Stanford CS231n實踐筆記（課時22卷積神經網絡工程實踐技巧與註意點 cnn in practise 上）
阿新 • • 發佈：2018-04-16
self ash 2gb ble 一個 time 縮放 對比度 cati 本課主要2個實踐內容：1、keras中數據集豐富，從數據集中提取更多特征（Data augmentation）2、遷移學習（Tranform learning）
代碼：https://github.com/jsxyhelu/DateSets1、keras中數據集豐富，從數據集中提取更多特征（Data augmentation）keras是比較現代化的DL工具，所以這方面的功能都是具備的。這裏首先將相關知識進行整理，然後將例子進行實現，特別註重結果的展示。具體內容包括：
參數非常多，用以生成一個batch的圖像數據，支持實時數據提升。訓練時該函數會無限生成數據，直到達到規定的epoch次數為止；那麽在訓練的時候肯定是需要結合fit_generate來使用，正好是最近研究的。主要方法：a、flow(self, X, y, batch_size=32, shuffle=True, seed=None, save_to_dir=None, save_prefix=‘‘, save_format=‘png‘)將會返回一個生成器，這個生成器用來擴充數據，每次都會產生batch_size個樣本。因為目前我們只導入了一張圖片，因此每次生成的圖片都是基於這張圖片而產生的，可以看到結果，旋轉、位移、放大縮小，統統都有。正如其名稱，這個生成器可以一直生成下去。 
from keras.preprocessing.image import ImageDataGeneratorfrom keras.preprocessing import imageimport matplotlib.pyplot as pltimport numpy as np
# 指定參數# rotation_range 旋轉# width_shift_range 左右平移# height_shift_range 上下平移# zoom_range 隨機放大或縮小img_generator = ImageDataGenerator( rotation_range = 90, width_shift_range = 0.2, height_shift_range = 0.2, zoom_range = 0.3 )# 導入並顯示圖片img_path = ‘e:/template/lena.jpg‘img = image.load_img(img_path)plt.imshow(img)plt.show()
# 將圖片轉為數組x = image.img_to_array(img)# 擴充一個維度x = np.expand_dims(x, axis=0)# 生成圖片gen = img_generator.flow(x, batch_size=1)
# 顯示生成的圖片plt.figure()for i in range(3): for j in range(3): x_batch = next(gen) idx = (3*i) + j plt.subplot(3, 3, idx+1) plt.imshow(x_batch[0]/256)x_batch.shapeplt.show()
代碼中的絕對核心一句gen = img_generator.flow(x, batch_size=1)這樣,gen相當於是img_generatror的一個生成器，下面直接使用next(gen)就可以不斷生成。e.g2驗證DateAugment性能。都說這個DateAugment好，但是不之際動手實踐一下，總是不放心。from keras.datasets import cifar10from keras.layers.core import Dense, Flatten, Activation, Dropoutfrom keras.layers.convolutional import Conv2Dfrom keras.layers.pooling import MaxPooling2Dfrom keras.layers.normalization import BatchNormalizationfrom keras.models import Sequentialfrom keras.utils import np_utilsfrom keras.preprocessing.image import ImageDataGeneratorfrom keras.preprocessing import imageimport matplotlib.pyplot as pltimport numpy as npfrom keras.utils import generic_utils
(x_train, y_train),(x_test, y_test) = cifar10.load_data()print(x_train.shape, y_train.shape, x_test.shape, y_test.shape)
def preprocess_data(x): x /= 255 x -= 0.5 x *= 2 return x
# 預處理x_train = x_train.astype(np.float32)x_test = x_test.astype(np.float32)
x_train = preprocess_data(x_train)x_test = preprocess_data(x_test)
# one-hot encodingn_classes = 10y_train = np_utils.to_categorical(y_train, n_classes)y_test = np_utils.to_categorical(y_test, n_classes)
# 取 20% 的訓練數據x_train_part = x_train[:10000]y_train_part = y_train[:10000]
print(x_train_part.shape, y_train_part.shape)
# 建立一個簡單的卷積神經網絡,序貫結構def build_model(): model = Sequential()
model.add(Conv2D(64, (3,3), input_shape=(32,32,3))) model.add(Activation(‘relu‘)) model.add(BatchNormalization(scale=False, center=False))
model.add(Conv2D(32, (3,3))) model.add(Activation(‘relu‘)) model.add(MaxPooling2D((2,2))) model.add(Dropout(0.2)) model.add(BatchNormalization(scale=False, center=False))
model.add(Flatten()) model.add(Dense(256)) model.add(Activation(‘relu‘)) model.add(Dropout(0.2)) model.add(BatchNormalization())
model.add(Dense(n_classes)) model.add(Activation(‘softmax‘))
return model
# 訓練參數batch_size = 128#epochs = 20epochs = 2#cifar-10 20%數據,訓練結果,繪圖model = build_model()model.compile(optimizer=‘adam‘, loss=‘categorical_crossentropy‘, metrics=[‘accuracy‘])model.fit(x_train_part, y_train_part, epochs=epochs, batch_size=batch_size, verbose=1, validation_split=0.1)
loss, acc = model.evaluate(x_test, y_test, batch_size=32)print(‘Loss: ‘, loss)print(‘Accuracy: ‘, acc)
#cifar-10 20%數據 + Data Augmentation.訓練結果# 設置生成參數img_generator = ImageDataGenerator( rotation_range = 20, width_shift_range = 0.2, height_shift_range = 0.2, zoom_range = 0.2 )model_2 = build_model()model_2.compile(optimizer=‘adam‘, loss=‘categorical_crossentropy‘, metrics=[‘accuracy‘])
#采用文檔中提示自動方法img_generator.fit(x_train_part)
# fits the model_2 on batches with real-time data augmentation:model_2.fit_generator(img_generator.flow(x_train_part, y_train_part, batch_size=batch_size), steps_per_epoch=len(x_train_part), epochs=epochs)初步比較結果：原始Loss: 1.6311466661453247Accuracy: 0.6169調整Epoch 1/20 2184/10000 [=====>........................] - ETA: 11:11 - loss: 1.0548 - acc: 0.6254雖然沒有走完，但是已經初見端倪。
2、遷移學習（Tranform learning）基於前面已經實現的代碼，在fasion_mnist上實現vgg的模型遷移。import numpy as npfrom keras.datasets import fashion_mnistimport gc
from keras.models import Sequential, Modelfrom keras.layers import Input, Dense, Dropout, Flattenfrom keras.layers.convolutional import Conv2D, MaxPooling2Dfrom keras.applications.vgg16 import VGG16from keras.optimizers import SGDimport matplotlib.pyplot as pltimport os
import cv2import h5py as h5py import numpy as npdef tran_y(y): y_ohe = np.zeros(10) y_ohe[y] = 1 return y_oheepochs = 1
# 如果硬件配置較高，比如主機具備32GB以上內存，GPU具備8GB以上顯存，可以適當增大這個值。VGG要求至少48像素ishape=48(X_train, y_train), (X_test, y_test) = fashion_mnist.load_data()
X_train = [cv2.cvtColor(cv2.resize(i, (ishape, ishape)), cv2.COLOR_GRAY2BGR) for i in X_train] X_train = np.concatenate([arr[np.newaxis] for arr in X_train]).astype(‘float32‘) X_train /= 255.0
X_test = [cv2.cvtColor(cv2.resize(i, (ishape, ishape)), cv2.COLOR_GRAY2BGR) for i in X_test] X_test = np.concatenate([arr[np.newaxis] for arr in X_test]).astype(‘float32‘)X_test /= 255.0
y_train_ohe = np.array([tran_y(y_train[i]) for i in range(len(y_train))]) y_test_ohe = np.array([tran_y(y_test[i]) for i in range(len(y_test))])y_train_ohe = y_train_ohe.astype(‘float32‘)y_test_ohe = y_test_ohe.astype(‘float32‘)
model_vgg = VGG16(include_top = False, weights = ‘imagenet‘, input_shape = (ishape, ishape, 3)) for layer in model_vgg.layers: layer.trainable = Falsemodel = Flatten()(model_vgg.output) model = Dense(4096, activation=‘relu‘, name=‘fc1‘)(model)model = Dense(4096, activation=‘relu‘, name=‘fc2‘)(model)model = Dropout(0.5)(model)model = Dense(10, activation = ‘softmax‘, name=‘prediction‘)(model) model_vgg_mnist_pretrain = Model(model_vgg.input, model, name = ‘vgg16_pretrain‘)model_vgg_mnist_pretrain.summary()sgd = SGD(lr = 0.05, decay = 1e-5) model_vgg_mnist_pretrain.compile(loss = ‘categorical_crossentropy‘, optimizer = sgd, metrics = [‘accuracy‘])log = model_vgg_mnist_pretrain.fit(X_train, y_train_ohe, validation_data = (X_test, y_test_ohe), epochs = epochs, batch_size = 64)
score = model_vgg_mnist_pretrain.evaluate(X_test, y_test_ohe, verbose=0)print(‘Test loss:‘, score[0])print(‘Test accuracy:‘, score[1])
plt.figure(‘acc‘) plt.subplot(2, 1, 1) plt.plot(log.history[‘acc‘],‘r--‘,label=‘Training Accuracy‘) plt.plot(log.history[‘val_acc‘],‘r-‘,label=‘Validation Accuracy‘) plt.legend(loc=‘best‘) plt.xlabel(‘Epochs‘) plt.axis([0, epochs, 0.9, 1]) plt.figure(‘loss‘) plt.subplot(2, 1, 2) plt.plot(log.history[‘loss‘],‘b--‘,label=‘Training Loss‘) plt.plot(log.history[‘val_loss‘],‘b-‘,label=‘Validation Loss‘) plt.legend(loc=‘best‘) plt.xlabel(‘Epochs‘) plt.axis([0, epochs, 0, 1]) plt.show() os.system("pause")其中絕對核心為：model = Flatten()(model_vgg.output) model = Dense(4096, activation=‘relu‘, name=‘fc1‘)(model)model = Dense(4096, activation=‘relu‘, name=‘fc2‘)(model)model = Dropout(0.5)(model)model = Dense(10, activation = ‘softmax‘, name=‘prediction‘)(model) 遷移學習非常重要（也只有這裏需要編碼）就是你遷移的那幾層，如何進行修改。可能是要和你實際操作的數據的模式是相關的。
小結：本課主要2個實踐內容：1、keras中數據集豐富，從數據集中提取更多特征（Data augmentation）2、遷移學習（Tranform learning）分別使用了cifar10和fasion_mnist數據集。這兩個數據集能夠很方便地用以實驗，一方面是因為它們都是已經規整好的數據集，另一方面，兩者都是“單”數據集，也就是一個標簽就是一個圖片類型的情況。這些都可能被進一步復雜。
來自為知筆記(Wiz)
代碼：https://github.com/jsxyhelu/DateSets1、keras中數據集豐富，從數據集中提取更多特征（Data augmentation）keras是比較現代化的DL工具，所以這方面的功能都是具備的。這裏首先將相關知識進行整理，然後將例子進行實現，特別註重結果的展示。具體內容包括：
- 旋轉 | 反射變換(Rotation/reflection): 隨機旋轉圖像一定角度; 改變圖像內容的朝向;
- 翻轉變換(flip): 沿著水平或者垂直方向翻轉圖像;
- 縮放變換(zoom): 按照一定的比例放大或者縮小圖像;
- 平移變換(shift): 在圖像平面上對圖像以一定方式進行平移;
- 可以采用隨機或人為定義的方式指定平移範圍和平移步長, 沿水平或豎直方向進行平移. 改變圖像內容的位置;
- 尺度變換(scale): 對圖像按照指定的尺度因子, 進行放大或縮小; 或者參照SIFT特征提取思想, 利用指定的尺度因子對圖像濾波構造尺度空間. 改變圖像內容的大小或模糊程度;
- 對比度變換(contrast): 在圖像的HSV顏色空間，改變飽和度S和V亮度分量，保持色調H不變. 對每個像素的S和V分量進行指數運算(指數因子在0.25到4之間), 增加光照變化;
- 噪聲擾動(noise): 對圖像的每個像素RGB進行隨機擾動, 常用的噪聲模式是椒鹽噪聲和高斯噪聲;
使用的函數為：
參數非常多，用以生成一個batch的圖像數據，支持實時數據提升。訓練時該函數會無限生成數據，直到達到規定的epoch次數為止；那麽在訓練的時候肯定是需要結合fit_generate來使用，正好是最近研究的。主要方法：a、flow(self, X, y, batch_size=32, shuffle=True, seed=None, save_to_dir=None, save_prefix=‘‘, save_format=‘png‘)將會返回一個生成器，這個生成器用來擴充數據，每次都會產生batch_size個樣本。因為目前我們只導入了一張圖片，因此每次生成的圖片都是基於這張圖片而產生的，可以看到結果，旋轉、位移、放大縮小，統統都有。正如其名稱，這個生成器可以一直生成下去。from keras.preprocessing.image import ImageDataGeneratorfrom keras.preprocessing import imageimport matplotlib.pyplot as pltimport numpy as np# 指定參數# rotation_range 旋轉# width_shift_range 左右平移# height_shift_range 上下平移# zoom_range 隨機放大或縮小img_generator = ImageDataGenerator( rotation_range = 90, width_shift_range = 0.2, height_shift_range = 0.2, zoom_range = 0.3 )# 導入並顯示圖片img_path = ‘e:/template/lena.jpg‘img = image.load_img(img_path)plt.imshow(img)plt.show()
# 將圖片轉為數組x = image.img_to_array(img)# 擴充一個維度x = np.expand_dims(x, axis=0)# 生成圖片gen = img_generator.flow(x, batch_size=1)
# 顯示生成的圖片plt.figure()for i in range(3): for j in range(3): x_batch = next(gen) idx = (3*i) + j plt.subplot(3, 3, idx+1) plt.imshow(x_batch[0]/256)x_batch.shapeplt.show()
代碼中的絕對核心一句gen = img_generator.flow(x, batch_size=1)這樣,gen相當於是img_generatror的一個生成器，下面直接使用next(gen)就可以不斷生成。e.g2驗證DateAugment性能。都說這個DateAugment好，但是不之際動手實踐一下，總是不放心。from keras.datasets import cifar10from keras.layers.core import Dense, Flatten, Activation, Dropoutfrom keras.layers.convolutional import Conv2Dfrom keras.layers.pooling import MaxPooling2Dfrom keras.layers.normalization import BatchNormalizationfrom keras.models import Sequentialfrom keras.utils import np_utilsfrom keras.preprocessing.image import ImageDataGeneratorfrom keras.preprocessing import imageimport matplotlib.pyplot as pltimport numpy as npfrom keras.utils import generic_utils
(x_train, y_train),(x_test, y_test) = cifar10.load_data()print(x_train.shape, y_train.shape, x_test.shape, y_test.shape)
def preprocess_data(x): x /= 255 x -= 0.5 x *= 2 return x
# 預處理x_train = x_train.astype(np.float32)x_test = x_test.astype(np.float32)
x_train = preprocess_data(x_train)x_test = preprocess_data(x_test)
# one-hot encodingn_classes = 10y_train = np_utils.to_categorical(y_train, n_classes)y_test = np_utils.to_categorical(y_test, n_classes)
# 取 20% 的訓練數據x_train_part = x_train[:10000]y_train_part = y_train[:10000]
print(x_train_part.shape, y_train_part.shape)
# 建立一個簡單的卷積神經網絡,序貫結構def build_model(): model = Sequential()
model.add(Conv2D(64, (3,3), input_shape=(32,32,3))) model.add(Activation(‘relu‘)) model.add(BatchNormalization(scale=False, center=False))
model.add(Conv2D(32, (3,3))) model.add(Activation(‘relu‘)) model.add(MaxPooling2D((2,2))) model.add(Dropout(0.2)) model.add(BatchNormalization(scale=False, center=False))
model.add(Flatten()) model.add(Dense(256)) model.add(Activation(‘relu‘)) model.add(Dropout(0.2)) model.add(BatchNormalization())
model.add(Dense(n_classes)) model.add(Activation(‘softmax‘))
return model
# 訓練參數batch_size = 128#epochs = 20epochs = 2#cifar-10 20%數據,訓練結果,繪圖model = build_model()model.compile(optimizer=‘adam‘, loss=‘categorical_crossentropy‘, metrics=[‘accuracy‘])model.fit(x_train_part, y_train_part, epochs=epochs, batch_size=batch_size, verbose=1, validation_split=0.1)
loss, acc = model.evaluate(x_test, y_test, batch_size=32)print(‘Loss: ‘, loss)print(‘Accuracy: ‘, acc)
#cifar-10 20%數據 + Data Augmentation.訓練結果# 設置生成參數img_generator = ImageDataGenerator( rotation_range = 20, width_shift_range = 0.2, height_shift_range = 0.2, zoom_range = 0.2 )model_2 = build_model()model_2.compile(optimizer=‘adam‘, loss=‘categorical_crossentropy‘, metrics=[‘accuracy‘])
#采用文檔中提示自動方法img_generator.fit(x_train_part)
# fits the model_2 on batches with real-time data augmentation:model_2.fit_generator(img_generator.flow(x_train_part, y_train_part, batch_size=batch_size), steps_per_epoch=len(x_train_part), epochs=epochs)初步比較結果：原始Loss: 1.6311466661453247Accuracy: 0.6169調整Epoch 1/20 2184/10000 [=====>........................] - ETA: 11:11 - loss: 1.0548 - acc: 0.6254雖然沒有走完，但是已經初見端倪。
2、遷移學習（Tranform learning）基於前面已經實現的代碼，在fasion_mnist上實現vgg的模型遷移。import numpy as npfrom keras.datasets import fashion_mnistimport gc
from keras.models import Sequential, Modelfrom keras.layers import Input, Dense, Dropout, Flattenfrom keras.layers.convolutional import Conv2D, MaxPooling2Dfrom keras.applications.vgg16 import VGG16from keras.optimizers import SGDimport matplotlib.pyplot as pltimport os
import cv2import h5py as h5py import numpy as npdef tran_y(y): y_ohe = np.zeros(10) y_ohe[y] = 1 return y_oheepochs = 1
# 如果硬件配置較高，比如主機具備32GB以上內存，GPU具備8GB以上顯存，可以適當增大這個值。VGG要求至少48像素ishape=48(X_train, y_train), (X_test, y_test) = fashion_mnist.load_data()
X_train = [cv2.cvtColor(cv2.resize(i, (ishape, ishape)), cv2.COLOR_GRAY2BGR) for i in X_train] X_train = np.concatenate([arr[np.newaxis] for arr in X_train]).astype(‘float32‘) X_train /= 255.0
X_test = [cv2.cvtColor(cv2.resize(i, (ishape, ishape)), cv2.COLOR_GRAY2BGR) for i in X_test] X_test = np.concatenate([arr[np.newaxis] for arr in X_test]).astype(‘float32‘)X_test /= 255.0
y_train_ohe = np.array([tran_y(y_train[i]) for i in range(len(y_train))]) y_test_ohe = np.array([tran_y(y_test[i]) for i in range(len(y_test))])y_train_ohe = y_train_ohe.astype(‘float32‘)y_test_ohe = y_test_ohe.astype(‘float32‘)
model_vgg = VGG16(include_top = False, weights = ‘imagenet‘, input_shape = (ishape, ishape, 3)) for layer in model_vgg.layers: layer.trainable = Falsemodel = Flatten()(model_vgg.output) model = Dense(4096, activation=‘relu‘, name=‘fc1‘)(model)model = Dense(4096, activation=‘relu‘, name=‘fc2‘)(model)model = Dropout(0.5)(model)model = Dense(10, activation = ‘softmax‘, name=‘prediction‘)(model) model_vgg_mnist_pretrain = Model(model_vgg.input, model, name = ‘vgg16_pretrain‘)model_vgg_mnist_pretrain.summary()sgd = SGD(lr = 0.05, decay = 1e-5) model_vgg_mnist_pretrain.compile(loss = ‘categorical_crossentropy‘, optimizer = sgd, metrics = [‘accuracy‘])log = model_vgg_mnist_pretrain.fit(X_train, y_train_ohe, validation_data = (X_test, y_test_ohe), epochs = epochs, batch_size = 64)
score = model_vgg_mnist_pretrain.evaluate(X_test, y_test_ohe, verbose=0)print(‘Test loss:‘, score[0])print(‘Test accuracy:‘, score[1])
plt.figure(‘acc‘) plt.subplot(2, 1, 1) plt.plot(log.history[‘acc‘],‘r--‘,label=‘Training Accuracy‘) plt.plot(log.history[‘val_acc‘],‘r-‘,label=‘Validation Accuracy‘) plt.legend(loc=‘best‘) plt.xlabel(‘Epochs‘) plt.axis([0, epochs, 0.9, 1]) plt.figure(‘loss‘) plt.subplot(2, 1, 2) plt.plot(log.history[‘loss‘],‘b--‘,label=‘Training Loss‘) plt.plot(log.history[‘val_loss‘],‘b-‘,label=‘Validation Loss‘) plt.legend(loc=‘best‘) plt.xlabel(‘Epochs‘) plt.axis([0, epochs, 0, 1]) plt.show() os.system("pause")其中絕對核心為：model = Flatten()(model_vgg.output) model = Dense(4096, activation=‘relu‘, name=‘fc1‘)(model)model = Dense(4096, activation=‘relu‘, name=‘fc2‘)(model)model = Dropout(0.5)(model)model = Dense(10, activation = ‘softmax‘, name=‘prediction‘)(model) 遷移學習非常重要（也只有這裏需要編碼）就是你遷移的那幾層，如何進行修改。可能是要和你實際操作的數據的模式是相關的。
小結：本課主要2個實踐內容：1、keras中數據集豐富，從數據集中提取更多特征（Data augmentation）2、遷移學習（Tranform learning）分別使用了cifar10和fasion_mnist數據集。這兩個數據集能夠很方便地用以實驗，一方面是因為它們都是已經規整好的數據集，另一方面，兩者都是“單”數據集，也就是一個標簽就是一個圖片類型的情況。這些都可能被進一步復雜。
來自為知筆記(Wiz)
Stanford CS231n實踐筆記（課時22卷積神經網絡工程實踐技巧與註意點 cnn in practise 上）