配置环境
import matplotlib.pyplot as plt
import numpy as np
import warnings
warnings.filterwarnings('ignore')
from tensorflow.keras import layers
import tensorflow as tf
gpus = tf.config.list_physical_devices("GPU")
if gpus:
tf.config.experimental.set_memory_growth(gpus[0], True)
tf.config.set_visible_devices([gpus[0]],"GPU")
print(gpus)
[]
加载数据
data_dir = r"C:\Users\11054\Desktop\kLearning\t9_learning\data"
img_height = 224
img_width = 224
batch_size = 32
train_ds = tf.keras.preprocessing.image_dataset_from_directory(
data_dir,
validation_split=0.3,
subset="training",
seed=12,
image_size=(img_height, img_width),
batch_size=batch_size)
Found 3400 files belonging to 2 classes.
Using 2380 files for training.
- preprocessing.image_dataset_from_directory作用
- 将文件夹中的数据加载到tf.data.Dataset中,且加载的同时会打乱数据
val_ds = tf.keras.preprocessing.image_dataset_from_directory(
data_dir,
validation_split=0.3,
subset="training",
seed=12,
image_size=(img_height, img_width),
batch_size=batch_size)
Found 3400 files belonging to 2 classes.
Using 2380 files for training.
val_batches = tf.data.experimental.cardinality(val_ds)
test_ds = val_ds.take(val_batches // 5)
val_ds = val_ds.skip(val_batches // 5)
print('Number of validation batches: %d' % tf.data.experimental.cardinality(val_ds))
print('Number of test batches: %d' % tf.data.experimental.cardinality(test_ds))
Number of validation batches: 60
Number of test batches: 15
显示数据类
class_names = train_ds.class_names
print(class_names)
['cat', 'dog']
AUTOTUNE = tf.data.AUTOTUNE
def preprocess_image(image,label):
return (image/255.0,label)
train_ds = train_ds.map(preprocess_image, num_parallel_calls=AUTOTUNE)
val_ds = val_ds.map(preprocess_image, num_parallel_calls=AUTOTUNE)
test_ds = test_ds.map(preprocess_image, num_parallel_calls=AUTOTUNE)
train_ds = train_ds.cache().prefetch(buffer_size=AUTOTUNE)
val_ds = val_ds.cache().prefetch(buffer_size=AUTOTUNE)
显示数据
plt.figure(figsize=(15, 10))
for images, labels in train_ds.take(1):
for i in range(8):
ax = plt.subplot(5, 8, i + 1)
plt.imshow(images[i])
plt.title(class_names[labels[i]])
plt.axis("off")
data_augmentation = tf.keras.Sequential([
tf.keras.layers.RandomFlip("horizontal_and_vertical"),
tf.keras.layers.RandomRotation(0.2),
])
- 第一个层表示进行随机的水平和垂直翻转,而第二个层表示按照 0.2 的弧度值进行随机旋转。
image = tf.expand_dims(images[i], 0)
plt.figure(figsize=(8, 8))
for i in range(9):
augmented_image = data_augmentation(image)
ax = plt.subplot(3, 3, i + 1)
plt.imshow(augmented_image[0])
plt.axis("off")
增强方式一
model = tf.keras.Sequential([
data_augmentation,
layers.Conv2D(16, 3, padding='same', activation='relu'),
layers.MaxPooling2D(),
])
增强方式二
batch_size = 32
AUTOTUNE = tf.data.AUTOTUNE
def prepare(ds):
ds = ds.map(lambda x, y: (data_augmentation(x, training=True), y), num_parallel_calls=AUTOTUNE)
return ds
train_ds = prepare(train_ds)
训练模型
model = tf.keras.Sequential([
layers.Conv2D(16, 3, padding='same', activation='relu'),
layers.MaxPooling2D(),
layers.Conv2D(32, 3, padding='same', activation='relu'),
layers.MaxPooling2D(),
layers.Conv2D(64, 3, padding='same', activation='relu'),
layers.MaxPooling2D(),
layers.Flatten(),
layers.Dense(128, activation='relu'),
layers.Dense(len(class_names))
])
model.compile(optimizer='adam',
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=['accuracy'])
epochs=20
history = model.fit(
train_ds,
validation_data=val_ds,
epochs=epochs
)
Epoch 1/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m18s[0m 215ms/step - accuracy: 0.5271 - loss: 0.9259 - val_accuracy: 0.6942 - val_loss: 0.6044
Epoch 2/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m16s[0m 210ms/step - accuracy: 0.7156 - loss: 0.5635 - val_accuracy: 0.8395 - val_loss: 0.3700
Epoch 3/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m20s[0m 200ms/step - accuracy: 0.8441 - loss: 0.3581 - val_accuracy: 0.8858 - val_loss: 0.2847
Epoch 4/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 197ms/step - accuracy: 0.8801 - loss: 0.2856 - val_accuracy: 0.9005 - val_loss: 0.2370
Epoch 5/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m21s[0m 201ms/step - accuracy: 0.9018 - loss: 0.2407 - val_accuracy: 0.9242 - val_loss: 0.1893
Epoch 6/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 200ms/step - accuracy: 0.9204 - loss: 0.2070 - val_accuracy: 0.9263 - val_loss: 0.2077
Epoch 7/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 199ms/step - accuracy: 0.9304 - loss: 0.1740 - val_accuracy: 0.9137 - val_loss: 0.2569
Epoch 8/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 202ms/step - accuracy: 0.9367 - loss: 0.1704 - val_accuracy: 0.9426 - val_loss: 0.1460
Epoch 9/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m21s[0m 198ms/step - accuracy: 0.9519 - loss: 0.1344 - val_accuracy: 0.9295 - val_loss: 0.1940
Epoch 10/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 200ms/step - accuracy: 0.9348 - loss: 0.1784 - val_accuracy: 0.9532 - val_loss: 0.1367
Epoch 11/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m21s[0m 200ms/step - accuracy: 0.9485 - loss: 0.1273 - val_accuracy: 0.9395 - val_loss: 0.1620
Epoch 12/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 194ms/step - accuracy: 0.9307 - loss: 0.1438 - val_accuracy: 0.9442 - val_loss: 0.1544
Epoch 13/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m21s[0m 190ms/step - accuracy: 0.9524 - loss: 0.1435 - val_accuracy: 0.9432 - val_loss: 0.1624
Epoch 14/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m16s[0m 207ms/step - accuracy: 0.9424 - loss: 0.1413 - val_accuracy: 0.9626 - val_loss: 0.1091
Epoch 15/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m16s[0m 202ms/step - accuracy: 0.9562 - loss: 0.1152 - val_accuracy: 0.9411 - val_loss: 0.1616
Epoch 16/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m16s[0m 207ms/step - accuracy: 0.9519 - loss: 0.1249 - val_accuracy: 0.9463 - val_loss: 0.1593
Epoch 17/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m16s[0m 208ms/step - accuracy: 0.9473 - loss: 0.1339 - val_accuracy: 0.9626 - val_loss: 0.1135
Epoch 18/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m16s[0m 212ms/step - accuracy: 0.9604 - loss: 0.1100 - val_accuracy: 0.9589 - val_loss: 0.1094
Epoch 19/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m20s[0m 205ms/step - accuracy: 0.9584 - loss: 0.1125 - val_accuracy: 0.9516 - val_loss: 0.1344
Epoch 20/20
[1m75/75[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m16s[0m 208ms/step - accuracy: 0.9539 - loss: 0.1224 - val_accuracy: 0.9668 - val_loss: 0.0874
评估模型
import matplotlib.pyplot as plt
plt.plot(history.history['accuracy'], label='accuracy')
plt.plot(history.history['val_accuracy'], label = 'val_accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.ylim([0.5, 1])
plt.legend(loc='lower right')
plt.show()
loss, acc = model.evaluate(test_ds)
print("Accuracy", acc)
[1m15/15[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m1s[0m 68ms/step - accuracy: 0.9499 - loss: 0.1181
Accuracy 0.9624999761581421
模型预测
import numpy as np
plt.figure(figsize=(18, 3))
plt.suptitle("预测结果展示")
for images, labels in val_ds.take(1):
for i in range(8):
ax = plt.subplot(1,8, i + 1)
plt.imshow(images[i].numpy())
img_array = tf.expand_dims(images[i], 0)
predictions = model.predict(img_array)
plt.title(class_names[np.argmax(predictions)])
plt.axis("off")
[1m1/1[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 133ms/step
[1m1/1[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 26ms/step
[1m1/1[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 34ms/step
[1m1/1[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 33ms/step
[1m1/1[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 36ms/step
[1m1/1[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 135ms/step
[1m1/1[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 33ms/step
[1m1/1[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 42ms/step
自定义增加函数
import random
def aug_img(image):
seed = (random.randint(0,9), 0)
stateless_random_brightness = tf.image.stateless_random_contrast(image, lower=0.1, upper=1.0, seed=seed)
return stateless_random_brightness
改变像素
image = tf.expand_dims(images[3]*255, 0)
print("Min and max pixel values:", image.numpy().min(), image.numpy().max())
plt.figure(figsize=(8, 8))
for i in range(9):
augmented_image = aug_img(image)
ax = plt.subplot(3, 3, i + 1)
plt.imshow(augmented_image[0].numpy().astype("uint8"))
plt.axis("off")
个人总结
- 学习了数据增强 包括旋转、翻转、缩放、改变图像对比度、改变像素