def getPredictValueByImg(img):
img_w = img.shape[0]
img_h = img.shape[1]
resize_img = cv2.resize(img, (512, 512))
test_images = resize_img.astype(np.float)
# convert from [0:255] => [0.0:1.0]
test_images = np.multiply(test_images, 1.0 / 255.0)
unet2d = unet2dModule(512, 512, 3)
predictvalue = unet2d.prediction("./model/unet2dglandceil.pd", test_images)
return cv2.resize(predictvalue, (img_w, img_h))
def predict(filePath, targetFileName):
true_img = cv2.imread(filePath, cv2.IMREAD_COLOR)
test_images = true_img.astype(np.float)
# convert from [0:255] => [0.0:1.0]
test_images = np.multiply(test_images, 1.0 / 255.0)
unet2d = unet2dModule(512, 512, 3)
predictvalue = unet2d.prediction("./model/unet2dglandceil.pd", test_images)
cv2.imwrite(
"../mouth_picture/data/test/stardand/mask/mouth/" + targetFileName,
predictvalue)
def predict():
true_img = cv2.imread("D:\Data\GlandCeil\Test\Image\\testA_31.bmp",
cv2.IMREAD_COLOR)
test_images = true_img.astype(np.float)
# convert from [0:255] => [0.0:1.0]
test_images = np.multiply(test_images, 1.0 / 255.0)
unet2d = unet2dModule(512, 512, 3)
predictvalue = unet2d.prediction(
"D:\Project\python\GlandCeil_Unet\model\\unet2dglandceil.pd",
test_images)
cv2.imwrite("mask1.bmp", predictvalue)
def train():
'''
Preprocessing for dataset
'''
# Read data set (Train data from CSV file)
csvmaskdata = pd.read_csv('./GlandsMask.csv')
csvimagedata = pd.read_csv('./GlandsImage.csv')
maskdata = csvmaskdata.iloc[:, :].values
imagedata = csvimagedata.iloc[:, :].values
# shuffle imagedata and maskdata together
perm = np.arange(len(csvimagedata))
np.random.shuffle(perm)
imagedata = imagedata[perm]
maskdata = maskdata[perm]
unet2d = unet2dModule(512, 512, channels=3, costname="dice coefficient")
unet2d.train(imagedata, maskdata, "./model/unet2dglandceil.pd", "./log",
0.0005, 0.8, 1000, 2)
def train():
# Read data set (Train data from CSV file)
csvmaskdata = pd.read_csv('./GlandsMask.csv')
csvimagedata = pd.read_csv('./GlandsImage.csv')
# imagedata:二维列表
maskdata = csvmaskdata.iloc[:, :].values
imagedata = csvimagedata.iloc[:, :].values
# 把训练集打乱顺序
perm = np.arange(len(csvimagedata))
np.random.shuffle(perm)
imagedata = imagedata[perm]
maskdata = maskdata[perm]
# 图片大小为 720 * 720
unet2d = unet2dModule(720, 720, channels=3, costname="dice coefficient")
# 保存名称："./model/unet2dglandceil605.pd"
# 日志："./log"
# 学习率：0.0005
# drop_out：1
# 次数：40000
# batch_size： 2
unet2d.train(imagedata, maskdata, "./model/unet2dglandceil605.pd",
"./log", 0.0005, 1, 40000, 2)
def predict():
fromImageDir = "./JPEGImages/"
# 输出文件的保存位置
toImageDir = "./JR-604/"
unet2d = unet2dModule(720, 720, 3)
filelist = os.listdir(fromImageDir) # 该文件夹下所有的文件（包括文件夹
for file in filelist: # 遍历所有文件
if file < '000000':
continue
else:
Olddir = os.path.join(fromImageDir, file) # 原来的文件路径
if os.path.isdir(Olddir): # 如果是文件夹则跳过
continue
true_img = cv2.imread(Olddir, cv2.IMREAD_COLOR)
test_images = true_img.astype(np.float)
test_images = np.multiply(test_images, 1.0 / 255.0)
# predictvalue = unet2d.prediction("C:\\Users\\admin\\Desktop\\GlandCeil_Unet\\model\\unet2dglandceil27.pd",
# test_images)
predictvalue = unet2d.prediction(".\\model\\unet2dglandceil604.pd",
test_images)
cv2.imwrite(toImageDir + '/' + file, predictvalue)