Deep Learning with Python 读书笔记 (Part 4)

发布于 September 24, 2019

Deep Learning with Python 第五章：Deep learning for computer vision

主要讲以下几部分内容

  • 理解卷积神经网络（convolutional neural networks）
  • 使用数据增强（data augmentation）来避免过拟
  • 使用预训练的卷积神经网络模型来做特征提取（feature extraction）
  • 优化预训练的模型
  • 可视化

训练一个基于小样本的卷积神经网络

我们来看一个训练卷积神经网络模型的实际例子，来解决 Dogs vs. Cats 问题。样本中包含 25000 张猫或狗的图片（各 12500 张）。我们会从每个类别中取 1000 张作为 training set，500 张作为 validation set，500 张作为 test set。

  • 首先，训练一个简单的卷积神经网络，不采用任何正则化（regularization），作为 baseline。这个结果大概能有 71% 的准确率。
  • 使用 data augmentation 来解决过拟问题。这将提高到 82% 的准确率。
  • 然后，使用使用预训练的卷积神经网络模型来做特征提取。90% ~ 96 的准确率。
  • 最后，进行模型调优。最后能达到 97% 的准确率。

  1. 从 Kaggle 上下载 Dogs vs. Cats 数据。

     import os, shutil

 # Copying images to training, validation, and test directories
 original_dataset_dir = '/Data/kaggle_original_data'

 base_dir = '/Data/cats_and_dogs_small'
 os.mkdir(base_dir)

 train_dir = os.path.join(base_dir, 'train')
 os.mkdir(train_dir)
 validation_dir = os.path.join(base_dir, 'validation')
 os.mkdir(validation_dir)
 test_dir = os.path.join(base_dir, 'test')
 os.mkdir(test_dir)

 train_cats_dir = os.path.join(train_dir, 'cats')
 os.mkdir(train_cats_dir)

 train_dogs_dir = os.path.join(train_dir, 'dogs')
 os.mkdir(train_dogs_dir)

 validation_cats_dir = os.path.join(validation_dir, 'cats')
 os.mkdir(validation_cats_dir)

 validation_dogs_dir = os.path.join(validation_dir, 'dogs')
 os.mkdir(validation_dogs_dir)

 test_cats_dir = os.path.join(test_dir, 'cats')
 os.mkdir(test_cats_dir)

 test_dogs_dir = os.path.join(test_dir, 'dogs')
 os.mkdir(test_dogs_dir)

 fnames = ['cat.{}.jpg'.format(i) for i in range(1000)]
 for fname in fnames:
     src = os.path.join(original_dataset_dir, fname)
     dst = os.path.join(train_cats_dir, fname)
     shutil.copyfile(src, dst)

 fnames = ['cat.{}.jpg'.format(i) for i in range(1000, 1500)]
 for fname in fnames:
     src = os.path.join(original_dataset_dir, fname)
     dst = os.path.join(validation_cats_dir, fname)
     shutil.copyfile(src, dst)

 fnames = ['cat.{}.jpg'.format(i) for i in range(1500, 2000)]
 for fname in fnames:
     src = os.path.join(original_dataset_dir, fname)
     dst = os.path.join(test_cats_dir, fname)
     shutil.copyfile(src, dst)

 fnames = ['dog.{}.jpg'.format(i) for i in range(1000)]
 for fname in fnames:
     src = os.path.join(original_dataset_dir, fname)
     dst = os.path.join(train_dogs_dir, fname)
     shutil.copyfile(src, dst)
 fnames = ['dog.{}.jpg'.format(i) for i in range(1000, 1500)]
 for fname in fnames:
     src = os.path.join(original_dataset_dir, fname)
     dst = os.path.join(validation_dogs_dir, fname)
     shutil.copyfile(src, dst)

 fnames = ['dog.{}.jpg'.format(i) for i in range(1500, 2000)]
 for fname in fnames:
     src = os.path.join(original_dataset_dir, fname)
     dst = os.path.join(test_dogs_dir, fname)
     shutil.copyfile(src, dst)

 # a sanity test
 print('total training cat images:', len(os.listdir(train_cats_dir)))
 print('total training dog images:', len(os.listdir(train_dogs_dir)))
 print('total validation cat images:', len(os.listdir(validation_cats_dir)))
 print('total validation dog images:', len(os.listdir(validation_dogs_dir)))
 print('total test cat images:', len(os.listdir(test_cats_dir)))
 print('total test dog images:', len(os.listdir(test_dogs_dir)))

  2. 创建一个卷积模型

    from keras import layers
from keras import models
from keras import optimizers
from keras.preprocessing.image import ImageDataGenerator

# Instantiating a small convnet for dogs vs. cats classification
model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(150, 150, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))
model.summary()
model.compile(loss='binary_crossentropy',
          optimizer=optimizers.RMSprop(lr=1e-4),
          metrics=['acc'])

# data processing
train_datagen = ImageDataGenerator(rescale=1./255)
test_datagen = ImageDataGenerator(rescale=1./255)
train_generator = train_datagen.flow_from_directory(
    train_dir,
    target_size=(150, 150)
    batch_size=20,
    class_mode='binary')
validation_generator = test_datagen.flow_from_directory(
    validation_dir,
    target_size=(150, 150),
    batch_size=20,
    class_mode='binary')

# Fitting the model using a batch generator
history = model.fit_generator(
  train_generator,
  steps_per_epoch=100,
  epochs=30,
  validation_data=validation_generator,
  validation_steps=50)
model.save('cats_and_dogs_small_1.h5')

  3. 使用 data augmentation 来解决过拟

    model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu',
                        input_shape=(150, 150, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dropout(0.5))
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))

model.compile(loss='binary_crossentropy',
            optimizer=optimizers.RMSprop(lr=1e-4),
            metrics=['acc'])

# Training the convnet using data-augmentation generators
train_datagen = ImageDataGenerator(
    rescale=1./255,
    rotation_range=40,
    width_shift_range=0.2,
    height_shift_range=0.2,
    shear_range=0.2,
    zoom_range=0.2,
    horizontal_flip=True,)

test_datagen = ImageDataGenerator(rescale=1./255)
train_generator = train_datagen.flow_from_directory(
    train_dir,
    target_size=(150, 150),
    batch_size=32,
    class_mode='binary')
validation_generator = test_datagen.flow_from_directory(
    validation_dir,
    target_size=(150, 150),
    batch_size=32,
    class_mode='binary')

history = model.fit_generator(
    train_generator,
    steps_per_epoch=100,
    epochs=100,
    validation_data=validation_generator,
    validation_steps=50)
model.save('cats_and_dogs_small_2.h5')

  4. 两种方法来使用预训练模型：feature extraction 和 fine-tuning。代码略。

卷积神经网络的可视化

几种可视化的方法：

  • Visualizing intermediate convnet outputs (intermediate activations)

    intermediate_activations

  • Visualizing convnets filters

    filter_patterns

  • Visualizing heatmaps of class activation in an image

    heatmaps

总结

卷积神经网络是用于解决视觉分类问题的最好工具。

E