我正在尝试学习keras，而我所使用的代码都不是学习的。从使用Python进行深度学习的示例代码到https://medium.com/nybles/create-your-first-image-recognition-classifier-using-cnn-keras-and-tensorflow-backend-6eaab98d14dd上的代码。在最后一个链接中，我无法使用完整的10000个数据集，但是即使使用我的1589大小的训练数据集，精度始终始终保持在0.5。
我几乎开始认为问题出在我的超频CPU和ram上，但这更多是一个疯狂的猜测。
我最初以为问题是我有tensorflow2.0.0-alpha。但是，即使我去了常规的tensorflow-gpu，仍然没有任何学习的经验。
#Convolutional Neural Network
# Importing the Keras libraries and packages
from keras.models import Sequential
from keras.layers import Convolution2D
from keras.layers import MaxPooling2D
from keras.layers import Flatten
from keras.layers import Dense
from keras.models import model_from_json
import os
#initialize the cnn
classifier = Sequential()
#Step 1 convolution
classifier.add(Convolution2D(32, 3, 3, input_shape = (64, 64, 3), activation = 'relu'))
#Step 2 Pooling
classifier.add(MaxPooling2D(pool_size = (2,2)))
#Step 3 Flattening
classifier.add(Flatten())
#Step 4 Full Connection
classifier.add(Dense(output_dim = 128, activation = 'relu'))
classifier.add(Dense(output_dim = 64, activation = 'relu'))
classifier.add(Dense(output_dim = 32, activation = 'relu'))
classifier.add(Dense(output_dim = 1, activation = 'sigmoid'))
#Compiling the CNN
classifier.compile(optimizer = 'adam', loss = 'binary_crossentropy', metrics = ['accuracy'])
#Part 2 Fitting the CNN to the images
from keras.preprocessing.image import ImageDataGenerator
train_datagen = ImageDataGenerator(
rescale=1/.255,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True)
test_datagen = ImageDataGenerator(rescale=1./255)
training_set = train_datagen.flow_from_directory(
'dataset/training_set',
target_size=(64, 64),
batch_size=32,
class_mode='binary')
test_set = test_datagen.flow_from_directory(
'dataset/test_set',
target_size=(64, 64),
batch_size=32,
class_mode='binary')
from IPython.display import display
from PIL import Image
classifier.fit_generator(
training_set,
steps_per_epoch=1589,
epochs=10,
validation_data=test_set,
validation_steps=378)
import numpy as np
from keras.preprocessing import image
test_image = image.load_img('dataset/test_set/cats/cat.4012.jpg', target_size = (64,64))
test_image = image.img_to_array(test_image)
test_image = np.expand_dims(test_image, axis = 0)
result = classifier.predict(test_image)
training_set.class_indices
if result[0][0] >= 0.5:
prediction = 'dog'
else:
prediction = 'cat'
print(prediction)
使用python进行深度学习的示例
from keras.datasets import imdb
(train_data, train_labels), (test_data, test_labels) =
imdb.load_data(num_words = 10000)
import numpy as np
def vectorize_sequences(sequences, dimension=10000):
results = np.zeros((len(sequences), dimension))
for i,sequence in enumerate(sequences):
results[i, sequence]=1.
return results
x_train = vectorize_sequences(train_data)
x_test = vectorize_sequences(test_data)
y_train = np.asarray(train_labels).astype('float32')
y_test = np.asarray(test_labels).astype('float32')
from keras import models
from keras import layers
model = models.Sequential()
model.add(layers.Dense(16, activation='relu',input_shape=(10000,)))
model.add(layers.Dense(16, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))
x_val = x_train[:10000]
partial_x_train = x_train[10000:]
y_val = y_train[:10000]
partial_y_train = y_train[10000:]
model.compile(optimizer='rmsprop',
loss='binary_crossentropy',
metrics=['acc'])
history = model.fit(partial_x_train,
partial_y_train,
epochs=20,
batch_size=512,
validation_data=(x_val,y_val))
狗猫输出：
Epoch 1/10
1589/1589 [==============================] - 112s 70ms/step - loss: 7.8736 - acc: 0.5115 - val_loss: 7.9528 - val_acc: 0.4976
Epoch 2/10
1589/1589 [==============================] - 111s 70ms/step - loss: 7.8697 - acc: 0.5117 - val_loss: 7.9606 - val_acc: 0.4971
Epoch 3/10
1589/1589 [==============================] - 111s 70ms/step - loss: 7.8740 - acc: 0.5115 - val_loss: 7.9499 - val_acc: 0.4978
Epoch 4/10
1589/1589 [==============================] - 111s 70ms/step - loss: 7.8674 - acc: 0.5119 - val_loss: 7.9634 - val_acc: 0.4969
Epoch 5/10
1589/1589 [==============================] - 111s 70ms/step - loss: 7.8765 - acc: 0.5113 - val_loss: 7.9499 - val_acc: 0.4977
Epoch 6/10
1589/1589 [==============================] - 111s 70ms/step - loss: 7.8737 - acc: 0.5115 - val_loss: 7.9634 - val_acc: 0.4970
Epoch 7/10
1589/1589 [==============================] - 129s 81ms/step - loss: 7.8623 - acc: 0.5122 - val_loss: 7.9626 - val_acc: 0.4970
Epoch 8/10
1589/1589 [==============================] - 112s 71ms/step - loss: 7.8758 - acc: 0.5114 - val_loss: 7.9508 - val_acc: 0.4977
Epoch 9/10
1589/1589 [==============================] - 115s 72ms/step - loss: 7.8708 - acc: 0.5117 - val_loss: 7.9519 - val_acc: 0.4976
Epoch 10/10
1589/1589 [==============================] - 112s 70ms/step - loss: 7.8738 - acc: 0.5115 - val_loss: 7.9614 - val_acc: 0.4971
cat
deeplearning imdb示例输出： 警告：tensorflow：来自C：\ Users \ Mike \ Anaconda3 \ lib \ site-packages \ tensorflow \ python \ ops \ math_ops.py：3066：不推荐使用to_int32（来自tensorflow.python.ops.math_ops）并将其删除将来的版本。
更新说明： 请改用tf.cast。 训练15000个样本，验证10000个样本
Epoch 1/20
15000/15000 [==============================] - 4s 246us/step - loss: 0.6932 - acc: 0.4982 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 2/20
15000/15000 [==============================] - 2s 115us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 3/20
15000/15000 [==============================] - 2s 115us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 4/20
15000/15000 [==============================] - 2s 119us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 5/20
15000/15000 [==============================] - 2s 120us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 6/20
15000/15000 [==============================] - 2s 119us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6933 - val_acc: 0.4947
Epoch 7/20
15000/15000 [==============================] - 2s 113us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 8/20
15000/15000 [==============================] - 2s 113us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 9/20
15000/15000 [==============================] - 2s 119us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6933 - val_acc: 0.4947
Epoch 10/20
15000/15000 [==============================] - 2s 122us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6933 - val_acc: 0.4947
Epoch 11/20
15000/15000 [==============================] - 2s 116us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6933 - val_acc: 0.4947
Epoch 12/20
15000/15000 [==============================] - 2s 116us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6933 - val_acc: 0.4947
Epoch 13/20
15000/15000 [==============================] - 2s 121us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6933 - val_acc: 0.4947
Epoch 14/20
15000/15000 [==============================] - 2s 127us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 15/20
15000/15000 [==============================] - 2s 121us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 16/20
15000/15000 [==============================] - 2s 113us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 17/20
15000/15000 [==============================] - 2s 115us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 18/20
15000/15000 [==============================] - 2s 114us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 19/20
15000/15000 [==============================] - 2s 114us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947
Epoch 20/20
15000/15000 [==============================] - 2s 119us/step - loss: 0.6931 - acc: 0.5035 - val_loss: 0.6932 - val_acc: 0.4947