python - 深度学习图像分类器架构配置

Question

嗨，我是机器学习新手，只知道一些基础知识以及事物应该如何工作。所以我正在查看这篇关于使用 Python、TensorFlow 和 Keras 进行深度学习的教程并获得了这些代码

import tensorflow as tf
from keras.datasets import cifar10
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Conv2D, MaxPooling2D
import numpy as np
import matplotlib.pyplot as plt
import os
import cv2

DATADIR = "C:/Users/Acer/imagerec/MRI"

CATEGORIES = ["yes", "no"]

for category in CATEGORIES:
    path = os.path.join(DATADIR,category)
    for img in os.listdir(path):
        img_array = cv2.imread(os.path.join(path,img) ,cv2.IMREAD_GRAYSCALE)
        plt.imshow(img_array, cmap='gray')


        break
    break
print(img_array)
print(img_array.shape)

IMG_SIZE = 224

new_array = cv2.resize(img_array, (IMG_SIZE, IMG_SIZE))
plt.imshow(new_array, cmap='gray')


training_data = []

def create_training_data():
    for category in CATEGORIES:
        path = os.path.join(DATADIR, category)
        class_num = CATEGORIES.index(category)
        for img in os.listdir(path):
            try:
                img_array = cv2.imread(os.path.join(path, img), cv2.IMREAD_GRAYSCALE)
                new_array = cv2.resize(img_array, (IMG_SIZE, IMG_SIZE))
                training_data.append([new_array, class_num])
            except Exception as e:
                pass


create_training_data()

print(len(training_data))

import random

random.shuffle(training_data)
for sample in training_data[:10]:
    print(sample[1])

X = []
y = []
for features, label in training_data:
    X.append(features)
    y.append(label)

X = np.array(X).reshape(-1, IMG_SIZE, IMG_SIZE, 1)

import pickle

pickle_in = open("X.pickle","rb")
X = pickle.load(pickle_in)



pickle_in = open("y.pickle","rb")
y = pickle.load(pickle_in)

X = X/255.0

model = Sequential()

model.add(Conv2D(256, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Conv2D(256, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Flatten())

model.add(Dense(64))
model.add(Activation('relu'))

model.add(Dense(1))
model.add(Activation('sigmoid'))

model.compile(loss='binary_crossentropy',
              optimizer='adam',
              metrics=['accuracy'])
model.fit(X, y, batch_size=15, epochs=20, validation_split=0.1)

from sklearn.metrics import confusion_matrix
pred = model.predict(X)
pred = np.round(pred)

conf = confusion_matrix(y, pred)

import seaborn as sns
sns.heatmap(conf, annot=True)

plt.show()

所以运行这些代码给了我很好的结果，验证准确度为 76.9%，我需要做的是将这段代码的模型更改为 VGG16、VGG19 和 mobilenet，但我不知道如何导入预训练模型，所以我决定制作我自己的模型并进行训练，所以我查看了 VGG16 和 VGG19 的架构，查看了有多少个 conv 和 maxpooling，并得出了这段代码

import tensorflow as tf
from keras.datasets import cifar10
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Conv2D, MaxPooling2D
import numpy as np
import matplotlib.pyplot as plt
import os
import cv2

DATADIR = "C:/Users/Acer/imagerec/EDA"

CATEGORIES = ["yes", "no"]

for category in CATEGORIES:
    path = os.path.join(DATADIR,category)
    for img in os.listdir(path):
        img_array = cv2.imread(os.path.join(path,img) ,cv2.IMREAD_GRAYSCALE)
        plt.imshow(img_array, cmap='gray')
        plt.show()

        break
    break
print(img_array)
print(img_array.shape)

IMG_SIZE = 224

new_array = cv2.resize(img_array, (IMG_SIZE, IMG_SIZE))
plt.imshow(new_array, cmap='gray')
plt.show()

training_data = []

def create_training_data():
    for category in CATEGORIES:
        path = os.path.join(DATADIR, category)
        class_num = CATEGORIES.index(category)
        for img in os.listdir(path):
            try:
                img_array = cv2.imread(os.path.join(path, img), cv2.IMREAD_GRAYSCALE)
                new_array = cv2.resize(img_array, (IMG_SIZE, IMG_SIZE))
                training_data.append([new_array, class_num])
            except Exception as e:
                pass


create_training_data()

print(len(training_data))

import random

random.shuffle(training_data)
for sample in training_data[:10]:
    print(sample[1])

X = []
y = []
for features, label in training_data:
    X.append(features)
    y.append(label)

X = np.array(X).reshape(-1, IMG_SIZE, IMG_SIZE, 1)

import pickle

pickle_in = open("X.pickle","rb")
X = pickle.load(pickle_in)



pickle_in = open("y.pickle","rb")
y = pickle.load(pickle_in)

X = X/255.0

model = Sequential()

model.add(Conv2D(64, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(Conv2D(64, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(MaxPooling2D(pool_size=(1, 1)))

model.add(Conv2D(128, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(Conv2D(128, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(MaxPooling2D(pool_size=(1, 1)))

model.add(Conv2D(256, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(Conv2D(256, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(Conv2D(256, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(MaxPooling2D(pool_size=(1, 1)))

model.add(Conv2D(512, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(Conv2D(512, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(Conv2D(512, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(MaxPooling2D(pool_size=(1, 1)))

model.add(Conv2D(512, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(Conv2D(512, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(Conv2D(512, (3, 3), input_shape=X.shape[1:]))
model.add(Activation('relu'))
padding='same'
model.add(MaxPooling2D(pool_size=(1, 1)))

model.add(Flatten())

model.add(Dense(64))
model.add(Activation('relu'))

model.add(Dense(1))
model.add(Activation('sigmoid'))
model.add(Dense(1))
model.add(Activation('sigmoid'))
model.add(Dense(1))
model.add(Activation('sigmoid'))

model.compile(loss='binary_crossentropy',
              optimizer='adam',
              metrics=['accuracy'])
model.fit(X, y, batch_size=15, epochs=1, validation_split=0.1)

from sklearn.metrics import confusion_matrix
pred = model.predict(X)
pred = np.round(pred)

conf = confusion_matrix(y, pred)

import seaborn as sns
sns.heatmap(conf, annot=True)

plt.show()

但是在任何时期运行这个总是给我 57.69% 的 val 准确率，我做错了什么吗？还是我做错了一切？

编辑所以我现在使用预训练模型

import tensorflow as tf
from keras.datasets import cifar10
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Conv2D, MaxPooling2D
import numpy as np
import matplotlib.pyplot as plt
import os
import cv2

DATADIR = "C:/Users/Acer/imagerec/MRI"

CATEGORIES = ["yes", "no"]

for category in CATEGORIES:
    path = os.path.join(DATADIR,category)
    for img in os.listdir(path):
        img_array = cv2.imread(os.path.join(path,img) ,cv2.IMREAD_GRAYSCALE)
        plt.imshow(img_array, cmap='gray')


        break
    break
print(img_array)
print(img_array.shape)

IMG_SIZE = 224

new_array = cv2.resize(img_array, (IMG_SIZE, IMG_SIZE))
plt.imshow(new_array, cmap='gray')


training_data = []

def create_training_data():
    for category in CATEGORIES:
        path = os.path.join(DATADIR, category)
        class_num = CATEGORIES.index(category)
        for img in os.listdir(path):
            try:
                img_array = cv2.imread(os.path.join(path, img), cv2.IMREAD_GRAYSCALE)
                new_array = cv2.resize(img_array, (IMG_SIZE, IMG_SIZE))
                training_data.append([new_array, class_num])
            except Exception as e:
                pass


create_training_data()

print(len(training_data))

import random

random.shuffle(training_data)
for sample in training_data[:10]:
    print(sample[1])

X = []
y = []
for features, label in training_data:
    X.append(features)
    y.append(label)

X = np.array(X).reshape(-1, IMG_SIZE, IMG_SIZE, 1)

import pickle

pickle_in = open("X.pickle","rb")
X = pickle.load(pickle_in)



pickle_in = open("y.pickle","rb")
y = pickle.load(pickle_in)

X = X/255.0


def input_shape(args):
    pass


from keras.applications.vgg16 import VGG16
from keras.models import Model
from keras.layers import Dense

vgg = VGG16(include_top=False, weights='imagenet', input_shape=(), pooling='avg')
x = vgg.output
x = Dense(1, activation='sigmoid')(x)
model = Model(vgg.input, x)
model.summary()

from sklearn.metrics import confusion_matrix
pred = model.predict(X)
pred = np.round(pred)

conf = confusion_matrix(y, pred)

import seaborn as sns
sns.heatmap(conf, annot=True)

plt.show()

model.save('64x2-CNN.model')

并收到此错误型号:“model_1”

<小时/>

层(类型)输出形状参数#

input_1(输入层)(无、无、无、3)0

<小时/>

block1_conv1(Conv2D)(无，无，无，64)1792

<小时/>

block1_conv2 (Conv2D)(无、无、无、64)36928

<小时/>

block1_pool (MaxPooling2D)(无、无、无、64)0

<小时/>

block2_conv1 (Conv2D)(无、无、无、128)73856

<小时/>

block2_conv2 (Conv2D)(无、无、无、128)147584

<小时/>

block2_pool(MaxPooling2D)(无，无，无，128)0

<小时/>

block3_conv1(Conv2D)(无、无、无、256)295168

<小时/>

block3_conv2(Conv2D)(无、无、无、256)590080

<小时/>

block3_conv3(Conv2D)(无、无、无、256)590080

<小时/>

block3_pool(MaxPooling2D)(无，无，无，256)0

<小时/>

block4_conv1(Conv2D)(无、无、无、512)1180160

<小时/>

block4_conv2(Conv2D)(无、无、无、512)2359808

<小时/>

block4_conv3(Conv2D)(无、无、无、512)2359808

<小时/>

block4_pool(MaxPooling2D)(无，无，无，512)0

<小时/>

block5_conv1(Conv2D)(无、无、无、512)2359808

<小时/>

block5_conv2(Conv2D)(无、无、无、512)2359808

<小时/>

block5_conv3(Conv2D)(无、无、无、512)2359808

<小时/>

block5_pool(MaxPooling2D)(无，无，无，512)0

<小时/>

global_average_pooling2d_1((无，512)0

<小时/>

dense_1(密集)(无，1)513

总参数:14,715,201可训练参数:14,715,201不可训练参数:0

<小时/>

回溯(最近一次调用最后一次): 文件“C:/Users/Acer/PycharmProjects/condas/UwU.py”，第 95 行，位于 pred = model.predict(X) 文件“C:\Users\Acer\Anaconda3\envs\condas\lib\site-packages\keras\engine\training.py”，第 1441 行，在预测中 x, _, _ = self._standardize_user_data(x) 文件“C:\Users\Acer\Anaconda3\envs\condas\lib\site-packages\keras\engine\training.py”，第 579 行，位于 _standardize_user_data 异常前缀='输入') 文件“C:\Users\Acer\Anaconda3\envs\condas\lib\site-packages\keras\engine\training_utils.py”，第 145 行，位于 standardize_input_data 中 str(数据形状))ValueError:检查输入时出错:预期 input_1 具有形状 (None, None, 3)，但得到的数组具有形状 (50, 50, 1)

进程已完成，退出代码为 1

Answer 1

在 keras 顺序模型中，只有第一层需要知道它应该期望的 input_shape ，在您的情况下是它的 Conv2D 层。此外，使用 sigmoid 激活添加多个 Dense 层是没有意义的。

引用这个

model = Sequential([
Conv2D(64, (3, 3), input_shape=input_shape, padding='same', activation='relu'),
Conv2D(64, (3, 3), activation='relu', padding='same'),
MaxPooling2D(pool_size=(2, 2), strides=(2, 2)),
Conv2D(128, (3, 3), activation='relu', padding='same'),
Conv2D(128, (3, 3), activation='relu', padding='same',),
MaxPooling2D(pool_size=(2, 2), strides=(2, 2)),
Conv2D(256, (3, 3), activation='relu', padding='same',),
Conv2D(256, (3, 3), activation='relu', padding='same',),
Conv2D(256, (3, 3), activation='relu', padding='same',),
MaxPooling2D(pool_size=(2, 2), strides=(2, 2)),
Conv2D(512, (3, 3), activation='relu', padding='same',),
Conv2D(512, (3, 3), activation='relu', padding='same',),
Conv2D(512, (3, 3), activation='relu', padding='same',),
MaxPooling2D(pool_size=(2, 2), strides=(2, 2)),
Conv2D(512, (3, 3), activation='relu', padding='same',),
Conv2D(512, (3, 3), activation='relu', padding='same',),
Conv2D(512, (3, 3), activation='relu', padding='same',),
MaxPooling2D(pool_size=(2, 2), strides=(2, 2)),
Flatten(),
Dense(4096, activation='relu'),
Dense(4096, activation='relu'),
Dense(1, activation='sigmoid')
])

或者，您可以使用 keras 应用程序中预训练的 VGG 模型。

from keras.applications.vgg16 import VGG16
from keras.models import Model
from keras.layers import Dense

vgg = VGG16(include_top=False, weights='imagenet', input_shape=(), pooling='avg')
x = vgg.output
x = Dense(1, activation='sigmoid')(x)
model = Model(vgg.input, x)
model.summary()