U-Net

最新推荐文章于 2022-05-12 19:58:12 发布

Seasea77

最新推荐文章于 2022-05-12 19:58:12 发布

阅读量254

点赞数

分类专栏：深度学习网络文章标签： U-net opencv keras

本文链接：https://blog.csdn.net/weixin_41874898/article/details/100107593

版权

深度学习网络专栏收录该内容

3 篇文章 0 订阅

订阅专栏

U-Net

1 安装libtiff

问题：pip install libtiff没有办法直接安装
解决：手动安装，去https://www.lfd.uci.edu/~gohlke/pythonlibs/#pygame下载libtiff，转到对应下载的文件夹目录 pip install libtiff-0.4.2-cp36-cp36m-win_amd64.whl

2 rindex
返回子字符串 str 在字符串中最后出现的位置，如果没有匹配的字符串会报异常。

str1 = "this a is string example....wow!!!";
str2 = "is"

print(str1.rindex(str2))  # 7

str1 = "this is string example....wow!!!";
str2 = "is"

print(str1.rindex(str2))  # 5

imgname = "../deform/train\\0.tif"
print(imgname.rindex("/"))
midname = imgname[imgname.rindex("/") + 7:]  # 0.tif
print(midname)

1

（1）训练ssd目标检测模型，实现舌头检测
（2）调用模型，返回舌头坐标值，裁剪并保存舌头图片，并统一大小。
（3）labelme
批量处理json文件：
https://blog.csdn.net/yql_617540298/article/details/81110685
（注意：“三、执行与查看”，直接就可以在json所在的文件夹执行：labelme_json_to_dataset.exe .
或者
labelme_json_to_dataset .）

from keras.preprocessing.image import ImageDataGenerator, array_to_img, img_to_array, load_img
import numpy as np
import os
import glob
import cv2
from libtiff import TIFF


class myAugmentation(object):
    """
    一个用于图像增强的类：
    首先：分别读取训练的图片和标签，然后将图片和标签合并用于下一个阶段使用
    然后：使用Keras的预处理来增强图像
    最后：将增强后的图片分解开，分为训练图片和训练标签
    """

    def __init__(self, train_path="train", label_path="label", merge_path="merge", aug_merge_path="aug_merge",
                 aug_train_path="aug_train", aug_label_path="aug_label", img_type="tif"):
        """
        使用glob从路径中得到所有的“.img_type”文件，初始化类：__init__()
        """
        self.train_imgs = glob.glob(train_path + "/*." + img_type)
        self.label_imgs = glob.glob(label_path + "/*." + img_type)
        self.train_path = train_path
        self.label_path = label_path
        self.merge_path = merge_path
        self.img_type = img_type
        self.aug_merge_path = aug_merge_path
        self.aug_train_path = aug_train_path
        self.aug_label_path = aug_label_path
        self.slices = len(self.train_imgs)
        self.datagen = ImageDataGenerator(
            rotation_range=0.2,
            width_shift_range=0.05,
            height_shift_range=0.05,
            shear_range=0.05,
            zoom_range=0.05,
            horizontal_flip=True,
            fill_mode='nearest')

    def Augmentation(self):

        """
        Start augmentation.....
        """
        trains = self.train_imgs
        labels = self.label_imgs
        path_train = self.train_path
        path_label = self.label_path
        path_merge = self.merge_path
        imgtype = self.img_type
        path_aug_merge = self.aug_merge_path
        if len(trains) != len(labels) or len(trains) == 0 or len(trains) == 0:
            print("trains can't match labels")
            return 0
        for i in range(len(trains)):
            img_t = load_img(path_train + "/" + str(i) + "." + imgtype)
            img_l = load_img(path_label + "/" + str(i) + "." + imgtype)
            x_t = img_to_array(img_t)
            x_l = img_to_array(img_l)
            x_t[:, :, 2] = x_l[:, :, 0]
            img_tmp = array_to_img(x_t)
            img_tmp.save(path_merge + "/" + str(i) + "." + imgtype)
            img = x_t
            img = img.reshape((1,) + img.shape)
            savedir = path_aug_merge + "/" + str(i)
            if not os.path.lexists(savedir):
                os.mkdir(savedir)
            self.doAugmentate(img, savedir, str(i))

    def doAugmentate(self, img, save_to_dir, save_prefix, batch_size=1, save_format='tif', imgnum=30):
        # 增强一张图片的方法
        """
        augmentate one image
        """
        datagen = self.datagen
        i = 0
        for batch in datagen.flow(img,
                                  batch_size=batch_size,
                                  save_to_dir=save_to_dir,
                                  save_prefix=save_prefix,
                                  save_format=save_format):
            i += 1
            if i > imgnum:
                break

    def splitMerge(self):
        # 将合在一起的图片分开
        """
        split merged image apart
        """
        path_merge = self.aug_merge_path
        path_train = self.aug_train_path
        path_label = self.aug_label_path

        for i in range(self.slices):
            path = path_merge + "/" + str(i)
            train_imgs = glob.glob(path + "/*." + self.img_type)
            savedir = path_train + "/" + str(i)
            if not os.path.lexists(savedir):
                os.mkdir(savedir)
            savedir = path_label + "/" + str(i)

            if not os.path.lexists(savedir):
                os.mkdir(savedir)
            for imgname in train_imgs:
                midname = imgname[imgname.rindex("/") + 1:imgname.rindex("." + self.img_type)]
                img = cv2.imread(imgname)
                img_train = img[:, :, 2]  # cv2 read image rgb->bgr
                img_label = img[:, :, 0]
                cv2.imwrite(path_train + "/" + str(i) + "/" + midname + "_train" + "." + self.img_type, img_train)
                cv2.imwrite(path_label + "/" + str(i) + "/" + midname + "_label" + "." + self.img_type, img_label)

    def splitTransform(self):
        # 拆分透视变换后的图像
        """
        split perspective transform images
        """
        # path_merge = "transform"
        # path_train = "transform/data/"
        # path_label = "transform/label/"

        path_merge = "deform/deform_norm2"
        path_train = "deform/train/"
        path_label = "deform/label/"

        train_imgs = glob.glob(path_merge + "/*." + self.img_type)
        for imgname in train_imgs:
            midname = imgname[imgname.rindex("/") + 1:imgname.rindex("." + self.img_type)]
            img = cv2.imread(imgname)
            img_train = img[:, :, 2]  # cv2 read image rgb->bgr
            img_label = img[:, :, 0]
            cv2.imwrite(path_train + midname + "." + self.img_type, img_train)
            cv2.imwrite(path_label + midname + "." + self.img_type, img_label)


class dataProcess(object):
    def __init__(self, out_rows, out_cols, data_path="../deform/train", label_path="../deform/label",
                 test_path="../test", npy_path="../npydata", img_type="tif"):
        # 数据处理类，初始化
        self.out_rows = out_rows
        self.out_cols = out_cols
        self.data_path = data_path
        self.label_path = label_path
        self.img_type = img_type
        self.test_path = test_path
        self.npy_path = npy_path

    # 创建训练数据
    def create_train_data(self):
        i = 0
        print('-' * 30)
        print('Creating training images...')
        print('-' * 30)
        print(self.data_path + "/1." + self.img_type)
        imgs = glob.glob(self.data_path + "/*." + self.img_type)
        print(len(imgs))

        imgdatas = np.ndarray((len(imgs), self.out_rows, self.out_cols, 1), dtype=np.uint8)
        imglabels = np.ndarray((len(imgs), self.out_rows, self.out_cols, 1), dtype=np.uint8)
        # 类似于这样 b = np.zeros((2, 3, 3, 1), dtype=np.uint8)
        for imgname in imgs:
            midname = imgname[imgname.rindex("/") + 7:]
            # img = load_img(self.data_path + "/" + midname, grayscale=True)
            # label = load_img(self.label_path + "/" + midname, grayscale=True)
            # img = img_to_array(img)
            # label = img_to_array(label)
            img = cv2.imread(self.data_path + "/" + midname, cv2.IMREAD_GRAYSCALE)
            label = cv2.imread(self.label_path + "/" + midname, cv2.IMREAD_GRAYSCALE)
            img = cv2.resize(img, (256, 256))
            label = cv2.resize(label, (256, 256))
            img = img.reshape(256, 256, 1)
            label = label.reshape(256, 256, 1)
            # cv2.imshow("img", img)
            # cv2.imshow("label", label)
            # cv2.waitKey(0)
            imgdatas[i] = img
            imglabels[i] = label
            if i % 100 == 0:
                print('Done: {0}/{1} images'.format(i, len(imgs)))
            i += 1
        print('loading done')
        np.save(self.npy_path + '/imgs_train.npy', imgdatas)
        np.save(self.npy_path + '/imgs_mask_train.npy', imglabels)
        print('Saving to .npy files done.')

    # 创建测试数据
    def create_test_data(self):
        i = 0
        print('-' * 30)
        print('Creating test images...')
        print('-' * 30)
        imgs = glob.glob(self.test_path + "/*." + self.img_type)
        print(len(imgs))
        imgdatas = np.ndarray((len(imgs), self.out_rows, self.out_cols, 1), dtype=np.uint8)
        for imgname in imgs:
            midname = imgname[imgname.rindex("/") + 6:]
            img = cv2.imread(self.test_path + "/" + midname, cv2.IMREAD_GRAYSCALE)
            img = cv2.resize(img, (256, 256))
            img = img.reshape(256, 256, 1)
            # cv2.imshow("img", img)
            # cv2.waitKey(0)
            imgdatas[i] = img
            i += 1
        print('loading done')
        np.save(self.npy_path + '/imgs_test.npy', imgdatas)
        print('Saving to imgs_test.npy files done.')

    # 加载训练图片与mask
    def load_train_data(self):
        print('-' * 30)
        print('load train images...')
        print('-' * 30)
        imgs_train = np.load(self.npy_path + "/imgs_train.npy")
        imgs_mask_train = np.load(self.npy_path + "/imgs_mask_train.npy")
        # print(imgs_train.shape)
        # cv2.imshow("imgs_train", imgs_train[0])
        # cv2.waitKey(0)
        imgs_train = imgs_train.astype('float32')
        imgs_mask_train = imgs_mask_train.astype('float32')
        imgs_train /= 255
        print(imgs_train.shape)
        mean = imgs_train.mean(axis=0)
        print(mean.shape)
        imgs_train -= mean
        imgs_mask_train /= 255
        # 做一个阈值处理，输出的概率值大于0.5的就认为是对象，否则认为是背景
        imgs_mask_train[imgs_mask_train > 0.5] = 1
        imgs_mask_train[imgs_mask_train <= 0.5] = 0
        print(imgs_train.dtype)
        print(imgs_train.shape)
        print(imgs_mask_train.dtype)
        print(imgs_mask_train.shape)

        return imgs_train, imgs_mask_train

    # 加载测试图片
    def load_test_data(self):
        print('-' * 30)
        print('load test images...')
        print('-' * 30)
        imgs_test = np.load(self.npy_path + "/imgs_test.npy")
        imgs_test = imgs_test.astype('float32')
        imgs_test /= 255
        mean = imgs_test.mean(axis=0)
        imgs_test -= mean
        return imgs_test


if __name__ == "__main__":
    # 以下注释掉的部分为数据增强代码，通过他们可以将数据进行增强

    # aug = myAugmentation()
    # aug.Augmentation()
    # aug.splitMerge()
    # aug.splitTransform()

    mydata = dataProcess(256, 256)
    mydata.create_train_data()
    mydata.create_test_data()

    imgs_train, imgs_mask_train = mydata.load_train_data()
    print(imgs_train.shape, imgs_mask_train.shape)

import os
# os.environ["CUDA_VISIBLE_DEVICES"] = "0"
import numpy as np
from keras.models import *
from keras.layers import *
from keras.optimizers import *
from keras.callbacks import ModelCheckpoint, LearningRateScheduler
from keras.utils.vis_utils import plot_model
from keras import backend as keras
from keras.preprocessing.image import array_to_img
from data import *


class myUnet(object):
    def __init__(self, img_rows=256, img_cols=256):
        self.img_rows = img_rows
        self.img_cols = img_cols

    # 参数初始化定义
    def load_data(self):
        mydata = dataProcess(self.img_rows, self.img_cols)
        imgs_train, imgs_mask_train = mydata.load_train_data()
        imgs_test = mydata.load_test_data()
        return imgs_train, imgs_mask_train, imgs_test

    # 载入数据
    def get_unet(self):
        inputs = Input((self.img_rows, self.img_cols, 1))
        conv1 = Conv2D(64, 3, activation='relu', padding='same', kernel_initializer='he_normal')(inputs)
        print("conv1 shape:", conv1.shape)
        conv1 = Conv2D(64, 3, activation='relu', padding='same', kernel_initializer='he_normal')(conv1)
        print("conv1 shape:", conv1.shape)
        pool1 = MaxPooling2D(pool_size=(2, 2))(conv1)
        print("pool1 shape:", pool1.shape)

        conv2 = Conv2D(128, 3, activation='relu', padding='same', kernel_initializer='he_normal')(pool1)
        print("conv2 shape:", conv2.shape)
        conv2 = Conv2D(128, 3, activation='relu', padding='same', kernel_initializer='he_normal')(conv2)
        print("conv2 shape:", conv2.shape)
        pool2 = MaxPooling2D(pool_size=(2, 2))(conv2)
        print("pool2 shape:", pool2.shape)

        conv3 = Conv2D(256, 3, activation='relu', padding='same', kernel_initializer='he_normal')(pool2)
        print("conv3 shape:", conv3.shape)
        conv3 = Conv2D(256, 3, activation='relu', padding='same', kernel_initializer='he_normal')(conv3)
        print("conv3 shape:", conv3.shape)
        pool3 = MaxPooling2D(pool_size=(2, 2))(conv3)
        print("pool3 shape:", pool3.shape)

        conv4 = Conv2D(512, 3, activation='relu', padding='same', kernel_initializer='he_normal')(pool3)
        conv4 = Conv2D(512, 3, activation='relu', padding='same', kernel_initializer='he_normal')(conv4)
        drop4 = Dropout(0.5)(conv4)
        pool4 = MaxPooling2D(pool_size=(2, 2))(drop4)

        conv5 = Conv2D(1024, 3, activation='relu', padding='same', kernel_initializer='he_normal')(pool4)
        conv5 = Conv2D(1024, 3, activation='relu', padding='same', kernel_initializer='he_normal')(conv5)
        drop5 = Dropout(0.5)(conv5)

        up6 = Conv2D(512, 2, activation='relu', padding='same', kernel_initializer='he_normal')(
            UpSampling2D(size=(2, 2))(drop5))
        # merge6 = merge([drop4, up6], mode='concat', concat_axis=3)
        merge6 = concatenate([drop4, up6], axis=3)  # 使用concatenate代替merge
        conv6 = Conv2D(512, 3, activation='relu', padding='same', kernel_initializer='he_normal')(merge6)
        conv6 = Conv2D(512, 3, activation='relu', padding='same', kernel_initializer='he_normal')(conv6)

        up7 = Conv2D(256, 2, activation='relu', padding='same', kernel_initializer='he_normal')(
            UpSampling2D(size=(2, 2))(conv6))
        merge7 = concatenate([conv3, up7], axis=3)
        conv7 = Conv2D(256, 3, activation='relu', padding='same', kernel_initializer='he_normal')(merge7)
        conv7 = Conv2D(256, 3, activation='relu', padding='same', kernel_initializer='he_normal')(conv7)

        up8 = Conv2D(128, 2, activation='relu', padding='same', kernel_initializer='he_normal')(
            UpSampling2D(size=(2, 2))(conv7))
        merge8 = concatenate([conv2, up8], axis=3)
        conv8 = Conv2D(128, 3, activation='relu', padding='same', kernel_initializer='he_normal')(merge8)
        conv8 = Conv2D(128, 3, activation='relu', padding='same', kernel_initializer='he_normal')(conv8)

        up9 = Conv2D(64, 2, activation='relu', padding='same', kernel_initializer='he_normal')(
            UpSampling2D(size=(2, 2))(conv8))
        merge9 = concatenate([conv1, up9], axis=3)
        conv9 = Conv2D(64, 3, activation='relu', padding='same', kernel_initializer='he_normal')(merge9)
        conv9 = Conv2D(64, 3, activation='relu', padding='same', kernel_initializer='he_normal')(conv9)
        conv9 = Conv2D(2, 3, activation='relu', padding='same', kernel_initializer='he_normal')(conv9)

        conv10 = Conv2D(1, 1, activation='sigmoid')(conv9)

        model = Model(input=inputs, output=conv10)
        plot_model(model=model, to_file="./Resnet_model.png", show_shapes=True)
        model.compile(optimizer=Adam(lr=1e-4), loss='binary_crossentropy', metrics=['accuracy'])
        return model

    # 如果需要修改输入的格式，那么可以从以下开始修改，上面的结构部分不需要修改
    def train(self):
        print("loading data")
        imgs_train, imgs_mask_train, imgs_test = self.load_data()  # 返回数据为float32类型
        print("loading data done")
        model = self.get_unet()
        print("got unet")
        model_checkpoint = ModelCheckpoint('my_unet.hdf5', monitor='loss', verbose=1, save_best_only=True)
        print('Fitting model...')
        # model.fit(imgs_train, imgs_mask_train, batch_size=2, nb_epoch=10, verbose=1, validation_split=0.2, shuffle=True,
        #           callbacks=[model_checkpoint])
        print('predict test data')
        model = load_model('my_unet.hdf5')
        imgs_mask_test = model.predict(imgs_test, batch_size=1, verbose=1)
        np.save('results/imgs_mask_test.npy', imgs_mask_test)

    def save_img(self):
        print("array to image")
        imgs = np.load('results/imgs_mask_test.npy')
        for i in range(imgs.shape[0]):
            img = imgs[i]
            img = array_to_img(img)
            img.save("results/%d.jpg" % (i))


if __name__ == '__main__':
    myunet = myUnet()
    myunet.train()
    myunet.save_img()

from keras.models import *
from keras.layers import *
from keras.optimizers import *
from keras.utils.vis_utils import plot_model
from keras.callbacks import ModelCheckpoint, LearningRateScheduler
from keras import backend as keras


# 基于FCN的U-net，参考网址：https://blog.csdn.net/g11d111/article/details/78068413
def tong_model():
    input_size = (256, 256, 3)
    inputs = Input(input_size)
    conv1 = Conv2D(64, 3, activation='relu', padding = 'same', kernel_initializer = 'he_normal')(inputs)
    conv1 = Conv2D(64, 3, activation='relu', padding = 'same', kernel_initializer = 'he_normal')(conv1)
    pool1 = MaxPooling2D(pool_size=(2, 2))(conv1)
    conv2 = Conv2D(128, 3, activation='relu', padding = 'same', kernel_initializer = 'he_normal')(pool1)
    conv2 = Conv2D(128, 3, activation='relu', padding = 'same', kernel_initializer = 'he_normal')(conv2)
    pool2 = MaxPooling2D(pool_size=(2, 2))(conv2)
    conv3 = Conv2D(256, 3, activation='relu', padding = 'same', kernel_initializer = 'he_normal')(pool2)
    conv3 = Conv2D(256, 3, activation='relu', padding = 'same', kernel_initializer = 'he_normal')(conv3)
    pool3 = MaxPooling2D(pool_size=(2, 2))(conv3)
    conv4 = Conv2D(512, 3, activation='relu', padding = 'same', kernel_initializer = 'he_normal')(pool3)
    conv4 = Conv2D(512, 3, activation='relu', padding = 'same', kernel_initializer = 'he_normal')(conv4)

    drop4 = conv4

    pool4 = MaxPooling2D(pool_size=(2, 2))(drop4)
    conv5 = Conv2D(1024, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(pool4)
    conv5 = Conv2D(1024, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv5)
    drop5 = conv5
    up6 = Conv2D(512, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(drop5))
    merge6 = concatenate([drop4, up6], axis = 3)
    conv6 = Conv2D(512, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(merge6)
    conv6 = Conv2D(512, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv6)

    up7 = Conv2D(256, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv6))
    merge7 = concatenate([conv3, up7], axis = 3)
    conv7 = Conv2D(256, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(merge7)
    conv7 = Conv2D(256, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv7)

    up8 = Conv2D(128, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv7))
    merge8 = concatenate([conv2, up8], axis = 3)
    conv8 = Conv2D(128, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(merge8)
    conv8 = Conv2D(128, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv8)

    up9 = Conv2D(64, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv8))
    merge9 = concatenate([conv1, up9], axis = 3)
    conv9 = Conv2D(64, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(merge9)
    conv9 = Conv2D(64, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv9)
    conv9 = Conv2D(2, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv9)
    conv10 = Conv2D(filters=3, kernel_size=(1,1), strides=(1,1), padding='same', activation = 'sigmoid')(conv9)
    model = Model(inputs = inputs, outputs = conv10)
    model.compile(optimizer = Adam(lr = 1e-4), loss = 'binary_crossentropy', metrics = ['accuracy'])
    plot_model(model=model, to_file="./Resnet_model.png", show_shapes=True)
    return model


if __name__ == "__main__":
    tong_model()