批量从文件夹中读取图片和训练tensorflow2.0

批量从文件夹中读取图片和训练

from PIL import Image
import matplotlib.pyplot as plt
import numpy as np
import os
import tensorflow as tf
from tensorflow.keras import datasets, layers, models

先导入第一张

X_train = Image.open('C:/Users/31035/Desktop/Desktop/data/livergate/livergate_0.png')
X_train = X_train.resize((512,512),Image.ANTIALIAS)
X_train = np.array(X_train)
X_train = np.expand_dims(X_train, axis=0)
print(X_train.shape)

用concatenate的方式导入指定数量的图片

for dirname, _, filenames in os.walk('C:/Users/31035/Desktop/Desktop/data/livergate'):
    for filename in filenames:
        if X_train.shape[0] > 357:
            break
        try:
            im = Image.open(os.path.join(dirname, filename))
            im = im.resize((512,512),Image.ANTIALIAS)
            image_array = np.array(im)
            image_array = np.expand_dims(image_array, axis=0)
            X_train = np.concatenate((X_train, image_array), axis=0)
        except:
            pass
print(str(X_train.shape[0]))

同理导入另一个文件夹图片

X_train_n = Image.open('C:/Users/31035/Desktop/Desktop/data/nolivergate/other_0.png')
X_train_n = X_train_n.resize((512,512),Image.ANTIALIAS)
X_train_n = np.array(X_train_n)
X_train_n = np.expand_dims(X_train_n, axis=0)
print(X_train_n.shape)

for dirname, _, filenames in os.walk('C:/Users/31035/Desktop/Desktop/data/nolivergate'):
    for filename in filenames:
        if X_train_n.shape[0] > 820:
            break
        try:
            im = Image.open(os.path.join(dirname, filename))
            im = im.resize((512,512),Image.ANTIALIAS)
            image_array = np.array(im)
            image_array = np.expand_dims(image_array, axis=0)
            X_train_n = np.concatenate((X_train_n, image_array), axis=0)
        except:
            pass
print(str(X_train_n.shape[0]))

设置样本标签

Y_train = np.ones((358))
Y_train = Y_train.astype(int)
Y_train_n = np.zeros((821))
Y_train_n = Y_train_n.astype(int)
X = np.concatenate((X_train,X_train_n),axis=0)

打乱

np.random.seed(200)
np.random.shuffle(X) 
np.random.seed(200)
np.random.shuffle(Y)
#不打乱的话，训练结果会很差

随机展示图片

class_names = ['nolivergate','livergate']

plt.figure(figsize=(10,10))
for i in range(25):
    plt.subplot(5,5,i+1)
    plt.xticks([])
    plt.yticks([])
    plt.grid(False)
    plt.imshow(X[i])
    # The CIFAR labels happen to be arrays, 
    # which is why you need the extra index
    plt.xlabel(class_names[Y[i]])
plt.show()

建立模型

model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(512,512,1)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))

model.add(layers.Flatten()) # 3维 转为 1维
model.add(layers.Dense(64, activation='relu'))  # 激活函数relu
model.add(layers.Dense(2, activation='softmax'))  # 激活函数softmax CIFAR有10个类别输出，所以softmax这里参数设置为10
# 再看看模型情况
model.summary()

编译&训练

model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])
history = model.fit(X_train, Y_train, epochs=10, 
                    validation_data=(X_val, Y_val))