from tensorflow.keras.datasets import mnist
from myLibrary.util import one_hot
import numpy as np
np.set_printoptions(edgeitems=256)
(train_images, train_labels), (test_images, test_labels) = mnist.load_data()
train_images = train_images.reshape(-1, 28, 28, 1)
train_images = train_images / 255
train_labels = one_hot(train_labels, 10)
from tensorflow.keras import layers, models
model = models.Sequential()
model.add(
layers.Conv2D(64,
kernel_size=3,
activation='relu',
input_shape=(28, 28, 1)))
model.add(layers.MaxPool2D(pool_size=2))
model.add(layers.Flatten())
model.add(layers.Dense(10, activation='softmax'))
model.compile(optimizer='rmsprop',
loss='categorical_crossentropy',
metrics=['acc'])
model.fit(x=train_images,
y=train_labels,
batch_size=1000,
epochs=30,
validation_split=0.2)
for files in os.listdir(db):
if files != 'label' and os.path.isdir(db + files + '/'):
for img in os.listdir(db + files + '/'):
X.append(db + files + '/' + img)
Y.append(int(files))
return X, Y
trans = Transformer()
trans.add(lambda img: cv2.resize(img, (224, 224)))
trans.add(lambda img: cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
DB = 'animal_train_data/'
imgs, labels = read_data(DB)
labels = one_hot(labels, 4)
dataset = ImageDataset(imgs, labels, transformer=trans)
from tensorflow.keras import metrics, optimizers
net = VGG16(weights='imagenet', include_top=True)
'''
net = models.Sequential()
net.add(layers.Flatten())
net.add(layers.Dense(4096, activation='relu'))
net.add(layers.Dense(4096, activation='relu'))
net.add(layers.Dense(3, activation='softmax'))
net = Model(input = pretrained.input, output = net)'''
from tensorflow.keras.datasets import mnist
from tensorflow.keras.models import load_model
from myLibrary.util import one_hot
(train_images, train_labels), (test_images, test_labels) = mnist.load_data()
test_images = test_images.reshape(-1, 28, 28, 1)
test_images = test_images / 255
test_labels = one_hot(test_labels, 10)
net = load_model('mnist_cnn.h5')
predict = net.predict(test_images)
print(predict.argmax(axis=1))
print(test_labels.argmax(axis=1))
count = (predict.argmax(axis=1) == test_labels.argmax(axis=1)).sum()
print(count / len(test_images))
'''import matplotlib.pyplot as plt
img = train_images[0]
plt.imshow(img)
plt.show()
'''
tags = ['飛機', '汽車', '鳥', '貓', '鹿', '狗', '青蛙', '馬', '船', '卡車']
print(train_images.shape)
train_images = train_images / 255
test_images = test_images / 255
print(train_images.shape)
train_labels = one_hot(train_labels, 10)
test_labels = one_hot(test_labels, 10)
print(type(test_images))
from myLibrary.model import to_cifar10
net = to_cifar10(10)
net.compile(optimizer='rmsprop',
loss='categorical_crossentropy',
metrics=['acc'])
net.fit(x=train_images,
y=train_labels,
batch_size=1000,
epochs=20,
validation_split=0.2)