
lable_lists = [f for train_lable in train_img_lables for f in train_lable]
Dataset_dir_lists_test = [f1 for test_img in test_img_fileNames for f in test_img for f1 in f]
lable_lists_test = [f for test_lables in test_img_lables for f in test_lables]
print("9:",len(Dataset_dir_lists)) #1852
print("10:",len(lable_lists))#1852
print("11:",len(Dataset_dir_lists_test))#210
print("12:",len(lable_lists_test)) #210
print("13:",Dataset_dir_lists[0])
print("14:",lable_lists[0])
print("15:",Dataset_dir_lists_test[0])
print("16:",lable_lists_test[0])
# # 3、目標標籤類別ont_hot編碼轉換， 平滑處理
# labels = to_categorical(lable_list, num_classes)
# labels = smooth_labels(labels)
# 分割訓練集合驗證集合
# train_img_paths, validation_img_paths, train_labels, validation_labels \
# = train_test_split(Dataset_dir_list, labels, test_size=0.15, random_state=0)
# print(validation_img_paths)
# print(train_labels, validation_labels)
print("總共樣本數： %d , 訓練樣本數： %d, 驗證樣本數： %d" % (len(Dataset_dir_lists)+len(Dataset_dir_lists_test),
len(Dataset_dir_lists), len(Dataset_dir_lists_test)))
"""
要求訓練樣本數/標籤數和驗證樣本/標籤數數都必須是可以整除batch_size的，
否則在mixup中X = X1 * X_l + X2 * (1 - X_l)代碼再進行矩陣廣播運算的時候會報錯，因爲形狀不一致，
報錯信息：ValueError: operands could not be broadcast together with shapes (不滿batch_size大小,300,300,3) (batch_size,1,1,1)。
正因爲如此，所以要求訓練樣本數和驗證樣本數都必須是可以整除batch_size的，因爲在一個epoch中遍歷到最後一個step時，
剩餘的訓練樣本數不足以整除batch size，那麼就會報錯。
"""
i = -2
j = -2
while 1:
if (len(Dataset_dir_lists) % batch_size == 0):
break
else:
Dataset_dir_lists = Dataset_dir_lists[:i]
lable_lists = lable_lists[:i]
i -= 1
while 1:
if (len(Dataset_dir_lists_test) % batch_size == 0):
break
else:
Dataset_dir_lists_test = Dataset_dir_lists[:j]
lable_lists_test = lable_lists_test[:j]
j -= 1
print("處理後的總共樣本數： %d , 處理後的訓練樣本數： %d, 處理後的驗證樣本數： %d, 處理後的訓練樣本標籤數： %d, 處理後的驗證樣本標籤數： %d" %
(len(Dataset_dir_lists)+len(Dataset_dir_lists_test), len(Dataset_dir_lists), len(Dataset_dir_lists_test)
, len(lable_lists), len(lable_lists_test)
))
# 4、Sequence調用測試
train_sequence = GarbageDataSequence(Dataset_dir_lists, lable_lists, batch_size, [input_size, input_size], use_aug=True)
validation_sequence = GarbageDataSequence(Dataset_dir_lists_test, lable_lists_test, batch_size, [input_size, input_size], use_aug=False)
return train_sequence, validation_sequence
# if __name__ == '__main__':
# train_data_dir = "../data/garbage_classify/train_data"
# batch_size = 32
#
# train_sequence, validation_sequence = data_from_sequence(train_data_dir, batch_size, num_classes=40, input_size=300)
#
# for i in range(100):
# print("第 %d 批次數據" % i)
# batch_x, batch_y = train_sequence.__getitem__(i)
# print(batch_x, batch_y)
data_gen文件夾 random_eraser.py
import numpy as np
import tensorflow as tf
def get_random_eraser(p=0.5, s_l=0.02, s_h=0.4, r_1=0.3, r_2=1/0.3, v_l=0, v_h=255, pixel_level=False):
def eraser(input_img):
img_h, img_w, img_c = input_img.shape
p_1 = np.random.rand()
if p_1 > p:
return input_img
while True:
s = np.random.uniform(s_l, s_h) * img_h * img_w
r = np.random.uniform(r_1, r_2)
w = int(np.sqrt(s / r))
h = int(np.sqrt(s * r))
left = np.random.randint(0, img_w)
top = np.random.randint(0, img_h)
if left + w <= img_w and top + h <= img_h:
break
if pixel_level:
c = np.random.uniform(v_l, v_h, (h, w, img_c))
else:
c = np.random.uniform(v_l, v_h)
input_img[top:top + h, left:left + w, :] = c
return input_img
return eraser
utils文件夾 lr_scheduler.py
import numpy as np
import tensorflow as tf
from tensorflow.keras import backend as K
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
import os
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"
def cosine_decay_with_warmup(global_step,
learning_rate_base,
total_steps,
warmup_learning_rate=0.0,
warmup_steps=0,
hold_base_rate_steps=0):
"""
每批次帶有warmup餘弦退火學習率計算
:param global_step: 當前到達的步數
:param learning_rate_base: warmup之後的基礎學習率
:param total_steps: 總需要批次數
:param warmup_learning_rate: warmup開始的學習率
:param warmup_steps:warmup學習率 步數
:param hold_base_rate_steps: 預留總步數和warmup步數間隔
:return:
"""
if total_steps < warmup_steps:
raise ValueError("總步數要大於wamup步數")
# 1、餘弦退火學習率計算
learning_rate = 0.5 * learning_rate_base * (1 + np.cos(
np.pi * (global_step - warmup_steps - hold_base_rate_steps) / float(total_steps - warmup_steps - hold_base_rate_steps)
))
# 2、warmup之後的學習率計算
# 預留步數階段
# 如果預留大於0，判斷目前步數是否 > warmup步數+預留步數，是的話返回剛纔上面計算的學習率，不是的話使用warmup之後的基礎學習率
learning_rate = np.where(global_step > warmup_steps + hold_base_rate_steps, learning_rate, learning_rate_base)
# 3、warmup學習率計算，並判斷大小
# 第一個階段的學習率計算
if warmup_steps > 0:
if learning_rate_base < warmup_learning_rate:
raise ValueError("第二階段學習率要大於第一階段學習率")
slope = (learning_rate_base - warmup_learning_rate) / warmup_steps
warmup_rate = slope * global_step + warmup_learning_rate
learning_rate = np.where(global_step < warmup_steps, warmup_rate, learning_rate)
# 4、如果最後當前到達的步數大於總步數，則歸0，否則返回當前的計算出來的學習率（可能是warmup學習率也可能是餘弦衰減結果）
return np.where(global_step > total_steps, 0.0, learning_rate)
class WarmUpCosineDecayScheduler(tf.keras.callbacks.Callback):
"""帶有warnup的餘弦退火學習率實現
"""
def __init__(self,
learning_rate_base,
total_steps,
global_step_init=0,
warmup_learning_rate=0.0,
warmup_steps=0,
hold_base_rate_steps=0,
verbose=0):
super(WarmUpCosineDecayScheduler, self).__init__()
self.learning_rate_base = learning_rate_base
self.total_steps = total_steps
self.global_step = global_step_init
self.warmup_learning_rate = warmup_learning_rate
self.warmup_steps = warmup_steps
self.hold_base_rate_steps = hold_base_rate_steps
# 是否在每次訓練結束打印學習率
self.verbose = verbose
# 記錄所有批次下來的每次準確的學習率，可以用於打印顯示
self.learning_rates = []
def on_batch_end(self, batch, logs=None):
# 記錄當前訓練到走到第幾步數
self.global_step = self.global_step + 1
# 記錄下所有每次的學習到列表，要統計畫圖可以使用
lr = K.get_value(self.model.optimizer.lr)
self.learning_rates.append(lr)
def on_batch_begin(self, batch, logs=None):
# 計算這批次開始的學習率 lr
lr = cosine_decay_with_warmup(global_step=self.global_step,
learning_rate_base=self.learning_rate_base,
total_steps=self.total_steps,
warmup_learning_rate=self.warmup_learning_rate,
warmup_steps=self.warmup_steps,
hold_base_rate_steps=self.hold_base_rate_steps)
# 設置模型的學習率爲lr
K.set_value(self.model.optimizer.lr, lr)
if self.verbose > 0:
print('\n批次數 %05d: 設置學習率爲'
' %s.' % (self.global_step + 1, lr))
if __name__ == '__main__':
# 1、創建模型
model = Sequential()
model.add(Dense(32, activation='relu', input_dim=100))
model.add(Dense(10, activation='softmax'))
model.compile(optimizer='rmsprop',
loss='categorical_crossentropy',
metrics=['accuracy'])
# 2、參數設置
sample_count = 1000 # 樣本數
epochs = 4 # 總迭代次數
warmup_epoch = 3 # warmup 迭代次數
batch_size = 16 # 批次大小
learning_rate_base = 0.0001 # warmup後的初始學習率
total_steps = int(epochs * sample_count / batch_size) # 總迭代批次步數
warmup_steps = int(warmup_epoch * sample_count / batch_size) # warmup總批次數
# 3、創建測試數據
data = np.random.random((sample_count, 100))
labels = np.random.randint(10, size=(sample_count, 1))
# 轉換目標類別
one_hot_labels = tf.keras.utils.to_categorical(labels, num_classes=10)
# 5、創建餘弦warmup調度器
warm_up_lr = WarmUpCosineDecayScheduler(learning_rate_base=learning_rate_base,
total_steps=total_steps,
warmup_learning_rate=4e-06, # warmup開始學習率
warmup_steps=warmup_steps,
hold_base_rate_steps=0,
)
# 訓練模型
model.fit(data, one_hot_labels, epochs=epochs, batch_size=batch_size, verbose=0, callbacks=[warm_up_lr])
print(warm_up_lr.learning_rates)
EfficientNet模型B0~B7 源碼
initializers.py
import numpy as np
import tensorflow as tf
import keras.backend as K
from keras.initializers import Initializer
from keras.utils.generic_utils import get_custom_objects
class EfficientConv2DKernelInitializer(Initializer):
"""Initialization for convolutional kernels.
The main difference with tf.variance_scaling_initializer is that
tf.variance_scaling_initializer uses a truncated normal with an uncorrected
standard deviation, whereas here we use a normal distribution. Similarly,
tf.contrib.layers.variance_scaling_initializer uses a truncated normal with
a corrected standard deviation.
Args:
shape: shape of variable
dtype: dtype of variable
partition_info: unused
Returns:
an initialization for the variable
"""
def __call__(self, shape, dtype=K.floatx(), **kwargs):
kernel_height, kernel_width, _, out_filters = shape
fan_out = int(kernel_height * kernel_width * out_filters)
return tf.random_normal(
shape, mean=0.0, stddev=np.sqrt(2.0 / fan_out), dtype=dtype)
class EfficientDenseKernelInitializer(Initializer):
"""Initialization for dense kernels.
This initialization is equal to
tf.variance_scaling_initializer(scale=1.0/3.0, mode='fan_out',
distribution='uniform').
It is written out explicitly here for clarity.
Args:
shape: shape of variable
dtype: dtype of variable
Returns:
an initialization for the variable
"""
def __call__(self, shape, dtype=K.floatx(), **kwargs):
"""Initialization for dense kernels.
This initialization is equal to
tf.variance_scaling_initializer(scale=1.0/3.0, mode='fan_out',
distribution='uniform').
It is written out explicitly here for clarity.
Args:
shape: shape of variable
dtype: dtype of variable
Returns:
an initialization for the variable
"""
init_range = 1.0 / np.sqrt(shape[1])
return tf.random_uniform(shape, -init_range, init_range, dtype=dtype)
conv_kernel_initializer = EfficientConv2DKernelInitializer()
dense_kernel_initializer = EfficientDenseKernelInitializer()
get_custom_objects().update({
'EfficientDenseKernelInitializer': EfficientDenseKernelInitializer,
'EfficientConv2DKernelInitializer': EfficientConv2DKernelInitializer,
})
layers.py
"""
import tensorflow as tf
import numpy as np
x=np.array([[1.,8.,7.],[10.,14.,3.],[1.,2.,4.]])
tf.math.sigmoid(x)
<tf.Tensor: id=22, shape=(3, 3), dtype=float64, numpy=
array([[0.73105858, 0.99966465, 0.99908895],
[0.9999546 , 0.99999917, 0.95257413],
[0.73105858, 0.88079708, 0.98201379]])>
tf.compat.v1.disable_eager_execution()
x=np.array([[1.,8.,7.],[10.,14.,3.],[1.,2.,4.]])
tf.nn.swish(x)
----------------------------------------------------------------
1.tf.nn.swish(x) 等同於把 x * tf.sigmoid(beta * x) 封裝了。
如果使用了tf.nn.swish(x) 則需要同時使用tf.compat.v1.disable_eager_execution()。
如果使用x * tf.sigmoid(beta * x)來代替tf.nn.swish(x)的話，則可以不使用tf.compat.v1.disable_eager_execution()。
2.但注意此處可能環境問題使用tf.nn.swish(x)的話會報錯，所以此處使用x * tf.sigmoid(beta * x)來代替tf.nn.swish(x)
報錯信息如下：
tensorflow/core/grappler/utils/graph_view.cc:830] No registered 'swish_f32' OpKernel for GPU devices compatible with node
{{node swish_75/swish_f32}} Registered: <no registered kernels>
"""
import tensorflow as tf
import tensorflow.keras.backend as K
import tensorflow.keras.layers as KL
from tensorflow.keras.utils import get_custom_objects
"""
import tensorflow as tf
import numpy as np
x=np.array([[1.,8.,7.],[10.,14.,3.],[1.,2.,4.]])
tf.math.sigmoid(x)
<tf.Tensor: id=22, shape=(3, 3), dtype=float64, numpy=
array([[0.73105858, 0.99966465, 0.99908895],
[0.9999546 , 0.99999917, 0.95257413],
[0.73105858, 0.88079708, 0.98201379]])>
tf.compat.v1.disable_eager_execution()
x=np.array([[1.,8.,7.],[10.,14.,3.],[1.,2.,4.]])
tf.nn.swish(x)
----------------------------------------------------------------
1.tf.nn.swish(x) 等同於把 x * tf.sigmoid(beta * x) 封裝了。
如果使用了tf.nn.swish(x) 則需要同時使用tf.compat.v1.disable_eager_execution()。
如果使用x * tf.sigmoid(beta * x)來代替tf.nn.swish(x)的話，則可以不使用tf.compat.v1.disable_eager_execution()。
2.但注意此處可能環境問題使用tf.nn.swish(x)的話會報錯，所以此處使用x * tf.sigmoid(beta * x)來代替tf.nn.swish(x)
報錯信息如下：
tensorflow/core/grappler/utils/graph_view.cc:830] No registered 'swish_f32' OpKernel for GPU devices compatible with node
{{node swish_75/swish_f32}} Registered: <no registered kernels>
"""
class Swish(KL.Layer):
def __init__(self, **kwargs):
super().__init__(**kwargs)
def call(self, inputs, **kwargs):
# return tf.nn.swish(inputs)
return inputs * tf.math.sigmoid(inputs)
class DropConnect(KL.Layer):
def __init__(self, drop_connect_rate=0., **kwargs):
super().__init__(**kwargs)
self.drop_connect_rate = drop_connect_rate
def call(self, inputs, training=None):
def drop_connect():
keep_prob = 1.0 - self.drop_connect_rate
# Compute drop_connect tensor
batch_size = tf.shape(inputs)[0]
random_tensor = keep_prob
random_tensor += tf.random.uniform([batch_size, 1, 1, 1], dtype=inputs.dtype)
binary_tensor = tf.floor(random_tensor)
output = tf.math.divide(inputs, keep_prob) * binary_tensor
return output
return K.in_train_phase(drop_connect, inputs, training=training)
def get_config(self):
config = super().get_config()
config['drop_connect_rate'] = self.drop_connect_rate
return config
get_custom_objects().update({
'DropConnect': DropConnect,
'Swish': Swish,
})
model.py
# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Contains definitions for EfficientNet model.
[1] Mingxing Tan, Quoc V. Le
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks.
ICML'19, https://arxiv.org/abs/1905.11946
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import collections
import math
import numpy as np
import six
from six.moves import xrange # pylint: disable=redefined-builtin
import tensorflow as tf
import tensorflow.keras.backend as K
import tensorflow.keras.models as KM
import tensorflow.keras.layers as KL
from tensorflow.keras.utils import get_file
from tensorflow.keras.initializers import Initializer
from .layers import Swish, DropConnect
from .params import get_model_params, IMAGENET_WEIGHTS
from .initializers import conv_kernel_initializer, dense_kernel_initializer
__all__ = ['EfficientNet', 'EfficientNetB0', 'EfficientNetB1', 'EfficientNetB2', 'EfficientNetB3',
'EfficientNetB4', 'EfficientNetB5', 'EfficientNetB6', 'EfficientNetB7']
class ConvKernalInitializer(Initializer):
def __call__(self, shape, dtype=K.floatx(), partition_info=None):
"""Initialization for convolutional kernels.
The main difference with tf.variance_scaling_initializer is that
tf.variance_scaling_initializer uses a truncated normal with an uncorrected
standard deviation, whereas here we use a normal distribution. Similarly,
tf.contrib.layers.variance_scaling_initializer uses a truncated normal with
a corrected standard deviation.
Args:
shape: shape of variable
dtype: dtype of variable
partition_info: unused
Returns:
an initialization for the variable
"""
del partition_info
kernel_height, kernel_width, _, out_filters = shape
fan_out = int(kernel_height * kernel_width * out_filters)
return tf.random.normal(
shape, mean=0.0, stddev=np.sqrt(2.0 / fan_out), dtype=dtype)
class DenseKernalInitializer(Initializer):
def __call__(self, shape, dtype=K.floatx(), partition_info=None):
"""Initialization for dense kernels.
This initialization is equal to
tf.variance_scaling_initializer(scale=1.0/3.0, mode='fan_out',
distribution='uniform').
It is written out explicitly here for clarity.
Args:
shape: shape of variable
dtype: dtype of variable
partition_info: unused
Returns:
an initialization for the variable
"""
del partition_info
init_range = 1.0 / np.sqrt(shape[1])
return tf.random_uniform(shape, -init_range, init_range, dtype=dtype)
def round_filters(filters, global_params):
"""Round number of filters based on depth multiplier."""
orig_f = filters
multiplier = global_params.width_coefficient
divisor = global_params.depth_divisor
min_depth = global_params.min_depth
if not multiplier:
return filters
filters *= multiplier
min_depth = min_depth or divisor
new_filters = max(min_depth, int(filters + divisor / 2) // divisor * divisor)
# Make sure that round down does not go down by more than 10%.
if new_filters < 0.9 * filters:
new_filters += divisor
# print('round_filter input={} output={}'.format(orig_f, new_filters))
return int(new_filters)
def round_repeats(repeats, global_params):
"""Round number of filters based on depth multiplier."""
multiplier = global_params.depth_coefficient
if not multiplier:
return repeats
return int(math.ceil(multiplier * repeats))
def SEBlock(block_args, global_params):
num_reduced_filters = max(
1, int(block_args.input_filters * block_args.se_ratio))
filters = block_args.input_filters * block_args.expand_ratio
if global_params.data_format == 'channels_first':
channel_axis = 1
spatial_dims = [2, 3]
else:
channel_axis = -1
spatial_dims = [1, 2]
def block(inputs):
x = inputs
x = KL.Lambda(lambda a: K.mean(a, axis=spatial_dims, keepdims=True))(x)
x = KL.Conv2D(
num_reduced_filters,
kernel_size=[1, 1],
strides=[1, 1],
kernel_initializer=ConvKernalInitializer(),
padding='same',
use_bias=True
)(x)
x = Swish()(x)
# Excite
x = KL.Conv2D(
filters,
kernel_size=[1, 1],
strides=[1, 1],
kernel_initializer=ConvKernalInitializer(),
padding='same',
use_bias=True
)(x)
x = KL.Activation('sigmoid')(x)
out = KL.Multiply()([x, inputs])
return out
return block
def MBConvBlock(block_args, global_params, drop_connect_rate=None):
batch_norm_momentum = global_params.batch_norm_momentum
batch_norm_epsilon = global_params.batch_norm_epsilon
if global_params.data_format == 'channels_first':
channel_axis = 1
spatial_dims = [2, 3]
else:
channel_axis = -1
spatial_dims = [1, 2]
has_se = (block_args.se_ratio is not None) and (
block_args.se_ratio > 0) and (block_args.se_ratio <= 1)
filters = block_args.input_filters * block_args.expand_ratio
kernel_size = block_args.kernel_size
def block(inputs):
if block_args.expand_ratio != 1:
x = KL.Conv2D(
filters,
kernel_size=[1, 1],
strides=[1, 1],
kernel_initializer=ConvKernalInitializer(),
padding='same',
use_bias=False
)(inputs)
x = KL.BatchNormalization(
axis=channel_axis,
momentum=batch_norm_momentum,
epsilon=batch_norm_epsilon
)(x)
x = Swish()(x)
else:
x = inputs
x = KL.DepthwiseConv2D(
[kernel_size, kernel_size],
strides=block_args.strides,
depthwise_initializer=ConvKernalInitializer(),
padding='same',
use_bias=False
)(x)
x = KL.BatchNormalization(
axis=channel_axis,
momentum=batch_norm_momentum,
epsilon=batch_norm_epsilon
)(x)
x = Swish()(x)
if has_se:
x = SEBlock(block_args, global_params)(x)
# output phase
x = KL.Conv2D(
block_args.output_filters,
kernel_size=[1, 1],
strides=[1, 1],
kernel_initializer=ConvKernalInitializer(),
padding='same',
use_bias=False
)(x)
x = KL.BatchNormalization(
axis=channel_axis,
momentum=batch_norm_momentum,
epsilon=batch_norm_epsilon
)(x)
if block_args.id_skip:
if all(
s == 1 for s in block_args.strides
) and block_args.input_filters == block_args.output_filters:
# only apply drop_connect if skip presents.
if drop_connect_rate:
x = DropConnect(drop_connect_rate)(x)
x = KL.Add()([x, inputs])
return x
return block
def EfficientNet(input_shape, block_args_list, global_params, include_top=True, pooling=None):
batch_norm_momentum = global_params.batch_norm_momentum
batch_norm_epsilon = global_params.batch_norm_epsilon
if global_params.data_format == 'channels_first':
channel_axis = 1
else:
channel_axis = -1
# Stem part
inputs = KL.Input(shape=input_shape)
x = inputs
x = KL.Conv2D(
filters=round_filters(32, global_params),
kernel_size=[3, 3],
strides=[2, 2],
kernel_initializer=ConvKernalInitializer(),
padding='same',
use_bias=False
)(x)
x = KL.BatchNormalization(
axis=channel_axis,
momentum=batch_norm_momentum,
epsilon=batch_norm_epsilon
)(x)
x = Swish()(x)
# Blocks part
block_idx = 1
n_blocks = sum([block_args.num_repeat for block_args in block_args_list])
drop_rate = global_params.drop_connect_rate or 0
drop_rate_dx = drop_rate / n_blocks
for block_args in block_args_list:
assert block_args.num_repeat > 0
# Update block input and output filters based on depth multiplier.
block_args = block_args._replace(
input_filters=round_filters(block_args.input_filters, global_params),
output_filters=round_filters(block_args.output_filters, global_params),
num_repeat=round_repeats(block_args.num_repeat, global_params)
)
# The first block needs to take care of stride and filter size increase.
x = MBConvBlock(block_args, global_params,
drop_connect_rate=drop_rate_dx * block_idx)(x)
block_idx += 1
if block_args.num_repeat > 1:
block_args = block_args._replace(input_filters=block_args.output_filters, strides=[1, 1])
for _ in xrange(block_args.num_repeat - 1):
x = MBConvBlock(block_args, global_params,
drop_connect_rate=drop_rate_dx * block_idx)(x)
block_idx += 1
# Head part
x = KL.Conv2D(
filters=round_filters(1280, global_params),
kernel_size=[1, 1],
strides=[1, 1],
kernel_initializer=ConvKernalInitializer(),
padding='same',
use_bias=False
)(x)
x = KL.BatchNormalization(
axis=channel_axis,
momentum=batch_norm_momentum,
epsilon=batch_norm_epsilon
)(x)
x = Swish()(x)
if include_top:
x = KL.GlobalAveragePooling2D(data_format=global_params.data_format)(x)
if global_params.dropout_rate > 0:
x = KL.Dropout(global_params.dropout_rate)(x)
x = KL.Dense(global_params.num_classes, kernel_initializer=DenseKernalInitializer())(x)
x = KL.Activation('softmax')(x)
else:
if pooling == 'avg':
x = KL.GlobalAveragePooling2D(data_format=global_params.data_format)(x)
elif pooling == 'max':
x = KL.GlobalMaxPooling2D(data_format=global_params.data_format)(x)
outputs = x
model = KM.Model(inputs, outputs)
return model
def _get_model_by_name(model_name, input_shape=None, include_top=True, weights=None, classes=1000, pooling=None):
"""Re-Implementation of EfficientNet for Keras
Reference:
https://arxiv.org/abs/1807.11626
Args:
input_shape: optional, if ``None`` default_input_shape is used
EfficientNetB0 - (224, 224, 3)
EfficientNetB1 - (240, 240, 3)
EfficientNetB2 - (260, 260, 3)
EfficientNetB3 - (300, 300, 3)
EfficientNetB4 - (380, 380, 3)
EfficientNetB5 - (456, 456, 3)
EfficientNetB6 - (528, 528, 3)
EfficientNetB7 - (600, 600, 3)
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet).
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
pooling: optional [None, 'avg', 'max'], if ``include_top=False``
add global pooling on top of the network
- avg: GlobalAveragePooling2D
- max: GlobalMaxPooling2D
Returns:
A Keras model instance.
"""
if weights not in {None, 'imagenet'}:
raise ValueError('Parameter `weights` should be one of [None, "imagenet"]')
if weights == 'imagenet' and model_name not in IMAGENET_WEIGHTS:
raise ValueError('There are not pretrained weights for {} model.'.format(model_name))
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` and `include_top`'
' `classes` should be 1000')
block_agrs_list, global_params, default_input_shape = get_model_params(
model_name, override_params={'num_classes': classes}
)
if input_shape is None:
input_shape = (default_input_shape, default_input_shape, 3)
model = EfficientNet(input_shape, block_agrs_list, global_params, include_top=include_top, pooling=pooling)
model._name = model_name
if weights:
if not include_top:
weights_name = model_name + '-notop'
else:
weights_name = model_name
weights = IMAGENET_WEIGHTS[weights_name]
weights_path = get_file(
weights['name'],
weights['url'],
cache_subdir='models',
md5_hash=weights['md5'],
)
model.load_weights(weights_path)
return model
def EfficientNetB0(include_top=True, input_shape=None, weights=None, classes=1000, pooling=None):
return _get_model_by_name('efficientnet-b0', include_top=include_top, input_shape=input_shape,
weights=weights, classes=classes, pooling=pooling)
def EfficientNetB1(include_top=True, input_shape=None, weights=None, classes=1000, pooling=None):
return _get_model_by_name('efficientnet-b1', include_top=include_top, input_shape=input_shape,
weights=weights, classes=classes, pooling=pooling)
def EfficientNetB2(include_top=True, input_shape=None, weights=None, classes=1000, pooling=None):
return _get_model_by_name('efficientnet-b2', include_top=include_top, input_shape=input_shape,
weights=weights, classes=classes, pooling=pooling)
def EfficientNetB3(include_top=True, input_shape=None, weights=None, classes=1000, pooling=None):
return _get_model_by_name('efficientnet-b3', include_top=include_top, input_shape=input_shape,
weights=weights, classes=classes, pooling=pooling)
def EfficientNetB4(include_top=True, input_shape=None, weights=None, classes=1000, pooling=None):
return _get_model_by_name('efficientnet-b4', include_top=include_top, input_shape=input_shape,
weights=weights, classes=classes, pooling=pooling)
def EfficientNetB5(include_top=True, input_shape=None, weights=None, classes=1000, pooling=None):
return _get_model_by_name('efficientnet-b5', include_top=include_top, input_shape=input_shape,
weights=weights, classes=classes, pooling=pooling)
def EfficientNetB6(include_top=True, input_shape=None, weights=None, classes=1000, pooling=None):
return _get_model_by_name('efficientnet-b6', include_top=include_top, input_shape=input_shape,
weights=weights, classes=classes, pooling=pooling)
def EfficientNetB7(include_top=True, input_shape=None, weights=None, classes=1000, pooling=None):
return _get_model_by_name('efficientnet-b7', include_top=include_top, input_shape=input_shape,
weights=weights, classes=classes, pooling=pooling)
EfficientNetB0.__doc__ = _get_model_by_name.__doc__
EfficientNetB1.__doc__ = _get_model_by_name.__doc__
EfficientNetB2.__doc__ = _get_model_by_name.__doc__
EfficientNetB3.__doc__ = _get_model_by_name.__doc__
EfficientNetB4.__doc__ = _get_model_by_name.__doc__
EfficientNetB5.__doc__ = _get_model_by_name.__doc__
EfficientNetB6.__doc__ = _get_model_by_name.__doc__
EfficientNetB7.__doc__ = _get_model_by_name.__doc__
params.py
import os
import re
import collections
IMAGENET_WEIGHTS = {
'efficientnet-b0': {
'name': 'efficientnet-b0_imagenet_1000.h5',
'url': 'https://github.com/qubvel/efficientnet/releases/download/v0.0.1/efficientnet-b0_imagenet_1000.h5',
'md5': 'bca04d16b1b8a7c607b1152fe9261af7',
},
'efficientnet-b0-notop': {
'name': 'efficientnet-b0_imagenet_1000_notop.h5',
'url': 'https://github.com/qubvel/efficientnet/releases/download/v0.0.1/efficientnet-b0_imagenet_1000_notop.h5',
'md5': '45d2f3b6330c2401ef66da3961cad769',
},
'efficientnet-b1': {
'name': 'efficientnet-b1_imagenet_1000.h5',
'url': 'https://github.com/qubvel/efficientnet/releases/download/v0.0.1/efficientnet-b1_imagenet_1000.h5',
'md5': 'bd4a2b82f6f6bada74fc754553c464fc',
},
'efficientnet-b1-notop': {
'name': 'efficientnet-b1_imagenet_1000_notop.h5',
'url': 'https://github.com/qubvel/efficientnet/releases/download/v0.0.1/efficientnet-b1_imagenet_1000_notop.h5',
'md5': '884aed586c2d8ca8dd15a605ec42f564',
},
'efficientnet-b2': {
'name': 'efficientnet-b2_imagenet_1000.h5',
'url': 'https://github.com/qubvel/efficientnet/releases/download/v0.0.1/efficientnet-b2_imagenet_1000.h5',
'md5': '45b28b26f15958bac270ab527a376999',
},
'efficientnet-b2-notop': {
'name': 'efficientnet-b2_imagenet_1000_notop.h5',
'url': 'https://github.com/qubvel/efficientnet/releases/download/v0.0.1/efficientnet-b2_imagenet_1000_notop.h5',
'md5': '42fb9f2d9243d461d62b4555d3a53b7b',
},
'efficientnet-b3': {
'name': 'efficientnet-b3_imagenet_1000.h5',
'url': 'https://github.com/qubvel/efficientnet/releases/download/v0.0.1/efficientnet-b3_imagenet_1000.h5',
'md5': 'decd2c8a23971734f9d3f6b4053bf424',
},
'efficientnet-b3-notop': {
'name': 'efficientnet-b3_imagenet_1000_notop.h5',
'url': 'https://github.com/qubvel/efficientnet/releases/download/v0.0.1/efficientnet-b3_imagenet_1000_notop.h5',
'md5': '1f7d9a8c2469d2e3d3b97680d45df1e1',
},
}
GlobalParams = collections.namedtuple('GlobalParams', [
'batch_norm_momentum', 'batch_norm_epsilon', 'dropout_rate', 'data_format',
'num_classes', 'width_coefficient', 'depth_coefficient',
'depth_divisor', 'min_depth', 'drop_connect_rate',
])
GlobalParams.__new__.__defaults__ = (None,) * len(GlobalParams._fields)
BlockArgs = collections.namedtuple('BlockArgs', [
'kernel_size', 'num_repeat', 'input_filters', 'output_filters',
'expand_ratio', 'id_skip', 'strides', 'se_ratio'
])
# defaults will be a public argument for namedtuple in Python 3.7
# https://docs.python.org/3/library/collections.html#collections.namedtuple
BlockArgs.__new__.__defaults__ = (None,) * len(BlockArgs._fields)
def efficientnet_params(model_name):
"""Get efficientnet params based on model name."""
params_dict = {
# (width_coefficient, depth_coefficient, resolution, dropout_rate)
'efficientnet-b0': (1.0, 1.0, 224, 0.2),
'efficientnet-b1': (1.0, 1.1, 240, 0.2),
'efficientnet-b2': (1.1, 1.2, 260, 0.3),
'efficientnet-b3': (1.2, 1.4, 300, 0.3),
'efficientnet-b4': (1.4, 1.8, 380, 0.4),
'efficientnet-b5': (1.6, 2.2, 456, 0.4),
'efficientnet-b6': (1.8, 2.6, 528, 0.5),
'efficientnet-b7': (2.0, 3.1, 600, 0.5),
}
return params_dict[model_name]
class BlockDecoder(object):
"""Block Decoder for readability."""
def _decode_block_string(self, block_string):
"""Gets a block through a string notation of arguments."""
assert isinstance(block_string, str)
ops = block_string.split('_')
options = {}
for op in ops:
splits = re.split(r'(\d.*)', op)
if len(splits) >= 2:
key, value = splits[:2]
options[key] = value
if 's' not in options or len(options['s']) != 2:
raise ValueError('Strides options should be a pair of integers.')
return BlockArgs(
kernel_size=int(options['k']),
num_repeat=int(options['r']),
input_filters=int(options['i']),
output_filters=int(options['o']),
expand_ratio=int(options['e']),
id_skip=('noskip' not in block_string),
se_ratio=float(options['se']) if 'se' in options else None,
strides=[int(options['s'][0]), int(options['s'][1])])
def _encode_block_string(self, block):
"""Encodes a block to a string."""
args = [
'r%d' % block.num_repeat,
'k%d' % block.kernel_size,
's%d%d' % (block.strides[0], block.strides[1]),
'e%s' % block.expand_ratio,
'i%d' % block.input_filters,
'o%d' % block.output_filters
]
if block.se_ratio > 0 and block.se_ratio <= 1:
args.append('se%s' % block.se_ratio)
if block.id_skip is False:
args.append('noskip')
return '_'.join(args)
def decode(self, string_list):
"""Decodes a list of string notations to specify blocks inside the network.
Args:
string_list: a list of strings, each string is a notation of block.
Returns:
A list of namedtuples to represent blocks arguments.
"""
assert isinstance(string_list, list)
blocks_args = []
for block_string in string_list:
blocks_args.append(self._decode_block_string(block_string))
return blocks_args
def encode(self, blocks_args):
"""Encodes a list of Blocks to a list of strings.
Args:
blocks_args: A list of namedtuples to represent blocks arguments.
Returns:
a list of strings, each string is a notation of block.
"""
block_strings = []
for block in blocks_args:
block_strings.append(self._encode_block_string(block))
return block_strings
def efficientnet(width_coefficient=None,
depth_coefficient=None,
dropout_rate=0.2,
drop_connect_rate=0.2):
"""Creates a efficientnet model."""
blocks_args = [
'r1_k3_s11_e1_i32_o16_se0.25', 'r2_k3_s22_e6_i16_o24_se0.25',
'r2_k5_s22_e6_i24_o40_se0.25', 'r3_k3_s22_e6_i40_o80_se0.25',
'r3_k5_s11_e6_i80_o112_se0.25', 'r4_k5_s22_e6_i112_o192_se0.25',
'r1_k3_s11_e6_i192_o320_se0.25',
]
global_params = GlobalParams(
batch_norm_momentum=0.99,
batch_norm_epsilon=1e-3,
dropout_rate=dropout_rate,
drop_connect_rate=drop_connect_rate,
data_format='channels_last',
num_classes=1000,
width_coefficient=width_coefficient,
depth_coefficient=depth_coefficient,
depth_divisor=8,
min_depth=None)
decoder = BlockDecoder()
return decoder.decode(blocks_args), global_params
def get_model_params(model_name, override_params=None):
"""Get the block args and global params for a given model."""
if model_name.startswith('efficientnet'):
width_coefficient, depth_coefficient, input_shape, dropout_rate = (
efficientnet_params(model_name))
blocks_args, global_params = efficientnet(
width_coefficient, depth_coefficient, dropout_rate)
else:
raise NotImplementedError('model name is not pre-defined: %s' % model_name)
if override_params:
# ValueError will be raised here if override_params has fields not included
# in global_params.
global_params = global_params._replace(**override_params)
#print('global_params= %s', global_params)
#print('blocks_args= %s', blocks_args)
return blocks_args, global_params, input_shape
preprocessing.py
import numpy as np
from skimage.transform import resize
MEAN_RGB = [0.485 * 255, 0.456 * 255, 0.406 * 255]
STDDEV_RGB = [0.229 * 255, 0.224 * 255, 0.225 * 255]
MAP_INTERPOLATION_TO_ORDER = {
'nearest': 0,
'bilinear': 1,
'biquadratic': 2,
'bicubic': 3,
}
def center_crop_and_resize(image, image_size, crop_padding=32, interpolation='bicubic'):
assert image.ndim in {2, 3}
assert interpolation in MAP_INTERPOLATION_TO_ORDER.keys()
h, w = image.shape[:2]
padded_center_crop_size = int((image_size / (image_size + crop_padding)) * min(h, w))
offset_height = ((h - padded_center_crop_size) + 1) // 2
offset_width = ((w - padded_center_crop_size) + 1) // 2
image_crop = image[offset_height:padded_center_crop_size + offset_height,
offset_width:padded_center_crop_size + offset_width]
resized_image = resize(
image_crop,
(image_size, image_size),
order=MAP_INTERPOLATION_TO_ORDER[interpolation],
preserve_range=True,
)
return resized_image
def preprocess_input(x):
assert x.ndim in (3, 4)
assert x.shape[-1] == 3
x = x - np.array(MEAN_RGB)
x = x / np.array(STDDEV_RGB)
return x