[Kaggle]Pneumonia/Normal Classification(CNN)

220320 작성

<본 블로그는 Kaggle 을 참고해서 저만의 풀이를 작성하였습니다>

https://www.kaggle.com/code/rafetcan/pneumonia-normal-cnn-model/notebook

🩺 Pneumonia/Normal ? - CNN Model 🩺

 

 

 

 

 

1. Data

- 3 folders (train, test, val)

- each image category (Pneumonia/Normal)

- 5,863 X-Ray images (JPEG) and 2 categories (Pneumonia/Normal)

 

 

 

2. import libraries

import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import seaborn as sns
import matplotlib.pyplot as plt
import os
from glob import glob
import cv2

from keras.models import Sequential 
from keras.layers import Conv2D, MaxPooling2D, Activation, Dropout, Flatten, Dense
from keras.preprocessing.image import ImageDataGenerator, img_to_array, load_img
from tensorflow.keras.callbacks import EarlyStopping

# import warnings
import warnings
# filter warnings
warnings.filterwarnings('ignore')

import os
for dirname, _, filenames in os.walk('/kaggle/input'):
    for filename in filenames:
        print(os.path.join(dirname, filename))

 

 

 

3. file 경로 지정

train_dir = '/content/drive/MyDrive/2022/project/Data/chest_xray/train'
test_dir =  '/content/drive/MyDrive/2022/project/Data/chest_xray/test'
val_dir = '/content/drive/MyDrive/2022/project/Data/chest_xray/val'

 

 

 

4. read image & train_test split

def picture_separation(folder):
    y = []
    x = []
    image_list = []

    for foldername in os.listdir(folder):
        if not foldername.startswith('.'):
            if foldername == "NORMAL":
                label = 0
            elif foldername == "PNEUMONIA":
                label = 1
            else:
                label = 2
                
            for image_filename in os.listdir(folder + "/"+ foldername):
                img_file = cv2.imread(folder + "/" + foldername + '/' + image_filename,0)               
                

                if img_file is not None:
                    img = cv2.resize(img_file,(64,64))
                    img_arr = img_to_array(img) / 255
                    x.append(img_arr)
                    y.append(label)
                    image_list.append(foldername + '/' + image_filename)
                                        
    X = np.asarray(x)
    y = np.asarray(y)

    return X,y,image_list

 

- train_df

X_train, y_train, img_train = picture_separation(train_dir)

train_df = pd.DataFrame(img_train, columns=["images"])
train_df["target"] = y_train

 

train_df.info()

 

sns.countplot(train_df["target"])
plt.title("NORMAL/PNOMONİA")
plt.show()
print(train_df["target"].value_counts())

 

 

 

 

- val_df

X_val, y_val, img_val = picture_separation(val_dir)

val_df = pd.DataFrame(img_val, columns=["images"])
val_df["target"] = y_val

 

 

sns.countplot(val_df["target"])
plt.title("NORMAL/PNOMONİA")
plt.show()
print(val_df["target"].value_counts())

 

 

 

- test_df

X_test, y_test, img_test = picture_separation(test_dir)

test_df = pd.DataFrame(img_test, columns=["images"])
test_df["target"] = y_test

 

sns.countplot(test_df["target"])
plt.title("NORMAL/PNOMONİA")
plt.show()
print(test_df["target"].value_counts())

 

 

 

- full_data

 full_data = pd.concat([train_df, test_df, val_df], axis=0, ignore_index=True)

 

print(full_data.head())
print(full_data.tail())

 

- 시각화

plt.figure(figsize=(12,8))

img = load_img(train_dir + "/" + full_data["images"][3875])
plt.imshow(img)
plt.title("NORMAL", color = "green", size = 14)
plt.grid(color='#CCCCCC', linestyle='--')
plt.show()

 

plt.figure(figsize=(10,7))

img = load_img(train_dir + "/" + full_data["images"][0])
plt.imshow(img)
plt.title("PNEUMONIA", color = "green", size = 14)
plt.grid(color='#CCCCCC', linestyle='--')
plt.show()

 

 

 

 

sns.countplot(full_data["target"])
plt.title("NORMAL/PNOMONİA")
plt.show()
print(full_data["target"].value_counts())

 

 

 

 

 

5. 시각화

plt.figure(figsize=(15,8))

plt.subplot(2,3,1) 
img = load_img(train_dir + "/" + full_data["images"][0])
plt.imshow(img)
plt.title("PNEUMONIA", color = "blue", size = 14)
plt.axis("off")


plt.subplot(2,3,2) 
img = load_img(train_dir + "/" + full_data["images"][1])
plt.imshow(img)
plt.title("PNEUMONIA", color = "blue", size = 14)
plt.axis("off")

plt.subplot(2,3,3) 
img = load_img(train_dir + "/" + full_data["images"][10])
plt.imshow(img)
plt.title("PNEUMONIA", color = "blue", size = 14)
plt.axis("off")

plt.subplot(2,3,4) 
img = load_img(train_dir + "/" + full_data["images"][3875])
plt.imshow(img)
plt.title("NORMAL", color = "green", size = 14)
plt.axis("off")

plt.subplot(2,3,5) 
img = load_img(train_dir + "/" + full_data["images"][3876])
plt.imshow(img)
plt.title("NORMAL", color = "green", size = 14)
plt.axis("off")

plt.subplot(2,3,6) 
img = load_img(train_dir + "/" + full_data["images"][3877])
plt.imshow(img)
plt.title("NORMAL", color = "green", size = 14)
plt.axis("off")

plt.suptitle("NORMAL or PNEUMONIA", size = 16, color = "darkred")
plt.show()

 

 

 

X_train.shape

 

className = glob(train_dir + '/*' )
numberOfClass = len(className)
print("NumberOfClass: ",numberOfClass)

 

 

 

6. 데이터 증강

batch_size = 32

 

train_datagen = ImageDataGenerator(rescale= 1./255,
                   shear_range = 0.3,
                   horizontal_flip=True,
                   zoom_range = 0.3)

 

test_datagen = ImageDataGenerator(rescale= 1./255)

val_datagen = ImageDataGenerator(rescale= 1./255)

 

- train

train_generator = train_datagen.flow_from_directory(
        train_dir, 
        target_size=(64, 64),
        batch_size = batch_size,
        color_mode = "grayscale",
        class_mode= "binary")

- test

test_generator = test_datagen.flow_from_directory(
        test_dir, 
        target_size=(64, 64),
        batch_size = batch_size,
        color_mode = "grayscale",
        class_mode= "binary")

 

- val

val_generator = test_datagen.flow_from_directory(
        val_dir, 
        target_size=(64, 64),
        batch_size = batch_size,
        color_mode = "grayscale",
        class_mode= "binary")

 

 

7. model 생성

X_train.shape[1:]

 

model = Sequential()
model.add(Conv2D(32,(3,3),input_shape = X_train.shape[1:]))
model.add(Activation("relu"))
model.add(MaxPooling2D())

model.add(Conv2D(32,(3,3)))
model.add(Activation("relu"))
model.add(MaxPooling2D())

model.add(Conv2D(64,(3,3)))
model.add(Activation("relu"))
model.add(MaxPooling2D())

model.add(Flatten())
model.add(Dense(1024))
model.add(Activation("relu"))
model.add(Dropout(0.4))
model.add(Dense(1)) # output
model.add(Activation("sigmoid"))

model.compile(loss = "binary_crossentropy",
              optimizer = "rmsprop",
              metrics = ["accuracy"])

 

- early stopping

: 과대적합 되기전에 early! stop 하기

early_stopping = EarlyStopping(monitor='val_loss', mode='min', verbose=1,patience=2)

history = model.fit_generator(
                    train_generator,
                    steps_per_epoch=5216//32,
                    epochs=20,
                    validation_data=test_generator,
                    validation_steps=624//32,
                    callbacks=[early_stopping])

 

print("Accuracy of the model is - " , model.evaluate_generator(test_generator)[1]*100 , "%")
print("Loss of the model is - " , model.evaluate_generator(test_generator)[0])

 

 

8. 시각화

history.history.keys()

 

- loss

plt.figure()
plt.plot(history.history["loss"],label = "Train Loss")
plt.plot(history.history["val_loss"],label = "Validation Loss")
plt.legend()
plt.show()

 

- accuracy

plt.figure()
plt.plot(history.history["accuracy"],label = "Train Accuracy")
plt.plot(history.history["val_accuracy"],label = "Validation Accuracy")
plt.legend()
plt.show()