ライブラリを用いてよくつかう可視化めも
(一部うまくいかなくなってる(バージョン違ってるせいか?)
学習モデルで出力層をカットし, 最後の全結合層を出力とするモデルの構築
#cutting softmax layer
from keras.models import Model
res = Model(inputs=model.input,
outputs=model.get_layer('dense_1').output)
mel = res.predict(x_train)
PCA
from sklearn.decomposition import PCA
X_reduced = PCA(n_components=2).fit_transform(mel[:1000])
plt.scatter(X_reduced[:,0],X_reduced[:,1],cmap=plt.cm.RdYlGn,c=[np.argmax(i) for i in y_train[:1000]])
plt.colorbar()
T-SNE
from sklearn.manifold import TSNE
X_reduced = TSNE(n_components=2,random_state=0).fit_transform(mel[:1000])
plt.scatter(X_reduced[:,0],X_reduced[:,1],cmap=plt.cm.RdYlGn,c=[np.argmax(i) for i in y_train[:1000]])
plt.colorbar()
Tensorboardを用いた可視化
import os
import numpy as np
from PIL import Image
ROOT_DIR = 'logs'
mnist_sprites_path = os.path.join(ROOT_DIR,'images.jpg')
mnist_metadata_path = os.path.join(ROOT_DIR,'metadata.tsv')
# Save the sprite image
img_array = x_test.reshape(100, 100, 28, 28)
img_array_flat = np.concatenate([np.concatenate([x for x in row], axis=1) for row in img_array])
img = Image.fromarray(np.uint8(255 * (1. - img_array_flat)))
img.save(os.path.join(ROOT_DIR, 'images.jpg'))
import matplotlib.pyplot as plt
%matplotlib inline
plt.figure(figsize=(12,12))
plt.imshow(Image.open('logs/images.jpg'),cmap='gray')
# Save the metadata
with open(mnist_metadata_path,'w') as f:
f.write("Index\tLabel\n")
for index,label in enumerate(np.where(y_test)[1]):
f.write("%d\t%d\n" % (index,label))
%head metadata.tsv
Index Label
0 7
1 2
2 1
3 0
4 4
5 1
6 4
7 9
8 5
images.jpg
from tensorflow.contrib.tensorboard.plugins import projector
# Creating the embeddings
from keras.models import Model
res = Model(inputs=model.input,
outputs=model.get_layer('dense_15').output)
emb = res.predict(x_test)
embedding_var = tf.Variable(emb, name='mnist_embedding')
writer = tf.summary.FileWriter(ROOT_DIR)
# Create the embedding projectorc
config = projector.ProjectorConfig()
embedding = config.embeddings.add()
embedding.tensor_name = embedding_var.name
# Specify where you find the metadata
#embedding.metadata_path = mnist_metadata_path #metadata.tsv
embedding.metadata_path='metadata.tsv'
# Specify where you find the sprite
#embedding.sprite.image_path = mnist_sprites_path #images.jpg
embedding.sprite.image_path='images.jpg'
embedding.sprite.single_image_dim.extend([28,28])
projector.visualize_embeddings(writer, config)
# Saving the data
sess = tf.InteractiveSession()
sess.run(tf.global_variables_initializer())
saver = tf.train.Saver()
saver.save(sess, os.path.join(ROOT_DIR,'model.ckpt'),1)
実行コマンド
tensorboard --logdir=logs
以下↓, コード全文
kerasのCNN-MNISTのサンプルコード
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
import tensorflow as tf
import keras
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten
from keras.layers import Conv2D, MaxPooling2D
from keras import backend as K
batch_size = 128
num_classes = 10
epochs = 12
# input image dimensions
img_rows, img_cols = 28, 28
# the data, shuffled and split between train and test sets
(x_train, y_train), (x_test, y_test) = mnist.load_data()
if K.image_data_format() == 'channels_first':
x_train = x_train.reshape(x_train.shape[0], 1, img_rows, img_cols)
x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)
input_shape = (1, img_rows, img_cols)
else:
x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)
x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)
input_shape = (img_rows, img_cols, 1)
x_train = x_train.astype('float32')
x_test = x_test.astype('float32')
x_train /= 255
x_test /= 255
print('x_train shape:', x_train.shape)
print(x_train.shape[0], 'train samples')
print(x_test.shape[0], 'test samples')
# convert class vectors to binary class matrices
y_train = keras.utils.to_categorical(y_train, num_classes)
y_test = keras.utils.to_categorical(y_test, num_classes)
model = Sequential()
model.add(Conv2D(32, kernel_size=(3, 3),
activation='relu',
input_shape=input_shape))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(num_classes, activation='softmax'))
model.compile(loss=keras.losses.categorical_crossentropy,
optimizer=keras.optimizers.Adadelta(),
metrics=['accuracy'])
model.fit(x_train, y_train,
batch_size=batch_size,
epochs=epochs,
verbose=1,
validation_data=(x_test, y_test))
#for layer in model.layers:
# print(layer.output)
model.summary()
#cutting softmax layer
from keras.models import Model
res = Model(inputs=model.input,
outputs=model.get_layer('dense_1').output)
# vis
mel = res.predict(x_train)
model.summary()
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_19 (Conv2D) (None, 26, 26, 32) 320
_________________________________________________________________
conv2d_20 (Conv2D) (None, 24, 24, 64) 18496
_________________________________________________________________
max_pooling2d_10 (MaxPooling (None, 12, 12, 64) 0
_________________________________________________________________
dropout_18 (Dropout) (None, 12, 12, 64) 0
_________________________________________________________________
flatten_10 (Flatten) (None, 9216) 0
_________________________________________________________________
dense_15 (Dense) (None, 128) 1179776
_________________________________________________________________
dropout_19 (Dropout) (None, 128) 0
_________________________________________________________________
dense_16 (Dense) (None, 10) 1290
=================================================================
Total params: 1,199,882
Trainable params: 1,199,882
Non-trainable params: 0
_________________________________________________________________
%ls logs
checkpoint metadata.tsv model.ckpt-1.index projector_config.pbtxt
images.jpg model.ckpt-1.data-00000-of-00001 model.ckpt-1.meta