【人工智能项目】深度学习实现胸腔X光肺炎检测

本次主要是任务：训练模型正确识别肺炎X光图片，0=正常，1=肺炎。

那么代码走起！！！

导包

# 导包
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from glob import glob
from keras.models import *
from keras.layers import *
from keras.preprocessing.image import *
from keras.utils import *
from keras.optimizers import *
from keras.applications import *
from keras.applications import  imagenet_utils
from keras.callbacks import EarlyStopping,ReduceLROnPlateau,ModelCheckpoint,LearningRateScheduler
from efficientnet.keras import EfficientNetB3

EDA探索数据

# EDA探索数据
import matplotlib.pyplot as plt
import seaborn as sns
import ostrain_dir_path = "./xray_dataset/train"
train_normal_path = os.path.join(train_dir_path,"NORMAL")
train_pneumonia_path = os.path.join(train_dir_path,"PNEUMONIA")
train_normal_length = len(os.listdir(train_normal_path))
train_pneumonia_length = len(os.listdir(train_pneumonia_path))sns.set_style("whitegrid")
sns.barplot(x = ["NORMAL","PNEUMONIA"],y = np.array([train_normal_length,train_pneumonia_length]))

labels = "Normal","Pneumonia"
sizes = [train_normal_length,train_pneumonia_length]
colors = ["green","red"]
explode = (0.1,0)plt.pie(sizes,labels=labels,colors=colors,explode=explode)
plt.axis("equal")
plt.show()

读取数据

img_rows,img_cols = 224,224
batch_size= 2

# ImageDataGenerator读取数据
from keras.preprocessing.image import ImageDataGenerator
import ostrain_path = "C:\\Users\\LvChaoZhang\\00contest\\09 ai研习社胸腔X光肺炎检测\\xray_dataset\\train"# train_datagen = ImageDataGenerator(rescale = 1./255,
#                                                               shear_range = 0.2,
#                                                               zoom_range = 0.2,
#                                                               horizontal_flip = True,
#                                                               vertical_flip=True,
#                                                               validation_split=0.2,
#                                                               fill_mode="nearest")
train_datagen = ImageDataGenerator(rescale = 1./255,validation_split=0.2)train_set = train_datagen.flow_from_directory(train_path,target_size=(img_rows,img_cols),batch_size=batch_size,color_mode="rgb",class_mode="binary",shuffle=True,seed=2019,subset="training")
val_set = train_datagen.flow_from_directory(train_path,target_size=(img_rows,img_cols),batch_size=batch_size,class_mode="binary",color_mode="rgb",shuffle=False,subset="validation")

Found 3280 images belonging to 2 classes.
Found 819 images belonging to 2 classes.

train_set.class_indices

{'NORMAL': 0, 'PNEUMONIA': 1}

模型

# 初始模型
model = Sequential()model.add(Conv2D(32,(3,3),input_shape=(img_rows,img_cols,3),activation='relu'))
model.add(Conv2D(32,(3,3),activation='relu'))model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.2))model.add(Conv2D(64,(3,3),activation='relu'))
model.add(Conv2D(64,(3,3),activation='relu'))model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.2))model.add(Conv2D(128,(3,3),activation='relu'))
model.add(Conv2D(128,(3,3),activation='relu'))model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.2))model.add(Flatten())
# model.add(Dense(units=256,activation='relu'))
# model.add(Dropout(0.2))
model.add(Dense(units=256,activation='relu'))
model.add(Dropout(0.2))
model.add(Dense(units=1,activation='sigmoid'))

# 模型 VGG161.0
base_model = VGG16(weights="imagenet",include_top=False,input_shape=(img_rows,img_cols,3))
for layer in base_model.layers:layer.trainable=False
x = base_model.output
x = GlobalAveragePooling2D()(x)
x = Dense(128,activation="relu")(x)
x = Dropout(0.5)(x)
out = Dense(1,activation="sigmoid")(x)
model = Model(base_model.input,out)

# 模型4.0model = Sequential()scale = 32# convolution 1
model.add(Conv2D(scale,(3,3),input_shape=(img_rows,img_cols,3)))
model.add(LeakyReLU(alpha=0.1))
# convolution 2
model.add(Conv2D(2*scale,(3,3)))
model.add(LeakyReLU(alpha=0.1))
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.1))# convolution 3
model.add(Conv2D(3*scale,(3,3)))
model.add(LeakyReLU(alpha=0.1))
# convolution 4
model.add(Conv2D(4*scale,(3,3)))
model.add(LeakyReLU(alpha=0.1))
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.1))# convolution 5
model.add(Conv2D(5*scale,(3,3)))
model.add(LeakyReLU(alpha=0.1))
# convolution 6
model.add(Conv2D(3*scale,(3,3)))
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.1))# convolution 7
model.add(Conv2D(6*scale,(3,3)))
model.add(LeakyReLU(alpha=0.1))
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.1))# convolution 8
model.add(Conv2D(7*scale,(3,3)))
model.add(LeakyReLU(alpha=0.1))
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.1))# flatten layer
model.add(Flatten())# first dense layer
model.add(Dense(units=15*scale))
model.add(LeakyReLU(alpha=0.1))
model.add(Dropout(0.5))# second dense layer
model.add(Dense(units=15*scale))
model.add(LeakyReLU(alpha=0.1))
model.add(Dropout(0.5))# third dense layer
model.add(Dense(units=15*scale))
model.add(LeakyReLU(alpha=0.1))
model.add(Dropout(0.5))# output layer
model.add(Dense(1,activation="sigmoid"))

# 试验模型Efficientnet
base_model = EfficientNetB3(weights="imagenet",include_top=False,input_shape=(img_rows,img_cols,3))
for layer in base_model.layers:layer.trainable=True
x = base_model.output
x = GlobalAveragePooling2D()(x)
x = Dense(10,activation="relu")(x)
x = Dropout(0.5)(x)
out = Dense(1,activation="sigmoid")(x)
model = Model(base_model.input,out)

模型编译

# 模型编译
import keras.backend as K# focal loss
def focal_loss(alpha=0.25,gamma=2.0):def focal_crossentropy(y_true, y_pred):bce = K.binary_crossentropy(y_true, y_pred)y_pred = K.clip(y_pred, K.epsilon(), 1.- K.epsilon())p_t = (y_true*y_pred) + ((1-y_true)*(1-y_pred))alpha_factor = 1modulating_factor = 1alpha_factor = y_true*alpha + ((1-alpha)*(1-y_true))modulating_factor = K.pow((1-p_t), gamma)# compute the final loss and returnreturn K.mean(alpha_factor*modulating_factor*bce, axis=-1)return focal_crossentropy

model.compile(loss=focal_loss(),metrics=["accuracy"],optimizer=Adam(lr=0.0001))

# Implement Learning rate decay
checkpoint = ModelCheckpoint("chest_xray_cnn3.h5",monitor="val_loss",mode="min",save_best_only = True,verbose=1)earlystop = EarlyStopping(monitor = 'val_loss', min_delta = 0, patience = 5,verbose = 1,restore_best_weights = True)reduce_lr = ReduceLROnPlateau(monitor = 'val_loss',factor = 0.2,patience = 3,verbose = 1)#min_delta = 0.00001)callbacks = [earlystop, checkpoint, reduce_lr]

# 训练
history = model.fit_generator(train_set,steps_per_epoch=3280//batch_size,epochs = 25,validation_data=val_set,validation_steps=819 //batch_size,shuffle=True,callbacks=callbacks,class_weight={0:1.84307554 ,1:0.68613994})

# 训练曲线
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline#Accuracy
plt.plot(history.history["acc"])
plt.plot(history.history["val_acc"])
plt.title("Model Accuracy")
plt.xlabel("Epoch")
plt.ylabel("Accuracy")
plt.legend(["Training Set","Validation Set"],loc="upper left")# Loss
plt.plot(history.history["loss"])
plt.plot(history.history["val_loss"])
plt.title("Model loss")
plt.xlabel("Epoch")
plt.ylabel("Loss")
plt.legend(["Training set","Test set"],loc="upper left")plt.show()

预测

import cv2
import numpy as np
import pandas as pd# 定义读取图片函数
def  get_img(file_path,img_rows,img_cols):img = cv2.imread(file_path)img = cv2.resize(img,(img_rows,img_cols),interpolation=cv2.INTER_CUBIC)
#     if img.shape[2] == 1:
#         img = np.dstack([img,img,img])img = cv2.cvtColor(img,cv2.COLOR_BGR2RGB)img = img.astype(np.float32)/255.return img

# 定义加载测试集的函数
def load_test_data(test_path,img_rows,img_cols):x_test_id = []x_test = []img_names = os.listdir(test_path)img_names.sort(key=lambda x:int(x[:-4]))for img_name in img_names:feature = get_img(os.path.join(test_path,img_name),img_rows,img_cols)id = img_namex_test_id.append(id)x_test.append(feature)#对x_test进行转换x_test = np.array(x_test)return x_test,x_test_id

#test_dir_path = "./xray_dataset/test/"
test_dir_path = "D:\\01\\05CT\\ct\\test\\"

x_test,x_id = load_test_data(test_dir_path,img_rows,img_cols)

model.load_weights("chest_xray_cnn1.h5")

y_pred = model.predict(x_test)

y_pred[:10]

for i in range(len(x_test)):if y_pred[i]>0.5:y_pred[i]=1;else:y_pred[i]=0;

id = np.arange(len(x_test))

df = pd.DataFrame({"id":id,"predict":y_pred[:,0]})

df.to_csv("submit1.csv",index=None,header=None)

小结

那么本次到此结束！下次见瓷！

【人工智能项目】深度学习实现胸腔X光肺炎检测相关推荐

“创新实践”项目介绍3：《利用深度学习进行VR手柄光点检测》
指导教师点评这个项目来源于企业的实际需求,所采用的技术方案是最新的深度学习物体检测技术.技术方案具有一定的难度,很有挑战.通过这个项目,学生可以接触真正的需求,提升分析问题和解决问题的能力. 项目名 ...
热门 | Google Brain前员工深度盘点2017人工智能和深度学习各大动态
翻译 | AI科技大本营参与 | shawn 编辑 | Donna 2017年是人工智能井喷的一年.Google Brain团队前成员Denny Britz在自己的博客WILDML上对过去一年人工智 ...
AlphaGo、人工智能、深度学习解读以及应用
经过几天的比拼,AlphaGo最终还是胜出,创造了人机大战历史上的一个新的里程碑.几乎所有的人都在谈论这件事情,这使得把"人工智能"."深度学习"的热潮推向了新 ...
H2O机器学习：一种强大的可扩展的人工智能和深度学习技术
书名:基于H2O的机器学习实用方法:一种强大的可扩展的人工智能和深度学习技术原书名:Practical Machine Learning with H2O:Powerful, Scalable Te ...
港科夜闻｜香港科大陈启峰教授:人工智能与深度学习交织-从竞赛开始,到无尽钻研...
关注并星标每周阅读港科夜闻建立新视野开启新思维 1.香港科大陈启峰教授:"人工智能与深度学习交织 - 从竞赛开始,到无尽钻研".人工智能与深度学习交织,未来的科技似乎将出现更 ...
【人工智能】深度学习、数据库选择和人工智能的革命；人工智能是解锁IoT潜力的钥匙
深度学习(DL)和人工智能(AI)已经不再是科幻小说中遥不可及的目标,目前已成为了互联网和大数据等领域的前沿研究内容. 由于云计算提供强的计算能力.提出的先进算法以及充裕的资金,这创造了五年前难以想象 ...
人工智能、深度学习和AIoT
1 引言如果从人类最初的幻想开始算起,人工智能的历史非常久远,也许能与人类文明比肩.而现代化的人工智能历史是从1956年达特茅斯会议开始的.在这之后,人工智能的研究几经起落,符号主义.联结主义.专家 ...
人工智能和深度学习发展趋势_AI在学习和发展中的作用
人工智能和深度学习发展趋势 In this series of blogs, AI in HR, we already understood what is AI, what is HR and be ...
深度学习(一): 人工智能-机器学习-深度学习的区别
人工智能-机器学习-深度学习他们之间是有区别的先来一张图做一下解释从发展历史上来看 AI:让机器展现出人类智力回到1956年夏天,在当时的会议上,AI先驱的梦想是建造一台复杂的机器(让当时刚出 ...

【人工智能项目】深度学习实现胸腔X光肺炎检测

【人工智能项目】深度学习实现胸腔X光肺炎检测

导包

EDA探索数据

读取数据

模型

预测

小结

【人工智能项目】深度学习实现胸腔X光肺炎检测相关推荐

最新文章

热门文章