In [62]:
Im = ['Negative', 'Positive']
num = [num_n, num_p]
plt.figure(figsize=(10, 6))
x = np.arange(2)
plt.bar(Im, num)
Out[62]:
<BarContainer object of 2 artists>
import matplotlib.pyplot as plt
import seaborn as sns
import keras
from keras.models import Sequential
from keras.layers import Dense, Conv2D , MaxPool2D , Flatten , Dropout
from keras.preprocessing.image import ImageDataGenerator
from keras.optimizers import Adam, RMSprop, Adagrad
from keras.layers import BatchNormalization
from sklearn.metrics import classification_report,confusion_matrix
import tensorflow as tf
import cv2
import os
import time
import numpy as np
import warnings
warnings.filterwarnings('ignore')
print(keras.__version__)
print(cv2.__version__)
print(sns.__version__)
print(np.__version__)
2.10.0 4.6.0 0.12.1 1.23.1
os.getcwd()
'D:\\Github\\DeepLearning_Basic_Class\\00_part06_02_CrackDetection'
labels = ['Negative', 'Positive']
img_size = 120
def read_images(data_dir):
data = []
for label in labels:
path = os.path.join(data_dir, label)
class_num = labels.index(label)
for img in os.listdir(path):
try:
img_arr = cv2.imread(os.path.join(path, img), cv2.IMREAD_GRAYSCALE)
resized_arr = cv2.resize(img_arr, (img_size, img_size)) # 이미지 변경
data.append([resized_arr, class_num])
except Exception as e:
print(e)
return np.array(data)
Dataset = read_images('../datasets/Surface_Crack_Detection_small')
print( Dataset.shape )
print( Dataset[0][0], Dataset[0][1]) # 피처와 Target
(990, 2) [[172 161 149 ... 183 182 184] [174 166 147 ... 176 175 177] [176 171 170 ... 180 176 176] ... [163 166 170 ... 175 175 173] [158 155 164 ... 172 173 171] [161 153 154 ... 170 172 170]] 0
# 폴더 내의 파일 수
data_dir = '../datasets/Surface_Crack_Detection_small'
path_negative = os.path.join(data_dir, "Negative")
path_positive = os.path.join(data_dir, "Positive")
print( len(os.listdir(path_negative) ))
print( len(os.listdir(path_positive) ))
num_n = len(os.listdir(path_negative) )
num_p = len(os.listdir(path_positive) )
num = [num_n, num_p]
num
490 500
[490, 500]
Im = ['Negative', 'Positive']
num = [num_n, num_p]
plt.figure(figsize=(10, 6))
x = np.arange(2)
plt.bar(Im, num)
<BarContainer object of 2 artists>
x = []
y = []
for feature, label in Dataset:
x.append(feature)
y.append(label)
x = np.array(x).reshape(-1, img_size, img_size, 1)
x = x / 255
y = np.array(y)
plt.subplot(1, 2, 1)
plt.imshow(x[300].reshape(img_size, img_size), cmap='gray')
plt.axis('off')
plt.subplot(1, 2, 2)
plt.imshow(x[500].reshape(img_size, img_size), cmap='gray')
plt.axis('off')
(-0.5, 119.5, 119.5, -0.5)
x.shape[1:]
(120, 120, 1)
model = Sequential()
model.add(Conv2D(64,3,padding="same", activation="relu", input_shape = x.shape[1:]))
model.add(MaxPool2D())
model.add(Conv2D(64, 3, padding="same", activation="relu"))
model.add(MaxPool2D())
model.add(Conv2D(128, 3, padding="same", activation="relu"))
model.add(MaxPool2D())
model.add(Flatten())
model.add(Dense(256,activation="relu"))
model.add(Dropout(0.5))
model.add(BatchNormalization())
model.add(Dense(1, activation="sigmoid"))
model.summary()
Model: "sequential_5"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_15 (Conv2D) (None, 120, 120, 64) 640
max_pooling2d_15 (MaxPoolin (None, 60, 60, 64) 0
g2D)
conv2d_16 (Conv2D) (None, 60, 60, 64) 36928
max_pooling2d_16 (MaxPoolin (None, 30, 30, 64) 0
g2D)
conv2d_17 (Conv2D) (None, 30, 30, 128) 73856
max_pooling2d_17 (MaxPoolin (None, 15, 15, 128) 0
g2D)
flatten_5 (Flatten) (None, 28800) 0
dense_10 (Dense) (None, 256) 7373056
dropout_5 (Dropout) (None, 256) 0
batch_normalization_5 (Batc (None, 256) 1024
hNormalization)
dense_11 (Dense) (None, 1) 257
=================================================================
Total params: 7,485,761
Trainable params: 7,485,249
Non-trainable params: 512
_________________________________________________________________
start = time.time()
opt = Adam(lr=1e-5)
model.compile(loss="binary_crossentropy",
optimizer=opt, metrics=["accuracy"])
history = model.fit(x, y, epochs = 15,
batch_size = 128, validation_split = 0.25, verbose=1)
print("소요시간", time.time() - start )
Epoch 1/15 6/6 [==============================] - 14s 2s/step - loss: 0.6723 - accuracy: 0.6051 - val_loss: 0.8110 - val_accuracy: 0.0000e+00 Epoch 2/15 6/6 [==============================] - 13s 2s/step - loss: 0.5471 - accuracy: 0.7129 - val_loss: 0.8786 - val_accuracy: 0.0000e+00 Epoch 3/15 6/6 [==============================] - 13s 2s/step - loss: 0.5044 - accuracy: 0.7615 - val_loss: 0.9086 - val_accuracy: 0.0000e+00 Epoch 4/15 6/6 [==============================] - 13s 2s/step - loss: 0.5025 - accuracy: 0.7655 - val_loss: 0.9090 - val_accuracy: 0.0000e+00 Epoch 5/15 6/6 [==============================] - 13s 2s/step - loss: 0.4722 - accuracy: 0.7978 - val_loss: 0.8947 - val_accuracy: 0.0000e+00 Epoch 6/15 6/6 [==============================] - 13s 2s/step - loss: 0.4586 - accuracy: 0.7911 - val_loss: 0.8742 - val_accuracy: 0.0000e+00 Epoch 7/15 6/6 [==============================] - 13s 2s/step - loss: 0.4354 - accuracy: 0.8059 - val_loss: 0.8462 - val_accuracy: 0.0000e+00 Epoch 8/15 6/6 [==============================] - 13s 2s/step - loss: 0.4236 - accuracy: 0.8127 - val_loss: 0.8175 - val_accuracy: 0.0000e+00 Epoch 9/15 6/6 [==============================] - 13s 2s/step - loss: 0.4013 - accuracy: 0.8369 - val_loss: 0.7776 - val_accuracy: 0.0000e+00 Epoch 10/15 6/6 [==============================] - 13s 2s/step - loss: 0.3868 - accuracy: 0.8464 - val_loss: 0.7484 - val_accuracy: 0.0444 Epoch 11/15 6/6 [==============================] - 13s 2s/step - loss: 0.3750 - accuracy: 0.8544 - val_loss: 0.7214 - val_accuracy: 0.2540 Epoch 12/15 6/6 [==============================] - 12s 2s/step - loss: 0.3551 - accuracy: 0.8625 - val_loss: 0.6943 - val_accuracy: 0.5403 Epoch 13/15 6/6 [==============================] - 13s 2s/step - loss: 0.3403 - accuracy: 0.8733 - val_loss: 0.6664 - val_accuracy: 0.7540 Epoch 14/15 6/6 [==============================] - 13s 2s/step - loss: 0.3122 - accuracy: 0.8881 - val_loss: 0.6420 - val_accuracy: 0.8831 Epoch 15/15 6/6 [==============================] - 13s 2s/step - loss: 0.3019 - accuracy: 0.8922 - val_loss: 0.6226 - val_accuracy: 0.9516 소요시간 194.53148007392883
plt.figure(figsize=(12, 12))
plt.style.use('ggplot')
plt.subplot(2,2,1)
plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])
plt.title('Accuracy of the Model')
plt.ylabel('Accuracy', fontsize=12)
plt.xlabel('Epoch', fontsize=12)
plt.legend(['train accuracy', 'validation accuracy'],
loc='lower right', prop={'size': 12})
plt.subplot(2,2,2)
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('Loss of the Model')
plt.ylabel('Loss', fontsize=12)
plt.xlabel('Epoch', fontsize=12)
plt.legend(['train loss', 'validation loss'],
loc='best', prop={'size': 12})
<matplotlib.legend.Legend at 0x1c336b075e0>
x.shape, y.shape
((990, 120, 120, 1), (990,))
from sklearn.metrics import classification_report,confusion_matrix
predictions = (model.predict(x) > 0.5).astype("int32")
print( predictions.shape )
31/31 [==============================] - 3s 106ms/step (990, 1)
print(classification_report(y, predictions, target_names = ['Negative','Positive']))
precision recall f1-score support
Negative 0.95 0.89 0.92 490
Positive 0.90 0.96 0.92 500
accuracy 0.92 990
macro avg 0.92 0.92 0.92 990
weighted avg 0.92 0.92 0.92 990