0

我设法为这段代码添加了扩充，但我不知道如何以及从哪里开始。数据集只有 200 张图像，所以我需要增强。所以我需要至少 5 次增强来制作至少 1000 张图像。

这是代码：

from keras.preprocessing.image import ImageDataGenerator

aug = ImageDataGenerator(rescale=1./255,
                         shear_range=0.2,
                         horizontal_flip=True,
                         rotation_range=45,
                         vertical_flip=True,
                         fill_mode='nearest',
                         height_shift_range=0.5,
                         width_shift_range=0.5
                         )

import os

base_dir ='/content/drive/My Drive/RESEARCH CENTER/Dataset'

train_dir = os.path.join(base_dir, 'train')
test_dir = os.path.join(base_dir, 'test')

train_normal_dir = os.path.join(train_dir, 'normal')
train_hem_dir = os.path.join(train_dir, 'hemorrhage')

test_normal_dir = os.path.join(test_dir, 'normal')
test_hem_dir = os.path.join(test_dir, 'hemorrhage')

print("total training normal brain images:", len(os.listdir(train_normal_dir)))
print("total training hemorrhage brain images:", len(os.listdir(train_hem_dir)))
print("total test normal brain images:", len(os.listdir(test_normal_dir)))
print("total test hemorrhage brain images:", len(os.listdir(test_hem_dir)))

import cv2
import numpy as np

# Gather data train
train_data = [] # disimpan dalam bentuk array
train_label = []
for r, d, f in os.walk(train_dir):
    for file in f:
        if ".png" in file:
            imagePath = os.path.join(r, file)
            image = cv2.imread(imagePath)
            image = cv2.resize(image, (224,224))
            color_mode="grayscale"
            label = imagePath.split(os.path.sep)[-2]
            train_data.append(image)
            train_label.append(label)

train_data = np.array(train_data)
train_label = np.array(train_label)

val_data = []
val_label = []
for r, d, f in os.walk(test_dir):
    for file in f:
        if ".png" in file:
            imagePath = os.path.join(r, file)
            image = cv2.imread(imagePath)
            image = cv2.resize(image, (224,224))
            color_mode="grayscale"
            label = imagePath.split(os.path.sep)[-2]
            val_label.append(label)
            val_data.append(image)

val_data = np.array(val_data)
val_label = np.array(val_label)

print("Train Data = ", train_data.shape)
print("Train Label = ", train_label.shape)
print("Validation Data = ", val_data.shape)
print("Validation Label = ", val_label.shape)

from sklearn.preprocessing import LabelEncoder

print("Data sebelum di-normalisasi ", train_data[0][0][0])

x_train = train_data.astype('float32') / 255.0
x_val = val_data.astype('float32') / 255.0
print("Data setelah di-normalisasi ", x_train[0][0][0])

from sklearn.preprocessing import LabelEncoder
print("Label sebelum di-encoder ", train_label[50:110])
lb = LabelEncoder()
y_train = lb.fit_transform(train_label)
y_val = lb.fit_transform(val_label)

print("Label setelah di-encoder", y_train[50:110])
print("Class Label encoder", lb.classes_)

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import InputLayer, Dense, Conv2D, MaxPool2D, GlobalMaxPool2D, GlobalAvgPool2D, Flatten, Dropout

model = Sequential()

model.add(InputLayer(input_shape=[224,224,3]))
model.add(Conv2D(filters=16, kernel_size=3, strides=1, padding='same', activation='relu'))
model.add(MaxPool2D(pool_size=2, padding='same'))
model.add(Conv2D(filters=32, kernel_size=3, strides=1, padding='same', activation='relu'))
model.add(MaxPool2D(pool_size=2, padding='same'))
model.add(Conv2D(filters=64, kernel_size=3, strides=1, padding='same', activation='relu'))
model.add(MaxPool2D(pool_size=2, padding='same'))
#model.add(Dropout(0.2))
model.add(GlobalMaxPool2D())
model.add(Flatten())

model.add(Dense(128, activation='relu'))
model.add(Dense(1, activation='sigmoid'))

from tensorflow.keras.optimizers import Adam

model.compile(optimizer=Adam(lr=0.001), 
              loss='binary_crossentropy', 
              metrics=['acc'])

H = model.fit(x_train, y_train, batch_size=20, epochs=15, validation_data=(x_val, y_val))

有人可以帮我添加增强功能吗？增强功能会在这段代码中起作用吗？

4

0 回答 0