diff --git a/LogicImplement/README.md b/LogicImplement/README.md
deleted file mode 100644
index ea7b369..0000000
--- a/LogicImplement/README.md
+++ /dev/null
@@ -1,2 +0,0 @@
-# Here is the code
-
diff --git a/LogicImplement/data_process/augmentation_handle.py b/LogicImplement/data_process/augmentation_handle.py
deleted file mode 100644
index 7508b54..0000000
--- a/LogicImplement/data_process/augmentation_handle.py
+++ /dev/null
@@ -1,129 +0,0 @@
-# -*- coding: utf-8 -*-
-
-import random
-
-import cv2
-import numpy as np
-import os.path
-import copy
-
-
-# 椒盐噪声
-def SaltAndPepper(src, percetage=0.2):
-    SP_NoiseImg = src.copy()
-    SP_NoiseNum = int(percetage * src.shape[0] * src.shape[1])
-    for i in range(SP_NoiseNum):
-        randR = np.random.randint(0, src.shape[0] - 1)
-        randG = np.random.randint(0, src.shape[1] - 1)
-        randB = np.random.randint(0, 3)
-        if np.random.randint(0, 1) == 0:
-            SP_NoiseImg[randR, randG, randB] = 0
-        else:
-            SP_NoiseImg[randR, randG, randB] = 255
-    return SP_NoiseImg
-
-
-# 高斯噪声
-def addGaussianNoise(image, percetage=0.2):
-    G_Noiseimg = image.copy()
-    w = image.shape[1]
-    h = image.shape[0]
-    G_NoiseNum = int(percetage * image.shape[0] * image.shape[1])
-    for i in range(G_NoiseNum):
-        temp_x = np.random.randint(0, h)
-        temp_y = np.random.randint(0, w)
-        G_Noiseimg[temp_x][temp_y][np.random.randint(3)] = np.random.randn(1)[0]
-    return G_Noiseimg
-
-
-# 昏暗
-def darker(image, percetage=0.9):
-    image_copy = image.copy()
-    w = image.shape[1]
-    h = image.shape[0]
-    # get darker
-    for xi in range(0, w):
-        for xj in range(0, h):
-            image_copy[xj, xi, 0] = int(image[xj, xi, 0] * percetage)
-            image_copy[xj, xi, 1] = int(image[xj, xi, 1] * percetage)
-            image_copy[xj, xi, 2] = int(image[xj, xi, 2] * percetage)
-    return image_copy
-
-
-# 亮度
-def brighter(image, percetage=1.5):
-    image_copy = image.copy()
-    w = image.shape[1]
-    h = image.shape[0]
-    # get brighter
-    for xi in range(0, w):
-        for xj in range(0, h):
-            image_copy[xj, xi, 0] = np.clip(int(image[xj, xi, 0] * percetage), a_max=255, a_min=0)
-            image_copy[xj, xi, 1] = np.clip(int(image[xj, xi, 1] * percetage), a_max=255, a_min=0)
-            image_copy[xj, xi, 2] = np.clip(int(image[xj, xi, 2] * percetage), a_max=255, a_min=0)
-    return image_copy
-
-
-# 旋转
-def rotate(image, angle, center=None, scale=1.0):
-    (h, w) = image.shape[:2]
-    # If no rotation center is specified, the center of the image is set as the rotation center
-    if center is None:
-        center = (w / 2, h / 2)
-    m = cv2.getRotationMatrix2D(center, angle, scale)
-    rotated = cv2.warpAffine(image, m, (w, h))
-    return rotated
-
-
-# 翻转
-def flip(image):
-    flipped_image = np.fliplr(image)
-    return flipped_image
-
-
-# 图片文件夹路径
-image_dir = r'C:\data\TianChiCTSeg\data\train\image\\'
-mask_dir = r'C:\data\TianChiCTSeg\data\train\mask\\'
-i = 0
-j = 0
-for img_name in os.listdir(image_dir):
-    num = random.randint(0, 10)
-    # print("num:", num)
-    i += 1
-    if num == 2 or num == 3:
-        j += 1
-        print("aug")
-        img_path = image_dir + img_name
-        mask_path = mask_dir + img_name
-        print(img_path)
-        img = cv2.imread(img_path)
-        mask = cv2.imread(mask_path)
-
-        rotated_90_image = rotate(img, 90)
-        rotated_90_mask = rotate(mask, 90)
-        cv2.imwrite(image_dir + img_name[0:-4] + '_r90.png', rotated_90_image)
-        cv2.imwrite(mask_dir.replace("image", "mask") + img_name[0:-4] + '_r90.png', rotated_90_mask)
-
-        brighter_image = brighter(img)
-        # brighter_mask = brighter(mask)
-        cv2.imwrite(image_dir + img_name[0:-4] + '_brighter.png', brighter_image)
-        cv2.imwrite(mask_dir.replace("image", "mask") + img_name[0:-4] + '_brighter.png', mask)
-
-        darker_image = darker(img)
-        # darker_mask = darker(mask)
-        cv2.imwrite(image_dir + img_name[0:-4] + '_darker.png', darker_image)
-        cv2.imwrite(mask_dir.replace("image", "mask") + img_name[0:-4] + '_darker.png', mask)
-
-        addGaussianNoise_image = addGaussianNoise(img, 90)
-        addGaussianNoise_mask = addGaussianNoise(mask, 90)
-        cv2.imwrite(image_dir + img_name[0:-4] + '_addGaussianNoise.png', addGaussianNoise_image)
-        cv2.imwrite(mask_dir.replace("image", "mask") + img_name[0:-4] + '_addGaussianNoise.png', addGaussianNoise_mask)
-        #
-        SaltAndPepper_image = SaltAndPepper(img, 90)
-        SaltAndPepper_mask = SaltAndPepper(mask, 90)
-        cv2.imwrite(image_dir + img_name[0:-4] + '_SaltAndPepper.png', SaltAndPepper_image)
-        cv2.imwrite(mask_dir.replace("image", "mask") + img_name[0:-4] + '_SaltAndPepper.png', SaltAndPepper_mask)
-    else:
-        print("不aug")
-    print(f"进行到{i}张")
-    print(f"已增强{j}张")
diff --git a/LogicImplement/data_process/data_split.py b/LogicImplement/data_process/data_split.py
deleted file mode 100644
index 98ef3f8..0000000
--- a/LogicImplement/data_process/data_split.py
+++ /dev/null
@@ -1,84 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*-
-import os
-import random
-
-import numpy as np
-from scipy import io
-import cv2
-import matplotlib.pyplot as plt
-import h5py
-import glob
-import pydicom
-
-
-def to255(img):     # 归一化到0-255代码
-    min_val = img.min()
-    max_val = img.max()
-    img = (img - min_val) / (max_val - min_val + 1e-5)  # 图像归一化
-    img = img * 255  # *255
-    return img
-
-src_path = r'C:\data\TianChiCTSeg\GroundTruth'  # mat文件的上级文件夹路径，英文路径
-save_path = r'C:\data\TianChiCTSeg\data' # 保存png路径
-
-train_image_path = save_path + '\\' + 'train\\' + 'image'
-test_image_path = save_path + '\\' + 'val\\' + 'image'
-train_mask_path = save_path + '\\' + 'train\\' + 'mask'
-test_mask_path = save_path + '\\' + 'val\\' + 'mask'
-
-os.makedirs(train_image_path, exist_ok=True)
-os.makedirs(test_image_path, exist_ok=True)
-os.makedirs(train_mask_path, exist_ok=True)
-os.makedirs(test_mask_path, exist_ok=True)
-
-names = glob.glob(src_path + '\\' + '*.mat')
-random.shuffle(names)
-order = 0
-for name in names:
-    # feature=h5py.File(name)               #读取mat文件
-    feature = io.loadmat(name)  # (512, 512, 183), ['__header__', '__version__', '__globals__', 'Mask']
-    # print(feature.keys())
-    # 获取病人的名字
-    this_pat_name = name.split('\\')[-1].split('.')[0]
-    # 寻找病人对应的image的路径
-    pat_image_root_path = r'C:\data\TianChiCTSeg\CT_scans 01' + '\\' + this_pat_name
-    # 首先获取mask
-    this_3D_mask = feature['Mask']
-    # 下面进行到255的转换
-    this_3D_mask[np.where(this_3D_mask == 1)] = 255
-    # 下面进行slice的切分
-    x, y, z = this_3D_mask.shape
-    if order % 5 == 0:
-        # 验证集
-        for slice_index in range(z):
-            this_2D_mask = this_3D_mask[:, :, slice_index]
-            # 进行保存
-            this_save_name = this_pat_name + '_' + 'D' + str(slice_index+1).zfill(4) + '.png'
-            cv2.imwrite(test_mask_path + '\\' + this_save_name, this_2D_mask)
-            # 读取原始dcm图像
-            ds = pydicom.dcmread(pat_image_root_path + '\\' + 'D' + str(slice_index+1).zfill(4) + '.dcm')
-            data = ds.pixel_array   # 0-4095， 肺部窗宽 1 300 Hu~1 700 Hu,窗位-600 Hu~-800 Hu,
-            ct_image = to255(data)
-            cv2.imwrite(test_image_path + '\\' + this_save_name, ct_image)
-    else:
-        # 训练集
-        for slice_index in range(z):
-            this_2D_mask = this_3D_mask[:, :, slice_index]
-            # 进行保存
-            this_save_name = this_pat_name + '_' + 'D' + str(slice_index+1).zfill(4) + '.png'
-            cv2.imwrite(train_mask_path + '\\' + this_save_name, this_2D_mask)
-            # 读取原始dcm图像
-            ds = pydicom.dcmread(pat_image_root_path + '\\' + 'D' + str(slice_index+1).zfill(4) + '.dcm')
-            data = ds.pixel_array   # 0-4095， 肺部窗宽 1 300 Hu~1 700 Hu,窗位-600 Hu~-800 Hu,
-
-            ct_image = to255(data)
-            cv2.imwrite(train_image_path + '\\' + this_save_name, ct_image)
-    print(order)
-    order += 1
-
-
-
-
-
-
diff --git a/LogicImplement/datasets/__init__.py b/LogicImplement/datasets/__init__.py
deleted file mode 100644
index 67f768c..0000000
--- a/LogicImplement/datasets/__init__.py
+++ /dev/null
@@ -1,2 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
diff --git a/LogicImplement/datasets/__pycache__/__init__.cpython-38.pyc b/LogicImplement/datasets/__pycache__/__init__.cpython-38.pyc
deleted file mode 100644
index 4555732..0000000
Binary files a/LogicImplement/datasets/__pycache__/__init__.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/datasets/__pycache__/lung.cpython-38.pyc b/LogicImplement/datasets/__pycache__/lung.cpython-38.pyc
deleted file mode 100644
index a86d3bf..0000000
Binary files a/LogicImplement/datasets/__pycache__/lung.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/datasets/lung.py b/LogicImplement/datasets/lung.py
deleted file mode 100644
index 48bf9ae..0000000
--- a/LogicImplement/datasets/lung.py
+++ /dev/null
@@ -1,71 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-from torch.utils.data import Dataset
-import torch
-from PIL import Image
-from utils.tools import process_binary_mask_tensor
-# from datasets.make_dataset import make_dataset
-import os
-import numpy as np
-
-
-class Lung(Dataset):
-
-    CHANNELS_NUM = 3
-    NUM_CLASSES = 2
-
-    MEAN = [0.70791537, 0.59156666, 0.54687498]
-    STD = [0.15324752, 0.16178547, 0.17681521]
-
-    def __init__(self, mode, transform=None, target_transform=None, BASE_PATH=""):
-        print(mode)
-        self.items_image, self.items_mask = make_dataset(mode, BASE_PATH)
-        self.transform = transform
-        self.target_transform = target_transform
-
-    def __len__(self):
-        return len(self.items_image)
-
-    def __str__(self):
-        return 'Stroke'
-
-    def __getitem__(self, index):
-        # image_path = self.items[index]['image_path']
-        # mask_path = self.items[index]['mask_path']
-        image_path = self.items_image[index]
-        mask_path = self.items_mask[index]
-
-        image = Image.open(image_path).convert('RGB')
-        mask = Image.open(mask_path).convert('L')
-
-        if self.transform:
-            image = self.transform(image)
-        if self.target_transform:
-            mask = self.target_transform(mask)
-
-        mask = process_binary_mask_tensor(mask)
-        # print(torch.unique(mask))
-
-        return image, mask
-
-
-def make_dataset(mode, base_path):
-    print(mode)
-
-    # assert mode in ['train', 'val']
-    #
-    # path = os.path.join(base_path, mode)
-    image_path = os.path.join(base_path, "image")
-    mask_path = os.path.join(base_path, "mask")
-    # print(image_path)
-    image_list = []
-    for file in os.listdir(image_path):
-        image_list.append(os.path.join(image_path, file))
-
-    mask_list = []
-    for file in os.listdir(mask_path):
-        mask_list.append(os.path.join(mask_path, file))
-
-    # print(image_list)
-    return image_list, mask_list
-
diff --git a/LogicImplement/datasets/make_dataset.py b/LogicImplement/datasets/make_dataset.py
deleted file mode 100644
index 5f1b66f..0000000
--- a/LogicImplement/datasets/make_dataset.py
+++ /dev/null
@@ -1,16 +0,0 @@
-# -*- coding: utf-8 -*-
-import os
-
-
-def make_dataset(mode, base_path, is_contain_augmented_data):
-    """从指定目录下读取用于training、validation的json文件"""
-
-    assert mode in ['train', 'val']
-
-    path = os.path.join(base_path, mode)
-    image_path = os.path.join(path, "image")
-    mask_path = os.path.join(path, "mask")
-
-    image_list = os.listdir(image_path)
-    mask_list = os.listdir(mask_path)
-    return image_list, mask_list
\ No newline at end of file
diff --git a/LogicImplement/generate_metric.py b/LogicImplement/generate_metric.py
deleted file mode 100644
index a957d3f..0000000
--- a/LogicImplement/generate_metric.py
+++ /dev/null
@@ -1,111 +0,0 @@
-# -*- coding: utf-8 -*-
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-from random import random
-
-import torch
-from torch.autograd import Variable
-from models import unet
-from torchvision import transforms
-import os
-from PIL import Image
-import argparse
-import numpy as np
-from collections import OrderedDict
-import cv2
-from utils.metrics import compute_metrics
-
-# 选择网络模型
-# net = xnet.XNet(num_classes=2, in_channels=3)
-net = unet.UNet(num_classes=2, in_channels=3)
-# net = unet.UNet(num_classes=2, in_channels=3, is_attention=True)
-
-# 加载模型
-ckpt = torch.load(r".pkl")
-ckpt = ckpt['model_state_dict']
-new_state_dict = OrderedDict()
-for k, v in ckpt.items():
-    # name = k[7:]  # remove `module.`，表面从第7个key值字符取到最后一个字符，正好去掉了module.
-    new_state_dict[k] = v  # 新字典的key值对应的value为一一对应的值。
-
-net.load_state_dict(new_state_dict)
-net.eval()
-
-pre_data_path = r"...image"
-dst_path = r"...preds"
-os.makedirs(dst_path, exist_ok=True)
-image_list = []
-for file in os.listdir(pre_data_path):
-    image_list.append(os.path.join(pre_data_path, file))
-
-# 验证
-loss_all = []
-predictions_all = []
-labels_all = []
-
-
-with torch.no_grad():
-    i = 0
-    for image in image_list:
-        print(i)
-        i += 1
-        name = image.split("\\")[-1]
-        # print(name)
-        # 读取原图
-        org_image = cv2.imread(image, 0)
-        org_mask = cv2.imread(image.replace('image', 'label'), 0)
-
-        # mask = Image.open(image.replace('image', 'label')).convert('L')
-        labels = np.zeros(org_mask.shape, np.uint8)
-        labels[org_mask != 0] = 1
-
-        image = Image.open(image).convert('RGB')
-        ori_size = image.size
-        # image = image.resize((224, 224), resample=NEAREST)
-        # image = img_to_tensor(image)
-        image = transforms.ToTensor()(image)
-        # print(image.shape)  # [3, 224, 224]
-        image = Variable(torch.unsqueeze(image, dim=0).float(), requires_grad=False)
-        # print(image.shape)  # [1, 3, 224, 224]
-        outputs = net(image)    # # [1, 2, 224, 224], 此2应该为类别数
-
-        if isinstance(outputs, list):
-            # 若使用deep supervision，用最后一个输出来进行预测
-            predictions = torch.max(outputs[-1], dim=1)[1].cpu().numpy().astype(np.int)
-        else:
-            # 将概率最大的类别作为预测的类别
-            predictions = torch.max(outputs, dim=1)[1].cpu().numpy().astype(np.int)
-        labels = labels.astype(np.int)
-
-        predictions_all.append(predictions)
-        labels_all.append(labels)
-        # softmax = torch.nn.Softmax(dim=1)
-        # outputs = softmax(outputs)
-        outputs = outputs.squeeze(0)    # [2, 224, 224]
-
-        # mask = np.argmax(outputs, axis=0)
-        mask = torch.max(outputs, 0)[1].cpu().numpy()
-        # print(mask.max())
-        # print(mask.shape)
-        a = np.zeros(mask.shape, np.float32)
-        a[mask == 1] = 255
-        # a[mask == 0] = 0
-        # for i in range(np.shape(a)[0]):
-        im = Image.fromarray(np.uint8(a[:, :]))
-        # 下面进行拼接展示
-
-        cat_img = np.hstack([org_image, org_mask, a])
-        print(dst_path + '\\' + str(name), cat_img)
-        ##cv2.imwrite(dst_path + '\\' + str(name), cat_img)
-
-
-    # 使用混淆矩阵计算语义分割中的指标
-    iou, miou, dsc, mdsc, ac, pc, mpc, se, mse, sp, msp, f1, mf1 = compute_metrics(predictions_all, labels_all,
-                                                                                   num_classes=2)
-    print('Training: MIoU: {:.4f}, MDSC: {:.4f}, MPC: {:.4f}, AC: {:.4f}, MSE: {:.4f}, MSP: {:.4f}, MF1: {:.4f}'.format(
-        miou, mdsc, mpc, ac, mse, msp, mf1
-    ))
-        # im = im.resize(ori_size, resample=NEAREST)     # 需要使用最近邻插值
-        # im.convert('RGB').save(dst_path + "/" + str(name))
-
-
diff --git a/LogicImplement/generate_predict.py b/LogicImplement/generate_predict.py
deleted file mode 100644
index 812a17f..0000000
--- a/LogicImplement/generate_predict.py
+++ /dev/null
@@ -1,82 +0,0 @@
-# -*- coding: utf-8 -*-
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-from random import random
-
-import torch
-import time
-
-import torchvision
-from PIL.Image import NEAREST
-from albumentations.pytorch.functional import img_to_tensor
-from torch.autograd import Variable
-from torch.utils.data import DataLoader
-from torch.utils.tensorboard import SummaryWriter
-from models import unet, nested_unet, loss_function, xnet, nfn_plus, unet_tiny
-from datasets import lung
-from torchvision import transforms
-import os
-from PIL import Image
-import argparse
-import numpy as np
-from collections import OrderedDict
-import cv2
-
-# 选择网络模型
-
-strat_time = time.time()
-# net = unet.UNet(num_classes=2, in_channels=3)
-save_name = ""
-net = unet.UNet(num_classes=2, in_channels=3)
-# 加载模型
-ckpt = torch.load(r".pkl")
-ckpt = ckpt['model_state_dict']
-# new_state_dict = OrderedDict()
-# for k, v in ckpt.items():
-#     name = k[7:]  # remove `module.`，表面从第7个key值字符取到最后一个字符，正好去掉了module.
-#     new_state_dict[name] = v  # 新字典的key值对应的value为一一对应的值。
-
-net.load_state_dict(ckpt)
-net.eval()
-
-pre_data_path = r".../val/image"
-dst_path = r"..." + "/" + save_name
-os.makedirs(dst_path, exist_ok=True)
-
-image_list = []
-for file in os.listdir(pre_data_path):
-    image_list.append(os.path.join(pre_data_path, file))
-
-with torch.no_grad():
-    i = 0
-    for image in image_list:
-        print(i)
-        i += 1
-        name = image.split("/")[-1]
-        gt_mask = cv2.imread(image.replace('image', 'mask'), 0)
-        if len(np.unique(gt_mask)) > 1:
-            image = Image.open(image).convert('RGB')
-            ori_size = image.size
-
-            image = transforms.ToTensor()(image)
-            # image = Variable(torch.unsqueeze(image, dim=0).float(), requires_grad=False)
-            image = torch.unsqueeze(image, dim=0)
-            outputs = net(image)
-            outputs = outputs.squeeze(0)    # [2, 224, 224]
-
-            if outputs.shape[0] == 2:
-                mask = torch.max(outputs, 0)[1].cpu().numpy()
-                a = np.zeros(mask.shape, np.uint8)
-                a[mask == 1] = 255
-                cat = np.hstack([gt_mask, a])
-                print(dst_path + "/" + str(name), cat)
-                #cv2.imwrite(dst_path + "/" + str(name), cat)
-        else:
-            pass
-
-
-end_time = time.time()
-seg_time = end_time - strat_time
-print(save_name + "总耗时为：" + str(seg_time))
-
-
diff --git a/LogicImplement/models/U2net.py b/LogicImplement/models/U2net.py
deleted file mode 100644
index 6440202..0000000
--- a/LogicImplement/models/U2net.py
+++ /dev/null
@@ -1,523 +0,0 @@
-# -*- coding: utf-8 -*-
-
-import torch
-from torch import nn
-
-import torch.nn.functional as F
-
-
-class REBNCONV(nn.Module):
-    def __init__(self, in_ch=3, out_ch=3, dirate=1):
-        super(REBNCONV, self).__init__()
-
-        self.conv_s1 = nn.Conv2d(in_ch, out_ch, 3, padding=1 * dirate, dilation=1 * dirate)
-        self.bn_s1 = nn.BatchNorm2d(out_ch)
-        self.relu_s1 = nn.ReLU(inplace=True)
-
-    def forward(self, x):
-        hx = x
-        xout = self.relu_s1(self.bn_s1(self.conv_s1(hx)))
-
-        return xout
-
-
-## upsample tensor 'src' to have the same spatial size with tensor 'tar'
-def _upsample_like(src, tar):
-    src = F.upsample(src, size=tar.shape[2:], mode='bilinear')
-
-    return src
-
-
-### RSU-7 ###
-class RSU7(nn.Module):  # UNet07DRES(nn.Module):
-
-    def __init__(self, in_ch=3, mid_ch=12, out_ch=3):
-        super(RSU7, self).__init__()
-
-        self.rebnconvin = REBNCONV(in_ch, out_ch, dirate=1)
-
-        self.rebnconv1 = REBNCONV(out_ch, mid_ch, dirate=1)
-        self.pool1 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv2 = REBNCONV(mid_ch, mid_ch, dirate=1)
-        self.pool2 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv3 = REBNCONV(mid_ch, mid_ch, dirate=1)
-        self.pool3 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv4 = REBNCONV(mid_ch, mid_ch, dirate=1)
-        self.pool4 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv5 = REBNCONV(mid_ch, mid_ch, dirate=1)
-        self.pool5 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv6 = REBNCONV(mid_ch, mid_ch, dirate=1)
-
-        self.rebnconv7 = REBNCONV(mid_ch, mid_ch, dirate=2)
-
-        self.rebnconv6d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv5d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv4d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv3d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv2d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv1d = REBNCONV(mid_ch * 2, out_ch, dirate=1)
-
-    def forward(self, x):
-        hx = x
-        hxin = self.rebnconvin(hx)
-
-        hx1 = self.rebnconv1(hxin)
-        hx = self.pool1(hx1)
-
-        hx2 = self.rebnconv2(hx)
-        hx = self.pool2(hx2)
-
-        hx3 = self.rebnconv3(hx)
-        hx = self.pool3(hx3)
-
-        hx4 = self.rebnconv4(hx)
-        hx = self.pool4(hx4)
-
-        hx5 = self.rebnconv5(hx)
-        hx = self.pool5(hx5)
-
-        hx6 = self.rebnconv6(hx)
-
-        hx7 = self.rebnconv7(hx6)
-
-        hx6d = self.rebnconv6d(torch.cat((hx7, hx6), 1))
-        hx6dup = _upsample_like(hx6d, hx5)
-
-        hx5d = self.rebnconv5d(torch.cat((hx6dup, hx5), 1))
-        hx5dup = _upsample_like(hx5d, hx4)
-
-        hx4d = self.rebnconv4d(torch.cat((hx5dup, hx4), 1))
-        hx4dup = _upsample_like(hx4d, hx3)
-
-        hx3d = self.rebnconv3d(torch.cat((hx4dup, hx3), 1))
-        hx3dup = _upsample_like(hx3d, hx2)
-
-        hx2d = self.rebnconv2d(torch.cat((hx3dup, hx2), 1))
-        hx2dup = _upsample_like(hx2d, hx1)
-
-        hx1d = self.rebnconv1d(torch.cat((hx2dup, hx1), 1))
-
-        return hx1d + hxin
-
-
-### RSU-6 ###
-class RSU6(nn.Module):  # UNet06DRES(nn.Module):
-
-    def __init__(self, in_ch=3, mid_ch=12, out_ch=3):
-        super(RSU6, self).__init__()
-
-        self.rebnconvin = REBNCONV(in_ch, out_ch, dirate=1)
-
-        self.rebnconv1 = REBNCONV(out_ch, mid_ch, dirate=1)
-        self.pool1 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv2 = REBNCONV(mid_ch, mid_ch, dirate=1)
-        self.pool2 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv3 = REBNCONV(mid_ch, mid_ch, dirate=1)
-        self.pool3 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv4 = REBNCONV(mid_ch, mid_ch, dirate=1)
-        self.pool4 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv5 = REBNCONV(mid_ch, mid_ch, dirate=1)
-
-        self.rebnconv6 = REBNCONV(mid_ch, mid_ch, dirate=2)
-
-        self.rebnconv5d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv4d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv3d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv2d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv1d = REBNCONV(mid_ch * 2, out_ch, dirate=1)
-
-    def forward(self, x):
-        hx = x
-
-        hxin = self.rebnconvin(hx)
-
-        hx1 = self.rebnconv1(hxin)
-        hx = self.pool1(hx1)
-
-        hx2 = self.rebnconv2(hx)
-        hx = self.pool2(hx2)
-
-        hx3 = self.rebnconv3(hx)
-        hx = self.pool3(hx3)
-
-        hx4 = self.rebnconv4(hx)
-        hx = self.pool4(hx4)
-
-        hx5 = self.rebnconv5(hx)
-
-        hx6 = self.rebnconv6(hx5)
-
-        hx5d = self.rebnconv5d(torch.cat((hx6, hx5), 1))
-        hx5dup = _upsample_like(hx5d, hx4)
-
-        hx4d = self.rebnconv4d(torch.cat((hx5dup, hx4), 1))
-        hx4dup = _upsample_like(hx4d, hx3)
-
-        hx3d = self.rebnconv3d(torch.cat((hx4dup, hx3), 1))
-        hx3dup = _upsample_like(hx3d, hx2)
-
-        hx2d = self.rebnconv2d(torch.cat((hx3dup, hx2), 1))
-        hx2dup = _upsample_like(hx2d, hx1)
-
-        hx1d = self.rebnconv1d(torch.cat((hx2dup, hx1), 1))
-
-        return hx1d + hxin
-
-
-### RSU-5 ###
-class RSU5(nn.Module):  # UNet05DRES(nn.Module):
-
-    def __init__(self, in_ch=3, mid_ch=12, out_ch=3):
-        super(RSU5, self).__init__()
-
-        self.rebnconvin = REBNCONV(in_ch, out_ch, dirate=1)
-
-        self.rebnconv1 = REBNCONV(out_ch, mid_ch, dirate=1)
-        self.pool1 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv2 = REBNCONV(mid_ch, mid_ch, dirate=1)
-        self.pool2 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv3 = REBNCONV(mid_ch, mid_ch, dirate=1)
-        self.pool3 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv4 = REBNCONV(mid_ch, mid_ch, dirate=1)
-
-        self.rebnconv5 = REBNCONV(mid_ch, mid_ch, dirate=2)
-
-        self.rebnconv4d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv3d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv2d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv1d = REBNCONV(mid_ch * 2, out_ch, dirate=1)
-
-    def forward(self, x):
-        hx = x
-
-        hxin = self.rebnconvin(hx)
-
-        hx1 = self.rebnconv1(hxin)
-        hx = self.pool1(hx1)
-
-        hx2 = self.rebnconv2(hx)
-        hx = self.pool2(hx2)
-
-        hx3 = self.rebnconv3(hx)
-        hx = self.pool3(hx3)
-
-        hx4 = self.rebnconv4(hx)
-
-        hx5 = self.rebnconv5(hx4)
-
-        hx4d = self.rebnconv4d(torch.cat((hx5, hx4), 1))
-        hx4dup = _upsample_like(hx4d, hx3)
-
-        hx3d = self.rebnconv3d(torch.cat((hx4dup, hx3), 1))
-        hx3dup = _upsample_like(hx3d, hx2)
-
-        hx2d = self.rebnconv2d(torch.cat((hx3dup, hx2), 1))
-        hx2dup = _upsample_like(hx2d, hx1)
-
-        hx1d = self.rebnconv1d(torch.cat((hx2dup, hx1), 1))
-
-        return hx1d + hxin
-
-
-### RSU-4 ###
-class RSU4(nn.Module):  # UNet04DRES(nn.Module):
-
-    def __init__(self, in_ch=3, mid_ch=12, out_ch=3):
-        super(RSU4, self).__init__()
-
-        self.rebnconvin = REBNCONV(in_ch, out_ch, dirate=1)
-
-        self.rebnconv1 = REBNCONV(out_ch, mid_ch, dirate=1)
-        self.pool1 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv2 = REBNCONV(mid_ch, mid_ch, dirate=1)
-        self.pool2 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.rebnconv3 = REBNCONV(mid_ch, mid_ch, dirate=1)
-
-        self.rebnconv4 = REBNCONV(mid_ch, mid_ch, dirate=2)
-
-        self.rebnconv3d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv2d = REBNCONV(mid_ch * 2, mid_ch, dirate=1)
-        self.rebnconv1d = REBNCONV(mid_ch * 2, out_ch, dirate=1)
-
-    def forward(self, x):
-        hx = x
-
-        hxin = self.rebnconvin(hx)
-
-        hx1 = self.rebnconv1(hxin)
-        hx = self.pool1(hx1)
-
-        hx2 = self.rebnconv2(hx)
-        hx = self.pool2(hx2)
-
-        hx3 = self.rebnconv3(hx)
-
-        hx4 = self.rebnconv4(hx3)
-
-        hx3d = self.rebnconv3d(torch.cat((hx4, hx3), 1))
-        hx3dup = _upsample_like(hx3d, hx2)
-
-        hx2d = self.rebnconv2d(torch.cat((hx3dup, hx2), 1))
-        hx2dup = _upsample_like(hx2d, hx1)
-
-        hx1d = self.rebnconv1d(torch.cat((hx2dup, hx1), 1))
-
-        return hx1d + hxin
-
-
-### RSU-4F ###
-class RSU4F(nn.Module):  # UNet04FRES(nn.Module):
-
-    def __init__(self, in_ch=3, mid_ch=12, out_ch=3):
-        super(RSU4F, self).__init__()
-
-        self.rebnconvin = REBNCONV(in_ch, out_ch, dirate=1)
-
-        self.rebnconv1 = REBNCONV(out_ch, mid_ch, dirate=1)
-        self.rebnconv2 = REBNCONV(mid_ch, mid_ch, dirate=2)
-        self.rebnconv3 = REBNCONV(mid_ch, mid_ch, dirate=4)
-
-        self.rebnconv4 = REBNCONV(mid_ch, mid_ch, dirate=8)
-
-        self.rebnconv3d = REBNCONV(mid_ch * 2, mid_ch, dirate=4)
-        self.rebnconv2d = REBNCONV(mid_ch * 2, mid_ch, dirate=2)
-        self.rebnconv1d = REBNCONV(mid_ch * 2, out_ch, dirate=1)
-
-    def forward(self, x):
-        hx = x
-
-        hxin = self.rebnconvin(hx)
-
-        hx1 = self.rebnconv1(hxin)
-        hx2 = self.rebnconv2(hx1)
-        hx3 = self.rebnconv3(hx2)
-
-        hx4 = self.rebnconv4(hx3)
-
-        hx3d = self.rebnconv3d(torch.cat((hx4, hx3), 1))
-        hx2d = self.rebnconv2d(torch.cat((hx3d, hx2), 1))
-        hx1d = self.rebnconv1d(torch.cat((hx2d, hx1), 1))
-
-        return hx1d + hxin
-
-
-##### U^2-Net ####
-class U2NET(nn.Module):
-
-    def __init__(self, in_ch=3, out_ch=1):
-        super(U2NET, self).__init__()
-
-        self.stage1 = RSU7(in_ch, 32, 64)
-        self.pool12 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.stage2 = RSU6(64, 32, 128)
-        self.pool23 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.stage3 = RSU5(128, 64, 256)
-        self.pool34 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.stage4 = RSU4(256, 128, 512)
-        self.pool45 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.stage5 = RSU4F(512, 256, 512)
-        self.pool56 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.stage6 = RSU4F(512, 256, 512)
-
-        # decoder
-        self.stage5d = RSU4F(1024, 256, 512)
-        self.stage4d = RSU4(1024, 128, 256)
-        self.stage3d = RSU5(512, 64, 128)
-        self.stage2d = RSU6(256, 32, 64)
-        self.stage1d = RSU7(128, 16, 64)
-
-        self.side1 = nn.Conv2d(64, out_ch, 3, padding=1)
-        self.side2 = nn.Conv2d(64, out_ch, 3, padding=1)
-        self.side3 = nn.Conv2d(128, out_ch, 3, padding=1)
-        self.side4 = nn.Conv2d(256, out_ch, 3, padding=1)
-        self.side5 = nn.Conv2d(512, out_ch, 3, padding=1)
-        self.side6 = nn.Conv2d(512, out_ch, 3, padding=1)
-
-        self.outconv = nn.Conv2d(6 * out_ch, out_ch, 1)
-
-    def forward(self, x):
-        hx = x
-
-        # stage 1
-        hx1 = self.stage1(hx)
-        hx = self.pool12(hx1)
-
-        # stage 2
-        hx2 = self.stage2(hx)
-        hx = self.pool23(hx2)
-
-        # stage 3
-        hx3 = self.stage3(hx)
-        hx = self.pool34(hx3)
-
-        # stage 4
-        hx4 = self.stage4(hx)
-        hx = self.pool45(hx4)
-
-        # stage 5
-        hx5 = self.stage5(hx)
-        hx = self.pool56(hx5)
-
-        # stage 6
-        hx6 = self.stage6(hx)
-        hx6up = _upsample_like(hx6, hx5)
-
-        # -------------------- decoder --------------------
-        hx5d = self.stage5d(torch.cat((hx6up, hx5), 1))
-        hx5dup = _upsample_like(hx5d, hx4)
-
-        hx4d = self.stage4d(torch.cat((hx5dup, hx4), 1))
-        hx4dup = _upsample_like(hx4d, hx3)
-
-        hx3d = self.stage3d(torch.cat((hx4dup, hx3), 1))
-        hx3dup = _upsample_like(hx3d, hx2)
-
-        hx2d = self.stage2d(torch.cat((hx3dup, hx2), 1))
-        hx2dup = _upsample_like(hx2d, hx1)
-
-        hx1d = self.stage1d(torch.cat((hx2dup, hx1), 1))
-
-        # side output
-        d1 = self.side1(hx1d)
-
-        d2 = self.side2(hx2d)
-        d2 = _upsample_like(d2, d1)
-
-        d3 = self.side3(hx3d)
-        d3 = _upsample_like(d3, d1)
-
-        d4 = self.side4(hx4d)
-        d4 = _upsample_like(d4, d1)
-
-        d5 = self.side5(hx5d)
-        d5 = _upsample_like(d5, d1)
-
-        d6 = self.side6(hx6)
-        d6 = _upsample_like(d6, d1)
-
-        d0 = self.outconv(torch.cat((d1, d2, d3, d4, d5, d6), 1))
-
-        return F.sigmoid(d0), F.sigmoid(d1), F.sigmoid(d2), F.sigmoid(d3), F.sigmoid(d4), F.sigmoid(d5), F.sigmoid(d6)
-
-
-### U^2-Net small ###
-class U2NETP(nn.Module):
-
-    def __init__(self, in_ch=3, out_ch=1):
-        super(U2NETP, self).__init__()
-
-        self.stage1 = RSU7(in_ch, 16, 64)
-        self.pool12 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.stage2 = RSU6(64, 16, 64)
-        self.pool23 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.stage3 = RSU5(64, 16, 64)
-        self.pool34 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.stage4 = RSU4(64, 16, 64)
-        self.pool45 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.stage5 = RSU4F(64, 16, 64)
-        self.pool56 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
-
-        self.stage6 = RSU4F(64, 16, 64)
-
-        # decoder
-        self.stage5d = RSU4F(128, 16, 64)
-        self.stage4d = RSU4(128, 16, 64)
-        self.stage3d = RSU5(128, 16, 64)
-        self.stage2d = RSU6(128, 16, 64)
-        self.stage1d = RSU7(128, 16, 64)
-
-        self.side1 = nn.Conv2d(64, out_ch, 3, padding=1)
-        self.side2 = nn.Conv2d(64, out_ch, 3, padding=1)
-        self.side3 = nn.Conv2d(64, out_ch, 3, padding=1)
-        self.side4 = nn.Conv2d(64, out_ch, 3, padding=1)
-        self.side5 = nn.Conv2d(64, out_ch, 3, padding=1)
-        self.side6 = nn.Conv2d(64, out_ch, 3, padding=1)
-
-        self.outconv = nn.Conv2d(6 * out_ch, out_ch, 1)
-
-    def forward(self, x):
-        hx = x      # [B, c_in, 128, 128]
-
-        # stage 1
-        hx1 = self.stage1(hx)   # [8, 64, 128, 128]
-        hx = self.pool12(hx1)   # [8, 64, 64, 64]
-
-        # stage 2
-        hx2 = self.stage2(hx)
-        hx = self.pool23(hx2)
-
-        # stage 3
-        hx3 = self.stage3(hx)
-        hx = self.pool34(hx3)
-
-        # stage 4
-        hx4 = self.stage4(hx)
-        hx = self.pool45(hx4)
-
-        # stage 5
-        hx5 = self.stage5(hx)
-        hx = self.pool56(hx5)
-
-        # stage 6
-        hx6 = self.stage6(hx)   # [8, 64, 4, 4]
-        hx6up = _upsample_like(hx6, hx5)
-
-        # decoder
-        hx5d = self.stage5d(torch.cat((hx6up, hx5), 1))
-        hx5dup = _upsample_like(hx5d, hx4)
-
-        hx4d = self.stage4d(torch.cat((hx5dup, hx4), 1))
-        hx4dup = _upsample_like(hx4d, hx3)
-
-        hx3d = self.stage3d(torch.cat((hx4dup, hx3), 1))
-        hx3dup = _upsample_like(hx3d, hx2)
-
-        hx2d = self.stage2d(torch.cat((hx3dup, hx2), 1))
-        hx2dup = _upsample_like(hx2d, hx1)
-
-        hx1d = self.stage1d(torch.cat((hx2dup, hx1), 1))
-
-        # side output
-        d1 = self.side1(hx1d)
-
-        d2 = self.side2(hx2d)
-        d2 = _upsample_like(d2, d1)
-
-        d3 = self.side3(hx3d)
-        d3 = _upsample_like(d3, d1)
-
-        d4 = self.side4(hx4d)
-        d4 = _upsample_like(d4, d1)
-
-        d5 = self.side5(hx5d)
-        d5 = _upsample_like(d5, d1)
-
-        d6 = self.side6(hx6)
-        d6 = _upsample_like(d6, d1)
-
-        d0 = self.outconv(torch.cat((d1, d2, d3, d4, d5, d6), 1))
-
-        return F.sigmoid(d0), F.sigmoid(d1), F.sigmoid(d2), F.sigmoid(d3), F.sigmoid(d4), F.sigmoid(d5), F.sigmoid(d6)
diff --git a/LogicImplement/models/__init__.py b/LogicImplement/models/__init__.py
deleted file mode 100644
index 67f768c..0000000
--- a/LogicImplement/models/__init__.py
+++ /dev/null
@@ -1,2 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
diff --git a/LogicImplement/models/__pycache__/U2net.cpython-38.pyc b/LogicImplement/models/__pycache__/U2net.cpython-38.pyc
deleted file mode 100644
index ec17f6d..0000000
Binary files a/LogicImplement/models/__pycache__/U2net.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/models/__pycache__/__init__.cpython-38.pyc b/LogicImplement/models/__pycache__/__init__.cpython-38.pyc
deleted file mode 100644
index 85b9541..0000000
Binary files a/LogicImplement/models/__pycache__/__init__.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/models/__pycache__/loss_function.cpython-38.pyc b/LogicImplement/models/__pycache__/loss_function.cpython-38.pyc
deleted file mode 100644
index 15e7a61..0000000
Binary files a/LogicImplement/models/__pycache__/loss_function.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/models/__pycache__/lovasz_losses.cpython-38.pyc b/LogicImplement/models/__pycache__/lovasz_losses.cpython-38.pyc
deleted file mode 100644
index 939fb98..0000000
Binary files a/LogicImplement/models/__pycache__/lovasz_losses.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/models/__pycache__/nested_unet.cpython-38.pyc b/LogicImplement/models/__pycache__/nested_unet.cpython-38.pyc
deleted file mode 100644
index 3db6671..0000000
Binary files a/LogicImplement/models/__pycache__/nested_unet.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/models/__pycache__/network_blocks.cpython-38.pyc b/LogicImplement/models/__pycache__/network_blocks.cpython-38.pyc
deleted file mode 100644
index 2eb7be6..0000000
Binary files a/LogicImplement/models/__pycache__/network_blocks.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/models/__pycache__/nfn_plus.cpython-38.pyc b/LogicImplement/models/__pycache__/nfn_plus.cpython-38.pyc
deleted file mode 100644
index 9f71eef..0000000
Binary files a/LogicImplement/models/__pycache__/nfn_plus.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/models/__pycache__/unet.cpython-38.pyc b/LogicImplement/models/__pycache__/unet.cpython-38.pyc
deleted file mode 100644
index c946e5b..0000000
Binary files a/LogicImplement/models/__pycache__/unet.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/models/__pycache__/unet_tiny.cpython-38.pyc b/LogicImplement/models/__pycache__/unet_tiny.cpython-38.pyc
deleted file mode 100644
index c7ff844..0000000
Binary files a/LogicImplement/models/__pycache__/unet_tiny.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/models/__pycache__/xnet.cpython-38.pyc b/LogicImplement/models/__pycache__/xnet.cpython-38.pyc
deleted file mode 100644
index 4fc06fe..0000000
Binary files a/LogicImplement/models/__pycache__/xnet.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/models/compute_flops.py b/LogicImplement/models/compute_flops.py
deleted file mode 100644
index 2655459..0000000
--- a/LogicImplement/models/compute_flops.py
+++ /dev/null
@@ -1,23 +0,0 @@
-# -*- coding: utf-8 -*-
-
-
-import torchvision.models as models
-import torch
-from ptflops import get_model_complexity_info
-from models import unet, nested_unet, loss_function, xnet, nfn_plus, lovasz_losses, unet_tiny, U2net
-
-
-with torch.cuda.device(0):
-    net = U2net.U2NET(out_ch=2, in_ch=3)
-    # net = unet.UNet(num_classes=2, in_channels=3, is_attention=True,
-    #                 is_recurrent_residual=False)
-# net = models.densenet161()
-    macs, params = get_model_complexity_info(net, (3, 128, 128), as_strings=True,
-                                           print_per_layer_stat=True, verbose=True)
-    print('{:<30}  {:<8}'.format('Computational complexity: ', macs))
-    print('{:<30}  {:<8}'.format('Number of parameters: ', params))
-
-"""
-U2NETP: Number of parameters:           1.13 M 
-U2NET:  Number of parameters:           44.02 M 
-"""
\ No newline at end of file
diff --git a/LogicImplement/models/loss_function.py b/LogicImplement/models/loss_function.py
deleted file mode 100644
index 618ebaf..0000000
--- a/LogicImplement/models/loss_function.py
+++ /dev/null
@@ -1,112 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-from models.lovasz_losses import lovasz_softmax
-from torch import nn
-from torch.nn import functional as F
-import torch
-
-
-class LovaszLoss(nn.Module):
-    """lovasz损失函数，https://github.com/bermanmaxim/LovaszSoftmax"""
-
-    def __init__(self):
-        super().__init__()
-
-    def forward(self, input, target):
-        probability = F.softmax(input, dim=1)
-        loss = lovasz_softmax(probability, target)
-        return loss
-
-
-class SoftDiceLoss(nn.Module):
-    """Dice损失函数，仅用于二分类，https://www.aiuai.cn/aifarm1159.html，https://github.com/pytorch/pytorch/issues/1249"""
-
-    def __init__(self, num_classes, smooth=1):
-        """:arg smooth Dice系数分母分子上添加的参数，防止除0错误"""
-
-        super().__init__()
-        self.num_classes = num_classes
-        self.smooth = smooth
-
-    def forward(self, input, target):
-        # 确保SoftDiceLoss只用于二分类问题
-        assert torch.max(target).item() <= 1, 'SoftDiceLoss()目前只能用于二分类'
-
-        batch_size = input.size(0)
-        probability = F.softmax(input, dim=1)
-
-        # 将probability和target矩阵flatten为B*(C*H*W)的矩阵
-        probability = probability.view(batch_size, -1)
-        # 这里的target label需要是one hot格式，one hot要显示传入类别数量
-        target_one_hot = F.one_hot(target, num_classes=self.num_classes).permute((0, 3, 1, 2))
-        target_one_hot = target_one_hot.contiguous().view(batch_size, -1)
-        # 计算相交区域的概率矩阵
-        intersection = probability * target_one_hot
-
-        # 计算每一个batch的Dice相似系数（Dice similar coefficient）
-        dsc = (2 * intersection.sum(dim=1) + self.smooth) / (
-                target_one_hot.sum(dim=1) + probability.sum(dim=1) + self.smooth)
-        loss = (1 - dsc).sum()
-
-        return loss
-
-
-# class BCELoss2d(nn.Module):
-#     """二进制交叉熵，计算概率的函数可选Sigmoid和Softmax，https://www.aiuai.cn/aifarm1159.html"""
-#
-#     def __init__(self, p_function='sigmoid', weight=None, size_average=True):
-#         super().__init__()
-#
-#         assert p_function in ['softmax', 'sigmoid']
-#
-#         self.p_function = p_function
-#         self.bce_loss_function = nn.BCELoss(weight=weight, size_average=size_average)
-#
-#     def forward(self, input, target):
-#         if self.p_function == 'softmax':
-#             probability = F.softmax(input, dim=1)
-#         else:
-#             probability = F.sigmoid(input)
-#         loss = self.bce_loss_function(probability, target)
-#         return loss
-
-class DiceAndBCELoss(nn.Module):
-    """Dice损失和BCE损失的混合"""
-
-    def __init__(self, num_classes):
-        super().__init__()
-        self.dice_loss = SoftDiceLoss(num_classes=num_classes)
-        self.bce_loss = nn.CrossEntropyLoss()
-
-    def forward(self, input, target):
-        return self.dice_loss(input, target) + self.bce_loss(input, target)
-
-
-class SCELoss(nn.Module):
-    """用于Learning with noisy labels的对称交叉熵损失函数"""
-
-    def __init__(self, num_classes, alpha=1.0, beta=1.0):
-        super().__init__()
-
-        self.alpha = alpha
-        self.beta = beta
-        self.num_classes = num_classes
-        self.cross_entropy = nn.CrossEntropyLoss()
-
-    def forward(self, input, target):
-        # CCE
-        ce = self.cross_entropy(input, target)
-
-        # RCE
-        input = F.softmax(input, dim=1)
-        input = torch.clamp(input, min=1e-7, max=1.0)
-        target_one_hot = F.one_hot(target, self.num_classes).permute((0, 3, 1, 2)).float()
-        target_one_hot = torch.clamp(target_one_hot, min=1e-4, max=1.0)
-        rce = (-1 * torch.sum(input * torch.log(target_one_hot), dim=1))
-
-        # Loss
-        loss = self.alpha * ce + self.beta * rce.mean()
-        return loss
-
-    def __str__(self):
-        return 'SCELoss(alpha={}, beta={})'.format(self.alpha, self.beta)
diff --git a/LogicImplement/models/lovasz_losses.py b/LogicImplement/models/lovasz_losses.py
deleted file mode 100644
index 57f61c3..0000000
--- a/LogicImplement/models/lovasz_losses.py
+++ /dev/null
@@ -1,254 +0,0 @@
-"""
-Lovasz-Softmax and Jaccard hinge loss in PyTorch
-Maxim Berman 2018 ESAT-PSI KU Leuven (MIT License)
-"""
-
-from __future__ import print_function, division
-
-import torch
-from torch.autograd import Variable
-import torch.nn.functional as F
-import numpy as np
-try:
-    from itertools import  ifilterfalse
-except ImportError:
-    from itertools import  filterfalse as ifilterfalse
-
-
-def lovasz_grad(gt_sorted):
-    """
-    Computes gradient of the Lovasz extension w.r.t sorted errors
-    See Alg. 1 in paper
-    """
-    p = len(gt_sorted)
-    gts = gt_sorted.sum()
-    intersection = gts - gt_sorted.float().cumsum(0)
-    union = gts + (1 - gt_sorted).float().cumsum(0)
-    jaccard = 1. - intersection / union
-    if p > 1: # cover 1-pixel case
-        jaccard[1:p] = jaccard[1:p] - jaccard[0:-1]
-    return jaccard
-
-
-def iou_binary(preds, labels, EMPTY=1., ignore=None, per_image=True):
-    """
-    IoU for foreground class
-    binary: 1 foreground, 0 background
-    """
-    if not per_image:
-        preds, labels = (preds,), (labels,)
-    ious = []
-    for pred, label in zip(preds, labels):
-        intersection = ((label == 1) & (pred == 1)).sum()
-        union = ((label == 1) | ((pred == 1) & (label != ignore))).sum()
-        if not union:
-            iou = EMPTY
-        else:
-            iou = float(intersection) / float(union)
-        ious.append(iou)
-    iou = mean(ious)    # mean accross images if per_image
-    return 100 * iou
-
-
-def iou(preds, labels, C, EMPTY=1., ignore=None, per_image=False):
-    """
-    Array of IoU for each (non ignored) class
-    """
-    if not per_image:
-        preds, labels = (preds,), (labels,)
-    ious = []
-    for pred, label in zip(preds, labels):
-        iou = []    
-        for i in range(C):
-            if i != ignore: # The ignored label is sometimes among predicted classes (ENet - CityScapes)
-                intersection = ((label == i) & (pred == i)).sum()
-                union = ((label == i) | ((pred == i) & (label != ignore))).sum()
-                if not union:
-                    iou.append(EMPTY)
-                else:
-                    iou.append(float(intersection) / float(union))
-        ious.append(iou)
-    ious = [mean(iou) for iou in zip(*ious)] # mean accross images if per_image
-    return 100 * np.array(ious)
-
-
-# --------------------------- BINARY LOSSES ---------------------------
-
-
-def lovasz_hinge(logits, labels, per_image=True, ignore=None):
-    """
-    Binary Lovasz hinge loss
-      logits: [B, H, W] Variable, logits at each pixel (between -\infty and +\infty)
-      labels: [B, H, W] Tensor, binary ground truth masks (0 or 1)
-      per_image: compute the loss per image instead of per batch
-      ignore: void class id
-    """
-    if per_image:
-        loss = mean(lovasz_hinge_flat(*flatten_binary_scores(log.unsqueeze(0), lab.unsqueeze(0), ignore))
-                          for log, lab in zip(logits, labels))
-    else:
-        loss = lovasz_hinge_flat(*flatten_binary_scores(logits, labels, ignore))
-    return loss
-
-
-def lovasz_hinge_flat(logits, labels):
-    """
-    Binary Lovasz hinge loss
-      logits: [P] Variable, logits at each prediction (between -\infty and +\infty)
-      labels: [P] Tensor, binary ground truth labels (0 or 1)
-      ignore: label to ignore
-    """
-    if len(labels) == 0:
-        # only void pixels, the gradients should be 0
-        return logits.sum() * 0.
-    signs = 2. * labels.float() - 1.
-    errors = (1. - logits * Variable(signs))
-    errors_sorted, perm = torch.sort(errors, dim=0, descending=True)
-    perm = perm.data
-    gt_sorted = labels[perm]
-    grad = lovasz_grad(gt_sorted)
-    loss = torch.dot(F.relu(errors_sorted), Variable(grad))
-    return loss
-
-
-def flatten_binary_scores(scores, labels, ignore=None):
-    """
-    Flattens predictions in the batch (binary case)
-    Remove labels equal to 'ignore'
-    """
-    scores = scores.view(-1)
-    labels = labels.view(-1)
-    if ignore is None:
-        return scores, labels
-    valid = (labels != ignore)
-    vscores = scores[valid]
-    vlabels = labels[valid]
-    return vscores, vlabels
-
-
-class StableBCELoss(torch.nn.modules.Module):
-    def __init__(self):
-         super(StableBCELoss, self).__init__()
-    def forward(self, input, target):
-         neg_abs = - input.abs()
-         loss = input.clamp(min=0) - input * target + (1 + neg_abs.exp()).log()
-         return loss.mean()
-
-
-def binary_xloss(logits, labels, ignore=None):
-    """
-    Binary Cross entropy loss
-      logits: [B, H, W] Variable, logits at each pixel (between -\infty and +\infty)
-      labels: [B, H, W] Tensor, binary ground truth masks (0 or 1)
-      ignore: void class id
-    """
-    logits, labels = flatten_binary_scores(logits, labels, ignore)
-    loss = StableBCELoss()(logits, Variable(labels.float()))
-    return loss
-
-
-# --------------------------- MULTICLASS LOSSES ---------------------------
-
-
-def lovasz_softmax(probas, labels, classes='present', per_image=False, ignore=None):
-    """
-    Multi-class Lovasz-Softmax loss
-      probas: [B, C, H, W] Variable, class probabilities at each prediction (between 0 and 1).
-              Interpreted as binary (sigmoid) output with outputs of size [B, H, W].
-      labels: [B, H, W] Tensor, ground truth labels (between 0 and C - 1)
-      classes: 'all' for all, 'present' for classes present in labels, or a list of classes to average.
-      per_image: compute the loss per image instead of per batch
-      ignore: void class labels
-    """
-    if per_image:
-        loss = mean(lovasz_softmax_flat(*flatten_probas(prob.unsqueeze(0), lab.unsqueeze(0), ignore), classes=classes)
-                          for prob, lab in zip(probas, labels))
-    else:
-        loss = lovasz_softmax_flat(*flatten_probas(probas, labels, ignore), classes=classes)
-    return loss
-
-
-def lovasz_softmax_flat(probas, labels, classes='present'):
-    """
-    Multi-class Lovasz-Softmax loss
-      probas: [P, C] Variable, class probabilities at each prediction (between 0 and 1)
-      labels: [P] Tensor, ground truth labels (between 0 and C - 1)
-      classes: 'all' for all, 'present' for classes present in labels, or a list of classes to average.
-    """
-    if probas.numel() == 0:
-        # only void pixels, the gradients should be 0
-        return probas * 0.
-    C = probas.size(1)
-    losses = []
-    class_to_sum = list(range(C)) if classes in ['all', 'present'] else classes
-    for c in class_to_sum:
-        fg = (labels == c).float() # foreground for class c
-        if (classes == 'present' and fg.sum() == 0):
-            continue
-        if C == 1:
-            if len(classes) > 1:
-                raise ValueError('Sigmoid output possible only with 1 class')
-            class_pred = probas[:, 0]
-        else:
-            class_pred = probas[:, c]
-        errors = (Variable(fg) - class_pred).abs()
-        errors_sorted, perm = torch.sort(errors, 0, descending=True)
-        perm = perm.data
-        fg_sorted = fg[perm]
-        loss = torch.dot(errors_sorted, Variable(lovasz_grad(fg_sorted)))
-        # loss = loss * (c + 1)
-        if c == 2:
-            loss = loss * 2
-        losses.append(loss)   # 3cls:0.6294, 0.5980
-    return mean(losses)
-
-
-def flatten_probas(probas, labels, ignore=None):
-    """
-    Flattens predictions in the batch
-    """
-    if probas.dim() == 3:
-        # assumes output of a sigmoid layer
-        B, H, W = probas.size()
-        probas = probas.view(B, 1, H, W)
-    B, C, H, W = probas.size()
-    probas = probas.permute(0, 2, 3, 1).contiguous().view(-1, C)  # B * H * W, C = P, C
-    labels = labels.view(-1)
-    if ignore is None:
-        return probas, labels
-    valid = (labels != ignore)
-    vprobas = probas[valid.nonzero().squeeze()]
-    vlabels = labels[valid]
-    return vprobas, vlabels
-
-def xloss(logits, labels, ignore=None):
-    """
-    Cross entropy loss
-    """
-    return F.cross_entropy(logits, Variable(labels), ignore_index=255)
-
-
-# --------------------------- HELPER FUNCTIONS ---------------------------
-def isnan(x):
-    return x != x
-    
-    
-def mean(l, ignore_nan=False, empty=0):
-    """
-    nanmean compatible with generators.
-    """
-    l = iter(l)
-    if ignore_nan:
-        l = ifilterfalse(isnan, l)
-    try:
-        n = 1
-        acc = next(l)
-    except StopIteration:
-        if empty == 'raise':
-            raise ValueError('Empty mean')
-        return empty
-    for n, v in enumerate(l, 2):
-        acc += v
-    if n == 1:
-        return acc
-    return acc / n
diff --git a/LogicImplement/models/nested_unet.py b/LogicImplement/models/nested_unet.py
deleted file mode 100644
index f8fd947..0000000
--- a/LogicImplement/models/nested_unet.py
+++ /dev/null
@@ -1,113 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-from torch import nn
-import torch
-from .network_blocks import DoubleConvBlock
-
-
-class NestedUNet(nn.Module):
-
-    def __init__(self, num_classes, in_channels, is_deep_supervision=True, is_dense_connections=True, l_num=4):
-        super().__init__()
-
-        assert l_num in [2, 3, 4], "目前实现的UNet++只有4层"
-
-        nb_filters = [32, 64, 128, 256, 512]
-
-        self.is_deep_supervision = is_deep_supervision
-        self.is_dense_connections = is_dense_connections
-        self.l_num = l_num
-
-        # 这2个层内没有需要学习的参数，可以共享
-        # 使用MaxPooling进行2倍下采样
-        self.downsampling = nn.MaxPool2d(kernel_size=2, stride=2)
-        # 使用双线性插值进行2倍上采样，对于pixel-wise任务，align_corners设置为True
-        self.upsampling = nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True)
-
-        self.conv0_0 = DoubleConvBlock(in_channels, nb_filters[0], nb_filters[0])
-        self.conv1_0 = DoubleConvBlock(nb_filters[0], nb_filters[1], nb_filters[1])
-        self.conv2_0 = DoubleConvBlock(nb_filters[1], nb_filters[2], nb_filters[2])
-        self.conv3_0 = DoubleConvBlock(nb_filters[2], nb_filters[3], nb_filters[3])
-        self.conv4_0 = DoubleConvBlock(nb_filters[3], nb_filters[4], nb_filters[4])
-
-        self.conv0_1 = DoubleConvBlock(nb_filters[0] + nb_filters[1], nb_filters[0], nb_filters[0])
-        self.conv1_1 = DoubleConvBlock(nb_filters[1] + nb_filters[2], nb_filters[1], nb_filters[1])
-        self.conv2_1 = DoubleConvBlock(nb_filters[2] + nb_filters[3], nb_filters[2], nb_filters[2])
-        self.conv3_1 = DoubleConvBlock(nb_filters[3] + nb_filters[4], nb_filters[3], nb_filters[3])
-
-        # 是否使用dense connections
-        if self.is_dense_connections:
-            self.conv0_2 = DoubleConvBlock(nb_filters[0] * 2 + nb_filters[1], nb_filters[0], nb_filters[0])
-            self.conv1_2 = DoubleConvBlock(nb_filters[1] * 2 + nb_filters[2], nb_filters[1], nb_filters[1])
-            self.conv2_2 = DoubleConvBlock(nb_filters[2] * 2 + nb_filters[3], nb_filters[2], nb_filters[2])
-
-            self.conv0_3 = DoubleConvBlock(nb_filters[0] * 3 + nb_filters[1], nb_filters[0], nb_filters[0])
-            self.conv1_3 = DoubleConvBlock(nb_filters[1] * 3 + nb_filters[2], nb_filters[1], nb_filters[1])
-
-            self.conv0_4 = DoubleConvBlock(nb_filters[0] * 4 + nb_filters[1], nb_filters[0], nb_filters[0])
-        else:
-            self.conv0_2 = DoubleConvBlock(nb_filters[0] + nb_filters[1], nb_filters[0], nb_filters[0])
-            self.conv1_2 = DoubleConvBlock(nb_filters[1] + nb_filters[2], nb_filters[1], nb_filters[1])
-            self.conv2_2 = DoubleConvBlock(nb_filters[2] + nb_filters[3], nb_filters[2], nb_filters[2])
-
-            self.conv0_3 = DoubleConvBlock(nb_filters[0] + nb_filters[1], nb_filters[0], nb_filters[0])
-            self.conv1_3 = DoubleConvBlock(nb_filters[1] + nb_filters[2], nb_filters[1], nb_filters[1])
-
-            self.conv0_4 = DoubleConvBlock(nb_filters[0] + nb_filters[1], nb_filters[0], nb_filters[0])
-
-        # 如果使用deep supervision，则每层UNet都有一个输出
-        if self.is_deep_supervision:
-            self.final1 = nn.Conv2d(nb_filters[0], num_classes, kernel_size=1)
-            self.final2 = nn.Conv2d(nb_filters[0], num_classes, kernel_size=1)
-            self.final3 = nn.Conv2d(nb_filters[0], num_classes, kernel_size=1)
-            self.final4 = nn.Conv2d(nb_filters[0], num_classes, kernel_size=1)
-        else:
-            self.final = nn.Conv2d(nb_filters[0], num_classes, kernel_size=1)
-
-        # 删除多余层的变量
-        if self.l_num == 2:
-            del self.conv3_0, self.conv2_1, self.conv1_2, self.conv0_3
-            del self.conv4_0, self.conv3_1, self.conv2_2, self.conv1_3, self.conv0_4
-        if self.l_num == 3:
-            del self.conv4_0, self.conv3_1, self.conv2_2, self.conv1_3, self.conv0_4
-
-    def forward(self, input):
-        x0_0 = self.conv0_0(input)
-        x1_0 = self.conv1_0(self.downsampling(x0_0))
-        x0_1 = self.conv0_1(torch.cat([x0_0, self.upsampling(x1_0)], dim=1))
-
-        x2_0 = self.conv2_0(self.downsampling(x1_0))
-        x1_1 = self.conv1_1(torch.cat([x1_0, self.upsampling(x2_0)], dim=1))
-        x0_2 = self.conv0_2(torch.cat([x0_0, x0_1, self.upsampling(x1_1)], dim=1))
-
-        if self.l_num == 2:
-            if self.is_deep_supervision:
-                return [self.final1(x0_1), self.final2(x0_2)]
-            else:
-                return self.final(x0_2)
-
-        x3_0 = self.conv3_0(self.downsampling(x2_0))
-        x2_1 = self.conv2_1(torch.cat([x2_0, self.upsampling(x3_0)], dim=1))
-        x1_2 = self.conv1_2(torch.cat([x1_0, x1_1, self.upsampling(x2_1)], dim=1))
-        x0_3 = self.conv0_3(torch.cat([x0_0, x0_1, x0_2, self.upsampling(x1_2)], dim=1))
-
-        if self.l_num == 3:
-            if self.is_deep_supervision:
-                return [self.final1(x0_1), self.final2(x0_2), self.final3(x0_3)]
-            else:
-                return self.final(x0_3)
-
-        x4_0 = self.conv4_0(self.downsampling(x3_0))
-        x3_1 = self.conv3_1(torch.cat([x3_0, self.upsampling(x4_0)], dim=1))
-        x2_2 = self.conv2_2(torch.cat([x2_0, x2_1, self.upsampling(x3_1)], dim=1))
-        x1_3 = self.conv1_3(torch.cat([x1_0, x1_1, x1_2, self.upsampling(x2_2)], dim=1))
-        x0_4 = self.conv0_4(torch.cat([x0_0, x0_1, x0_2, x0_3, self.upsampling(x1_3)], dim=1))
-
-        if self.is_deep_supervision:
-            return [self.final1(x0_1), self.final2(x0_2), self.final3(x0_3), self.final4(x0_4)]
-        else:
-            return self.final(x0_4)
-
-    def __str__(self):
-        return 'UNet++(is_deep_supervision={}, is_dense_connections={}, l_num={})'.format(
-            self.is_deep_supervision, self.is_dense_connections, self.l_num)
diff --git a/LogicImplement/models/network_blocks.py b/LogicImplement/models/network_blocks.py
deleted file mode 100644
index b7c6b8d..0000000
--- a/LogicImplement/models/network_blocks.py
+++ /dev/null
@@ -1,135 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-# 所用到的所有网络的基本模块
-from torch import nn
-from torch.nn import functional as F
-
-
-class SingleConvBlock(nn.Module):
-    """Conv-BN-ReLU的基本模块，其中的Conv并不会张量的尺寸"""
-
-    def __init__(self, in_channels, out_channels):
-        super().__init__()
-        self.block = nn.Sequential(
-            nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=1, padding=1),
-            nn.BatchNorm2d(out_channels),
-            nn.ReLU(inplace=True)
-        )
-
-    def forward(self, input):
-        return self.block(input)
-
-
-class DoubleConvBlock(nn.Module):
-    """由2个Conv-BN-ReLu组成的基本模块"""
-
-    def __init__(self, in_channels, middle_channels, out_channels, dropout=False, p=0.2):
-        super().__init__()
-        # layers = [
-        #     SingleConvBlock(in_channels, middle_channels),
-        #     SingleConvBlock(middle_channels, out_channels)
-        # ]
-        #
-        # if dropout:
-        #     layers.append(nn.Dropout())
-
-        # 将Dropout加在2个卷积层的中间
-        layers = [SingleConvBlock(in_channels, middle_channels)]
-        if dropout:
-            layers.append(nn.Dropout(p=p))
-        layers.append(SingleConvBlock(middle_channels, out_channels))
-
-        self.block = nn.Sequential(*layers)
-
-    def forward(self, input):
-        return self.block(input)
-
-
-class RecurrentSingleBlock(nn.Module):
-    """具有循环结构的SingleBlock，其in_channels=out_channels"""
-
-    def __init__(self, channels, t):
-        assert t > 1, "循环次数必须>1"
-        super().__init__()
-        self.t = t
-        self.block = SingleConvBlock(channels, channels)
-
-    def forward(self, input):
-        for i in range(self.t):
-            if i == 0:
-                output = self.block(input)
-            # 循环时，要加上最开始的输入
-            output = self.block(output + input)
-        return output
-
-
-class ResidualRecurrentBlock(nn.Module):
-    """由2个RecurrentBlock和残差连接组成的模块"""
-
-    def __init__(self, in_channels, out_channels, t=2):
-        super().__init__()
-        self.conv = nn.Conv2d(in_channels, out_channels, kernel_size=1, stride=1, padding=0)
-        self.r2_block = nn.Sequential(
-            # 在循环模块前加一个Conv用于改变数据的channel数
-            RecurrentSingleBlock(out_channels, t),
-            RecurrentSingleBlock(out_channels, t)
-        )
-
-    def forward(self, input):
-        # 残差连接
-        conv_input = self.conv(input)
-        return self.r2_block(conv_input) + conv_input
-
-
-class AttentionBlock2d(nn.Module):
-    """注意力模块，计算较粗糙、低语义的feature map中每个点的注意力系数，并返回乘以注意力系数的feature map，
-    按照论文作者的代码进行实现的
-    https://github.com/ozan-oktay/Attention-Gated-Networks/blob/master/models/layers/grid_attention_layer.py"""
-
-    def __init__(self, x_channels, gate_channels, inter_channels=None, subsample_factor=(2, 2)):
-        super().__init__()
-        assert isinstance(subsample_factor, tuple), "subsample_factor必须是tuple类型"
-
-        if inter_channels is None:
-            inter_channels = x_channels // 2
-            if inter_channels == 0:
-                inter_channels = 1
-
-        # TODO: 这里的下采样系数subsample_factor有什么用，bias为什么为False
-        self.x_conv = nn.Conv2d(x_channels, inter_channels, kernel_size=subsample_factor, stride=subsample_factor,
-                                padding=0, bias=False)
-        self.gate_conv = nn.Conv2d(gate_channels, inter_channels, kernel_size=1, stride=1, padding=0)
-        self.psi = nn.Conv2d(inter_channels, 1, kernel_size=1, stride=1, padding=0)
-
-        # 将最终乘以注意力系数的feature map再进行一次卷积
-        # TODO: 为什么后面还要接一个BN
-        self.end_transform = nn.Sequential(
-            nn.Conv2d(x_channels, x_channels, kernel_size=1, stride=1, padding=0),
-            nn.BatchNorm2d(x_channels)
-        )
-
-        # TODO: 是否需要初始化参数
-
-    def forward(self, x, gate):
-        origin_x = x
-        origin_x_size = x.size()
-
-        x = self.x_conv(x)
-        gate = self.gate_conv(gate)
-        # 将卷积后的gate上采样为x的尺寸
-        gate = F.upsample_bilinear(gate, size=x.size()[2:])
-
-        # 将feature map相加后进行ReLU
-        temp = F.relu(gate + x, inplace=True)
-        temp = self.psi(temp)
-        attention_coefficient = F.sigmoid(temp)
-
-        # 将注意力系数上采样为原始x的尺寸，并扩展为对应的channel数
-        attention_coefficient = F.upsample_bilinear(attention_coefficient, size=origin_x_size[2:])
-        attention_coefficient = attention_coefficient.expand_as(origin_x)
-
-        # 注意力系数乘原始的x
-        output = attention_coefficient * origin_x
-        output = self.end_transform(output)
-
-        return output
diff --git a/LogicImplement/models/nfn_plus.py b/LogicImplement/models/nfn_plus.py
deleted file mode 100644
index 3c783c9..0000000
--- a/LogicImplement/models/nfn_plus.py
+++ /dev/null
@@ -1,78 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-import torch
-from torch import nn
-from .network_blocks import DoubleConvBlock
-
-
-class OutputBlock(nn.Module):
-    """论文中提到的输出块"""
-
-    def __init__(self, in_channels, out_channels):
-        super().__init__()
-        self.block = nn.Sequential(
-            nn.Conv2d(in_channels, out_channels, kernel_size=1, stride=1),
-            nn.BatchNorm2d(out_channels),
-            nn.ReLU(inplace=True),
-        )
-
-    def forward(self, input):
-        return self.block(input)
-
-
-class NFNPlus(nn.Module):
-    """具有2个Stage的UNet结构的网络，添加了UNet间的skip connection，
-    https://www.sciencedirect.com/science/article/pii/S0893608020300721
-    """
-
-    def __init__(self, num_classes, in_channels, is_deep_supervision=True, p=0.2):
-        super().__init__()
-        nb_filters = [32, 64, 128]
-        self.is_deep_supervision = is_deep_supervision
-
-        # 标准的2倍上采样和下采样，因为没有可以学习的参数，可以共享
-        self.down = nn.MaxPool2d(kernel_size=2, stride=2)
-        self.up = nn.UpsamplingBilinear2d(scale_factor=2)
-
-        # Front Network
-        self.conv0_0 = DoubleConvBlock(in_channels, nb_filters[0], nb_filters[0], dropout=True, p=p)
-        self.conv0_1 = DoubleConvBlock(nb_filters[0], nb_filters[1], nb_filters[1], dropout=True, p=p)
-        self.conv0_2 = DoubleConvBlock(nb_filters[1], nb_filters[2], nb_filters[2], dropout=True, p=p)
-        self.conv0_3 = DoubleConvBlock(nb_filters[2] + nb_filters[1], nb_filters[1], nb_filters[1], dropout=True, p=p)
-        self.conv0_4 = DoubleConvBlock(nb_filters[1] + nb_filters[0], nb_filters[0], nb_filters[0], dropout=True, p=p)
-
-        # Followed Network
-        self.conv1_0 = DoubleConvBlock(nb_filters[0], nb_filters[0], nb_filters[0], dropout=True, p=p)
-        self.conv1_1 = DoubleConvBlock(nb_filters[0] + nb_filters[1], nb_filters[1], nb_filters[1], dropout=True, p=p)
-        self.conv1_2 = DoubleConvBlock(nb_filters[1] + nb_filters[2], nb_filters[2], nb_filters[2], dropout=True, p=p)
-        self.conv1_3 = DoubleConvBlock(nb_filters[2] + nb_filters[1], nb_filters[1], nb_filters[1], dropout=True, p=p)
-        self.conv1_4 = DoubleConvBlock(nb_filters[1] + nb_filters[0], nb_filters[0], nb_filters[0], dropout=True, p=p)
-
-        if is_deep_supervision:
-            self.final0 = OutputBlock(nb_filters[0], num_classes)
-            self.final1 = OutputBlock(nb_filters[0], num_classes)
-        else:
-            self.final = OutputBlock(nb_filters[0], num_classes)
-
-    def forward(self, input):
-        # Front Network
-        x0_0 = self.conv0_0(input)
-        x0_1 = self.conv0_1(self.down(x0_0))
-        x0_2 = self.conv0_2(self.down(x0_1))
-        x0_3 = self.conv0_3(torch.cat([x0_1, self.up(x0_2)], dim=1))
-        x0_4 = self.conv0_4(torch.cat([x0_0, self.up(x0_3)], dim=1))
-
-        # Followed Network
-        x1_0 = self.conv1_0(x0_4)
-        x1_1 = self.conv1_1(torch.cat([x0_3, self.down(x1_0)], dim=1))
-        x1_2 = self.conv1_2(torch.cat([x0_2, self.down(x1_1)], dim=1))
-        x1_3 = self.conv1_3(torch.cat([x1_1, self.up(x1_2)], dim=1))
-        x1_4 = self.conv1_4(torch.cat([x1_0, self.up(x1_3)], dim=1))
-
-        if self.is_deep_supervision:
-            return [self.final0(x0_4), self.final1(x1_4)]
-        else:
-            return self.final(x1_4)
-
-    def __str__(self):
-        return 'NFN+'
diff --git a/LogicImplement/models/unet.py b/LogicImplement/models/unet.py
deleted file mode 100644
index 0aa76de..0000000
--- a/LogicImplement/models/unet.py
+++ /dev/null
@@ -1,89 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-import torch
-from torch import nn
-from .network_blocks import DoubleConvBlock, AttentionBlock2d, ResidualRecurrentBlock
-
-
-class UNet(nn.Module):
-    """基本的UNet网络，其中的卷积并不会改变张量的尺寸"""
-
-    def __init__(self, num_classes, in_channels, is_attention=False, is_recurrent_residual=False):
-        super().__init__()
-        nb_filters = [32, 64, 128, 256, 512]
-        self.is_attention = is_attention
-        self.is_recurrent_residual = is_recurrent_residual
-
-        # 标准的2倍上采样和下采样，因为没有可以学习的参数，可以共享
-        self.down = nn.MaxPool2d(kernel_size=2, stride=2)
-        self.up = nn.UpsamplingBilinear2d(scale_factor=2)
-
-        if is_recurrent_residual:
-            # 下采样的模块
-            self.conv0_0 = ResidualRecurrentBlock(in_channels, nb_filters[0])
-            self.conv1_0 = ResidualRecurrentBlock(nb_filters[0], nb_filters[1])
-            self.conv2_0 = ResidualRecurrentBlock(nb_filters[1], nb_filters[2])
-            self.conv3_0 = ResidualRecurrentBlock(nb_filters[2], nb_filters[3])
-            self.conv4_0 = ResidualRecurrentBlock(nb_filters[3], nb_filters[4])
-
-            # 上采样的模块
-            self.conv3_1 = ResidualRecurrentBlock(nb_filters[4] + nb_filters[3], nb_filters[3])
-            self.conv2_2 = ResidualRecurrentBlock(nb_filters[3] + nb_filters[2], nb_filters[2])
-            self.conv1_3 = ResidualRecurrentBlock(nb_filters[2] + nb_filters[1], nb_filters[1])
-            self.conv0_4 = ResidualRecurrentBlock(nb_filters[1] + nb_filters[0], nb_filters[0])
-        else:
-            # 下采样的模块
-            self.conv0_0 = DoubleConvBlock(in_channels, nb_filters[0], nb_filters[0])
-            self.conv1_0 = DoubleConvBlock(nb_filters[0], nb_filters[1], nb_filters[1])
-            self.conv2_0 = DoubleConvBlock(nb_filters[1], nb_filters[2], nb_filters[2])
-            self.conv3_0 = DoubleConvBlock(nb_filters[2], nb_filters[3], nb_filters[3])
-            self.conv4_0 = DoubleConvBlock(nb_filters[3], nb_filters[4], nb_filters[4])
-
-            # 上采样的模块
-            self.conv3_1 = DoubleConvBlock(nb_filters[4] + nb_filters[3], nb_filters[3], nb_filters[3])
-            self.conv2_2 = DoubleConvBlock(nb_filters[3] + nb_filters[2], nb_filters[2], nb_filters[2])
-            self.conv1_3 = DoubleConvBlock(nb_filters[2] + nb_filters[1], nb_filters[1], nb_filters[1])
-            self.conv0_4 = DoubleConvBlock(nb_filters[1] + nb_filters[0], nb_filters[0], nb_filters[0])
-
-        # 最后接一个Conv计算在所有类别上的分数
-        self.final = nn.Conv2d(nb_filters[0], num_classes, kernel_size=1, stride=1)
-
-        if is_attention:
-            self.attention3_1 = AttentionBlock2d(nb_filters[3], nb_filters[4])
-            self.attention2_2 = AttentionBlock2d(nb_filters[2], nb_filters[3])
-            self.attention1_3 = AttentionBlock2d(nb_filters[1], nb_filters[2])
-            self.attention0_4 = AttentionBlock2d(nb_filters[0], nb_filters[1])
-
-    def forward(self, input):
-        # 下采样编码
-        x0_0 = self.conv0_0(input)
-        x1_0 = self.conv1_0(self.down(x0_0))
-        x2_0 = self.conv2_0(self.down(x1_0))
-        x3_0 = self.conv3_0(self.down(x2_0))
-        x4_0 = self.conv4_0(self.down(x3_0))
-
-        if self.is_attention:
-            # 特征融合并进行上采样解码，使用concatenate进行特征融合
-            x3_1 = self.conv3_1(torch.cat([self.attention3_1(x3_0, x4_0), self.up(x4_0)], dim=1))
-            x2_2 = self.conv2_2(torch.cat([self.attention2_2(x2_0, x3_1), self.up(x3_1)], dim=1))
-            x1_3 = self.conv1_3(torch.cat([self.attention1_3(x1_0, x2_2), self.up(x2_2)], dim=1))
-            x0_4 = self.conv0_4(torch.cat([self.attention0_4(x0_0, x1_3), self.up(x1_3)], dim=1))
-        else:
-            # 特征融合并进行上采样解码，使用concatenate进行特征融合
-            x3_1 = self.conv3_1(torch.cat([x3_0, self.up(x4_0)], dim=1))
-            x2_2 = self.conv2_2(torch.cat([x2_0, self.up(x3_1)], dim=1))
-            x1_3 = self.conv1_3(torch.cat([x1_0, self.up(x2_2)], dim=1))
-            x0_4 = self.conv0_4(torch.cat([x0_0, self.up(x1_3)], dim=1))
-
-        # 计算每个类别上的得分
-        output = self.final(x0_4)
-
-        return output
-
-    def __str__(self):
-        ret = 'U-Net'
-        if self.is_recurrent_residual:
-            ret = 'R2U-Net'
-        if self.is_attention:
-            ret = 'Attention ' + ret
-        return ret
diff --git a/LogicImplement/models/unet_resnet34.py b/LogicImplement/models/unet_resnet34.py
deleted file mode 100644
index ca13a22..0000000
--- a/LogicImplement/models/unet_resnet34.py
+++ /dev/null
@@ -1,100 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-# 基于ResNet34实现的UNet，将ResNet34作为UNet的Encoder模块
-import torch
-from torch import nn
-from torch.nn import functional as F
-from torchvision import models
-
-
-class _DecoderBlock(nn.Module):
-    """UNet中的解码模块，标准UNet中的Decoder模块"""
-
-    def __init__(self, in_channels, middle_channels, out_channels=None):
-        super().__init__()
-
-        # 进行2次卷积后，再进行上采样（除了最后的输出的那一层）
-        layers = [
-            nn.Conv2d(in_channels, middle_channels, kernel_size=3),
-            nn.BatchNorm2d(middle_channels),
-            nn.ReLU(inplace=True),
-            nn.Conv2d(middle_channels, middle_channels, kernel_size=3),
-            nn.BatchNorm2d(middle_channels),
-            nn.ReLU(inplace=True)
-        ]
-
-        if out_channels:
-            # 最后进行2倍上采样
-            layers.append(nn.ConvTranspose2d(middle_channels, out_channels, kernel_size=2, stride=2))
-
-        self.decode = nn.Sequential(*layers)
-
-    def forward(self, input):
-        return self.decode(input)
-
-
-class UNet_ResNet34(nn.Module):
-    """使用预训练的ResNet34（整体上分为4个stage，和UNet的下采样过程相似）作为Encoder实现的UNet"""
-
-    def __init__(self, num_classes):
-        super().__init__()
-
-        # 加载预训练的ResNet34
-        resnet34 = models.resnet34(pretrained=True)
-
-        self.enc1 = nn.Sequential(
-            resnet34.conv1,
-            resnet34.bn1,
-            resnet34.relu,
-            resnet34.maxpool,
-            # ResNet34中的layer1并不会进行下采样，所以将其和前面的层合并到一起
-            resnet34.layer1
-        )
-        # 使用ResNet34中主要的4个stage最为UNet的Encoder层
-        self.enc2 = resnet34.layer2
-        self.enc3 = resnet34.layer3
-        self.enc4 = resnet34.layer4
-
-        self.center = _DecoderBlock(512, 1024, 512)
-
-        self.dec4 = _DecoderBlock(1024, 512, 256)
-        self.dec3 = _DecoderBlock(512, 256, 128)
-        self.dec2 = _DecoderBlock(256, 128, 64)
-        # 最后一层不进行上采样与降维
-        self.dec1 = _DecoderBlock(128, 64)
-
-        # 获得每个像素，在每个类别上的得分
-        self.final = nn.Conv2d(64, num_classes, kernel_size=1)
-
-    def forward(self, input):
-        enc1 = self.enc1(input)
-        enc2 = self.enc2(enc1)
-        enc3 = self.enc3(enc2)
-        enc4 = self.enc4(enc3)
-
-        center = self.center(enc4)
-
-        # 进行特征融合与上采样
-        dec4 = self.dec4(self.feature_fusion(enc4, center))
-        dec3 = self.dec3(self.feature_fusion(enc3, dec4))
-        dec2 = self.dec2(self.feature_fusion(enc2, dec3))
-        dec1 = self.dec1(self.feature_fusion(enc1, dec2))
-
-        # 计算每个像素在每个类别上的得分
-        final = self.final(dec1)
-
-        # 最后通过上采样，将feature map还原到输入图像的尺寸
-        return F.upsample_bilinear(final, input.shape[2:])
-
-    def feature_fusion(self, t1, t2):
-        """特征融合，将其中的一个张量进行下采样后（使其与另一个张量保持相同的shape），与另一个张量进行拼接"""
-
-        # 假设输入的张量都是正方形的
-        if t1.shape[-1] >= t2.shape[-1]:
-            max_t = t1
-            min_t = t2
-        else:
-            max_t = t2
-            min_t = t1
-
-        return torch.cat([F.upsample_bilinear(max_t, min_t.shape[2:]), min_t], dim=1)
diff --git a/LogicImplement/models/unet_tiny.py b/LogicImplement/models/unet_tiny.py
deleted file mode 100644
index 013ba8a..0000000
--- a/LogicImplement/models/unet_tiny.py
+++ /dev/null
@@ -1,241 +0,0 @@
-# -*- coding: utf-8 -*-
-
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-import torch
-from torch import nn
-from .network_blocks import DoubleConvBlock, AttentionBlock2d, ResidualRecurrentBlock
-
-class UNet3tiny(nn.Module):
-    """基本的UNet网络，其中的卷积并不会改变张量的尺寸"""
-
-    def __init__(self, num_classes, in_channels, is_attention=False, is_recurrent_residual=False):
-        super().__init__()
-        nb_filters = [32, 64, 128, 256, 512]
-        self.is_attention = is_attention
-        self.is_recurrent_residual = is_recurrent_residual
-
-        # 标准的2倍上采样和下采样，因为没有可以学习的参数，可以共享
-        self.down = nn.MaxPool2d(kernel_size=2, stride=2)
-        self.up = nn.UpsamplingBilinear2d(scale_factor=2)
-
-        if is_recurrent_residual:
-            # 下采样的模块
-            self.conv0_0 = ResidualRecurrentBlock(in_channels, nb_filters[0])
-            self.conv1_0 = ResidualRecurrentBlock(nb_filters[0], nb_filters[1])
-            self.conv2_0 = ResidualRecurrentBlock(nb_filters[1], nb_filters[2])
-            self.conv3_0 = ResidualRecurrentBlock(nb_filters[2], nb_filters[3])
-            self.conv4_0 = ResidualRecurrentBlock(nb_filters[3], nb_filters[4])
-
-            # 上采样的模块
-            self.conv3_1 = ResidualRecurrentBlock(nb_filters[4] + nb_filters[3], nb_filters[3])
-            self.conv2_2 = ResidualRecurrentBlock(nb_filters[3] + nb_filters[2], nb_filters[2])
-            self.conv1_3 = ResidualRecurrentBlock(nb_filters[2] + nb_filters[1], nb_filters[1])
-            self.conv0_4 = ResidualRecurrentBlock(nb_filters[1] + nb_filters[0], nb_filters[0])
-        else:
-            # 下采样的模块
-            self.conv0_0 = DoubleConvBlock(in_channels, nb_filters[0], nb_filters[0])
-            self.conv1_0 = DoubleConvBlock(nb_filters[0], nb_filters[1], nb_filters[1])
-            self.conv2_0 = DoubleConvBlock(nb_filters[1], nb_filters[2], nb_filters[2])
-            self.conv3_0 = DoubleConvBlock(nb_filters[2], nb_filters[3], nb_filters[3])
-
-            # 上采样的模块
-            self.conv2_2 = DoubleConvBlock(nb_filters[3] + nb_filters[2], nb_filters[2], nb_filters[2])
-            self.conv1_3 = DoubleConvBlock(nb_filters[2] + nb_filters[1], nb_filters[1], nb_filters[1])
-            self.conv0_4 = DoubleConvBlock(nb_filters[1] + nb_filters[0], nb_filters[0], nb_filters[0])
-
-        # 最后接一个Conv计算在所有类别上的分数
-        self.final = nn.Conv2d(nb_filters[0], num_classes, kernel_size=1, stride=1)
-
-        if is_attention:
-            self.attention3_1 = AttentionBlock2d(nb_filters[3], nb_filters[4])
-            self.attention2_2 = AttentionBlock2d(nb_filters[2], nb_filters[3])
-            self.attention1_3 = AttentionBlock2d(nb_filters[1], nb_filters[2])
-            self.attention0_4 = AttentionBlock2d(nb_filters[0], nb_filters[1])
-
-    def forward(self, input):
-        # 下采样编码
-        x0_0 = self.conv0_0(input)
-        x1_0 = self.conv1_0(self.down(x0_0))
-        x2_0 = self.conv2_0(self.down(x1_0))
-        x3_0 = self.conv3_0(self.down(x2_0))
-        # x4_0 = self.conv4_0(self.down(x3_0))
-
-        if self.is_attention:
-            # 特征融合并进行上采样解码，使用concatenate进行特征融合
-            x3_1 = self.conv3_1(torch.cat([self.attention3_1(x3_0, x4_0), self.up(x4_0)], dim=1))
-            x2_2 = self.conv2_2(torch.cat([self.attention2_2(x2_0, x3_1), self.up(x3_1)], dim=1))
-            x1_3 = self.conv1_3(torch.cat([self.attention1_3(x1_0, x2_2), self.up(x2_2)], dim=1))
-            x0_4 = self.conv0_4(torch.cat([self.attention0_4(x0_0, x1_3), self.up(x1_3)], dim=1))
-        else:
-            # 特征融合并进行上采样解码，使用concatenate进行特征融合
-            # x3_1 = self.conv3_1(torch.cat([x3_0, self.up(x4_0)], dim=1))
-            x2_2 = self.conv2_2(torch.cat([x2_0, self.up(x3_0)], dim=1))
-            x1_3 = self.conv1_3(torch.cat([x1_0, self.up(x2_2)], dim=1))
-            x0_4 = self.conv0_4(torch.cat([x0_0, self.up(x1_3)], dim=1))
-
-        # 计算每个类别上的得分
-        output = self.final(x0_4)
-
-        return output
-
-    def __str__(self):
-        ret = 'U-Net'
-        if self.is_recurrent_residual:
-            ret = 'R2U-Net'
-        if self.is_attention:
-            ret = 'Attention ' + ret
-        return ret
-
-class UNet2tiny(nn.Module):
-    """基本的UNet网络，其中的卷积并不会改变张量的尺寸"""
-
-    def __init__(self, num_classes, in_channels, is_attention=False, is_recurrent_residual=False):
-        super().__init__()
-        nb_filters = [32, 64, 128, 256, 512]
-        self.is_attention = is_attention
-        self.is_recurrent_residual = is_recurrent_residual
-
-        # 标准的2倍上采样和下采样，因为没有可以学习的参数，可以共享
-        self.down = nn.MaxPool2d(kernel_size=2, stride=2)
-        self.up = nn.UpsamplingBilinear2d(scale_factor=2)
-
-        if is_recurrent_residual:
-            # 下采样的模块
-            self.conv0_0 = ResidualRecurrentBlock(in_channels, nb_filters[0])
-            self.conv1_0 = ResidualRecurrentBlock(nb_filters[0], nb_filters[1])
-            self.conv2_0 = ResidualRecurrentBlock(nb_filters[1], nb_filters[2])
-            self.conv3_0 = ResidualRecurrentBlock(nb_filters[2], nb_filters[3])
-            self.conv4_0 = ResidualRecurrentBlock(nb_filters[3], nb_filters[4])
-
-            # 上采样的模块
-            self.conv3_1 = ResidualRecurrentBlock(nb_filters[4] + nb_filters[3], nb_filters[3])
-            self.conv2_2 = ResidualRecurrentBlock(nb_filters[3] + nb_filters[2], nb_filters[2])
-            self.conv1_3 = ResidualRecurrentBlock(nb_filters[2] + nb_filters[1], nb_filters[1])
-            self.conv0_4 = ResidualRecurrentBlock(nb_filters[1] + nb_filters[0], nb_filters[0])
-        else:
-            # 下采样的模块
-            self.conv0_0 = DoubleConvBlock(in_channels, nb_filters[0], nb_filters[0])
-            self.conv1_0 = DoubleConvBlock(nb_filters[0], nb_filters[1], nb_filters[1])
-            self.conv2_0 = DoubleConvBlock(nb_filters[1], nb_filters[2], nb_filters[2])
-
-
-            # 上采样的模块
-
-            self.conv1_3 = DoubleConvBlock(nb_filters[2] + nb_filters[1], nb_filters[1], nb_filters[1])
-            self.conv0_4 = DoubleConvBlock(nb_filters[1] + nb_filters[0], nb_filters[0], nb_filters[0])
-
-        # 最后接一个Conv计算在所有类别上的分数
-        self.final = nn.Conv2d(nb_filters[0], num_classes, kernel_size=1, stride=1)
-
-        if is_attention:
-            self.attention3_1 = AttentionBlock2d(nb_filters[3], nb_filters[4])
-            self.attention2_2 = AttentionBlock2d(nb_filters[2], nb_filters[3])
-            self.attention1_3 = AttentionBlock2d(nb_filters[1], nb_filters[2])
-            self.attention0_4 = AttentionBlock2d(nb_filters[0], nb_filters[1])
-
-    def forward(self, input):
-        # 下采样编码
-        x0_0 = self.conv0_0(input)
-        x1_0 = self.conv1_0(self.down(x0_0))
-        x2_0 = self.conv2_0(self.down(x1_0))
-        # x3_0 = self.conv3_0(self.down(x2_0))
-        # x4_0 = self.conv4_0(self.down(x3_0))
-
-        if self.is_attention:
-            # 特征融合并进行上采样解码，使用concatenate进行特征融合
-            x3_1 = self.conv3_1(torch.cat([self.attention3_1(x3_0, x4_0), self.up(x4_0)], dim=1))
-            x2_2 = self.conv2_2(torch.cat([self.attention2_2(x2_0, x3_1), self.up(x3_1)], dim=1))
-            x1_3 = self.conv1_3(torch.cat([self.attention1_3(x1_0, x2_2), self.up(x2_2)], dim=1))
-            x0_4 = self.conv0_4(torch.cat([self.attention0_4(x0_0, x1_3), self.up(x1_3)], dim=1))
-        else:
-            # 特征融合并进行上采样解码，使用concatenate进行特征融合
-            # x3_1 = self.conv3_1(torch.cat([x3_0, self.up(x4_0)], dim=1))
-            # x2_2 = self.conv2_2(torch.cat([x2_0, self.up(x3_1)], dim=1))
-            x1_3 = self.conv1_3(torch.cat([x1_0, self.up(x2_0)], dim=1))
-            x0_4 = self.conv0_4(torch.cat([x0_0, self.up(x1_3)], dim=1))
-
-        # 计算每个类别上的得分
-        output = self.final(x0_4)
-
-        return output
-
-    def __str__(self):
-        ret = 'U-Net'
-        if self.is_recurrent_residual:
-            ret = 'R2U-Net'
-        if self.is_attention:
-            ret = 'Attention ' + ret
-        return ret
-
-class UNet1tiny(nn.Module):
-    """基本的UNet网络，其中的卷积并不会改变张量的尺寸"""
-
-    def __init__(self, num_classes, in_channels, is_attention=False, is_recurrent_residual=False):
-        super().__init__()
-        nb_filters = [32, 64, 128, 256, 512]
-        self.is_attention = is_attention
-        self.is_recurrent_residual = is_recurrent_residual
-
-        # 标准的2倍上采样和下采样，因为没有可以学习的参数，可以共享
-        self.down = nn.MaxPool2d(kernel_size=2, stride=2)
-        self.up = nn.UpsamplingBilinear2d(scale_factor=2)
-
-        if is_recurrent_residual:
-            # 下采样的模块
-            self.conv0_0 = ResidualRecurrentBlock(in_channels, nb_filters[0])
-            self.conv1_0 = ResidualRecurrentBlock(nb_filters[0], nb_filters[1])
-            self.conv2_0 = ResidualRecurrentBlock(nb_filters[1], nb_filters[2])
-            self.conv3_0 = ResidualRecurrentBlock(nb_filters[2], nb_filters[3])
-            self.conv4_0 = ResidualRecurrentBlock(nb_filters[3], nb_filters[4])
-
-            # 上采样的模块
-            self.conv3_1 = ResidualRecurrentBlock(nb_filters[4] + nb_filters[3], nb_filters[3])
-            self.conv2_2 = ResidualRecurrentBlock(nb_filters[3] + nb_filters[2], nb_filters[2])
-            self.conv1_3 = ResidualRecurrentBlock(nb_filters[2] + nb_filters[1], nb_filters[1])
-            self.conv0_4 = ResidualRecurrentBlock(nb_filters[1] + nb_filters[0], nb_filters[0])
-        else:
-            # 下采样的模块
-            self.conv0_0 = DoubleConvBlock(in_channels, nb_filters[0], nb_filters[0])
-            self.conv1_0 = DoubleConvBlock(nb_filters[0], nb_filters[1], nb_filters[1])
-
-
-            # 上采样的模块
-
-            self.conv0_4 = DoubleConvBlock(nb_filters[1] + nb_filters[0], nb_filters[0], nb_filters[0])
-
-        # 最后接一个Conv计算在所有类别上的分数
-        self.final = nn.Conv2d(nb_filters[0], num_classes, kernel_size=1, stride=1)
-
-        if is_attention:
-            self.attention3_1 = AttentionBlock2d(nb_filters[3], nb_filters[4])
-            self.attention2_2 = AttentionBlock2d(nb_filters[2], nb_filters[3])
-            self.attention1_3 = AttentionBlock2d(nb_filters[1], nb_filters[2])
-            self.attention0_4 = AttentionBlock2d(nb_filters[0], nb_filters[1])
-
-    def forward(self, input):
-        # 下采样编码
-        x0_0 = self.conv0_0(input)
-        x1_0 = self.conv1_0(self.down(x0_0))
-
-        if self.is_attention:
-            # 特征融合并进行上采样解码，使用concatenate进行特征融合
-            x3_1 = self.conv3_1(torch.cat([self.attention3_1(x3_0, x4_0), self.up(x4_0)], dim=1))
-            x2_2 = self.conv2_2(torch.cat([self.attention2_2(x2_0, x3_1), self.up(x3_1)], dim=1))
-            x1_3 = self.conv1_3(torch.cat([self.attention1_3(x1_0, x2_2), self.up(x2_2)], dim=1))
-            x0_4 = self.conv0_4(torch.cat([self.attention0_4(x0_0, x1_3), self.up(x1_3)], dim=1))
-        else:
-            x0_4 = self.conv0_4(torch.cat([x0_0, self.up(x1_0)], dim=1))
-
-        # 计算每个类别上的得分, x0_4:[16, 32, 128, 128]
-        output = self.final(x0_4)
-
-        return output   # [16, 3, 128, 128]
-
-    def __str__(self):
-        ret = 'U-Net'
-        if self.is_recurrent_residual:
-            ret = 'R2U-Net'
-        if self.is_attention:
-            ret = 'Attention ' + ret
-        return ret
\ No newline at end of file
diff --git a/LogicImplement/models/xnet.py b/LogicImplement/models/xnet.py
deleted file mode 100644
index d48b9c1..0000000
--- a/LogicImplement/models/xnet.py
+++ /dev/null
@@ -1,70 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-import torch
-from torch import nn
-from .network_blocks import SingleConvBlock
-
-
-class XNet(nn.Module):
-    """用于数据量小的X-Ray图像分割的网络，架构上是将2个SegNet拼接了起来，并添加了一些skip connection，
-    https://github.com/JosephPB/XNet，https://arxiv.org/abs/1812.00548
-    """
-
-    def __init__(self, num_classes, in_channels):
-        super().__init__()
-        # TODO: nb_filters需要确认下
-        nb_filters = [32, 64, 128, 256]
-
-        # 标准的2倍上采样和下采样，因为没有可以学习的参数，可以共享
-        self.down = nn.MaxPool2d(kernel_size=2, stride=2)
-        self.up = nn.UpsamplingBilinear2d(scale_factor=2)
-
-        # 第一个SegNet
-        self.conv0 = SingleConvBlock(in_channels, nb_filters[0])
-        self.conv1 = SingleConvBlock(nb_filters[0], nb_filters[1])
-        self.conv2 = SingleConvBlock(nb_filters[1], nb_filters[2])
-        self.conv3 = nn.Sequential(
-            SingleConvBlock(nb_filters[2], nb_filters[3]),
-            SingleConvBlock(nb_filters[3], nb_filters[3])
-        )
-        self.conv4 = SingleConvBlock(nb_filters[3] + nb_filters[2], nb_filters[2])
-        self.conv5 = SingleConvBlock(nb_filters[2] + nb_filters[1], nb_filters[1])
-
-        # 第2个SegNet
-        self.conv6 = SingleConvBlock(nb_filters[1], nb_filters[1])
-        self.conv7 = SingleConvBlock(nb_filters[1], nb_filters[2])
-        self.conv8 = nn.Sequential(
-            SingleConvBlock(nb_filters[2], nb_filters[3]),
-            SingleConvBlock(nb_filters[3], nb_filters[3])
-        )
-        self.conv9 = SingleConvBlock(nb_filters[3] + nb_filters[2], nb_filters[2])
-        self.conv10 = SingleConvBlock(nb_filters[2] + nb_filters[1], nb_filters[1])
-        self.conv11 = SingleConvBlock(nb_filters[1] + nb_filters[0], nb_filters[0])
-
-        # 最后接一个Conv计算在所有类别上的分数
-        self.final = nn.Conv2d(nb_filters[0], num_classes, kernel_size=1, stride=1)
-
-    def forward(self, input):
-        # 第一个SegNet
-        x0 = self.conv0(input)
-        x1 = self.conv1(self.down(x0))
-        x2 = self.conv2(self.down(x1))
-        x3 = self.conv3(self.down(x2))
-        x4 = self.conv4(torch.cat([x2, self.up(x3)], dim=1))
-        x5 = self.conv5(torch.cat([x1, self.up(x4)], dim=1))
-
-        # 第2个SegNet
-        x6 = self.conv6(x5)
-        x7 = self.conv7(self.down(x6))
-        x8 = self.conv8(self.down(x7))
-        x9 = self.conv9(torch.cat([x7, self.up(x8)], dim=1))
-        x10 = self.conv10(torch.cat([x6, self.up(x9)], dim=1))
-        x11 = self.conv11(torch.cat([x0, self.up(x10)], dim=1))
-
-        # 计算每个类别上的得分
-        output = self.final(x11)
-
-        return output
-
-    def __str__(self):
-        return 'X-Net'
diff --git a/LogicImplement/train.py b/LogicImplement/train.py
deleted file mode 100644
index d727f4e..0000000
--- a/LogicImplement/train.py
+++ /dev/null
@@ -1,307 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-import os
-# 
-os.environ["CUDA_VISIBLE_DEVICES"] = "0"
-
-import torch
-import time
-from torch.utils.data import DataLoader
-from torch.utils.tensorboard import SummaryWriter
-from models import unet, nested_unet, loss_function, xnet, nfn_plus, lovasz_losses, unet_tiny
-from datasets import lung
-from torchvision import transforms
-import os
-import argparse
-import numpy as np
-from utils.metrics import compute_metrics
-from utils.tools import create_directory
-
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-# 命令行参数设置
-parser = argparse.ArgumentParser()
-# 模型选择
-parser.add_argument('--model', type=str, default='unet', choices=['unet', 'r2unet', 'attention_unet', 'attention_r2unet', 'nested_unet',
-                                    'xnet', 'nfn_plus'])
-# 数据集选择
-parser.add_argument('--dataset', type=str, default='lung', choices=['lung'])
-# 损失选择
-parser.add_argument('--loss', type=str, default='Dice', choices=['DiceBCE', 'CE', 'SCE', 'Dice', 'Lovasz'])
-# tools.py 增强
-parser.add_argument('--noisy_rate', type=float, choices=[0.2, 0.3, 0.4])
-parser.add_argument('--noisy_type', type=str, choices=['sy', 'asy'])
-
-parser.add_argument('--checkpoint', type=str, default=None)
-parser.add_argument('--gpu', type=str, default='0', choices=['0', '1'])
-parser.add_argument('--parallel', type=str, default='False', choices=['True', 'False'])
-parser.add_argument('--num_workers', type=int, default=0, choices=list(range(17)))
-parser.add_argument('--epoch', type=int, default=300)
-parser.add_argument('--batch_size', type=int, default=2)
-# Learning rate
-parser.add_argument('--lr', type=float, default=1e-4)
-# 学习率衰减
-parser.add_argument('--weight_decay', type=float, default=0)
-parser.add_argument('--print_frequency', type=int, default=4)
-parser.add_argument('--save_frequency', type=int, default=4)
-args = parser.parse_args()
-
-# 其他准备
-BASE_PATH = r'C:\Users\29185\Desktop\大创-肺部疾病检测\TianChiCTSeg'
-format = '{}_{}_{}'.format(args.dataset, args.model, args.loss)
-# ？
-loss_weights = False
-if loss_weights:
-    format += "_unet_dice_debug"
-
-log_path = os.path.join(BASE_PATH, 'log', format)
-os.makedirs(log_path, exist_ok=True)
-checkpoint_path_prefix = os.path.join(BASE_PATH, 'checkpoint', format)
-os.makedirs(checkpoint_path_prefix, exist_ok=True)
-
-
-# DEVICE = 'cuda'
-DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-
-# 加载数据
-print('Loading data...')
-# 选择数据集
-if args.dataset == 'lung':
-    dataset = lung.Lung
-else:
-    print('数据集异常')
-    pass
-
-# 对image和mask进行resize
-transform = transforms.Compose([transforms.ToTensor()])
-target_transform = transforms.Compose([transforms.ToTensor()])
-
-# noisy_chaos可以设置噪声率和噪声类型
-if args.dataset == 'lung':
-    train_data = dataset(mode='train', transform=transform, target_transform=target_transform,
-                         BASE_PATH=r"C:\Users\29185\Desktop\大创-肺部疾病检测\TianChiCTSeg\data\train")
-    val_data = dataset(mode='val', transform=transform, target_transform=target_transform,
-                       BASE_PATH=r"C:\Users\29185\Desktop\大创-肺部疾病检测\TianChiCTSeg\data\val")
-
-train_loader = DataLoader(dataset=train_data, batch_size=args.batch_size, shuffle=True,
-                          num_workers=args.num_workers, pin_memory=True)
-val_loader = DataLoader(dataset=val_data, batch_size=args.batch_size, shuffle=False,
-                        num_workers=args.num_workers, pin_memory=True)
-print('Create model...')
-
-# 选择网络模型
-if args.model == 'unet':
-    net = unet.UNet(num_classes=dataset.NUM_CLASSES, in_channels=dataset.CHANNELS_NUM)
-elif args.model == 'r2unet':
-    net = unet.UNet(num_classes=dataset.NUM_CLASSES, in_channels=dataset.CHANNELS_NUM, is_recurrent_residual=True)
-elif args.model == 'attention_unet':
-    net = unet.UNet(num_classes=dataset.NUM_CLASSES, in_channels=dataset.CHANNELS_NUM, is_attention=True)
-elif args.model == 'attention_r2unet':
-    net = unet.UNet(num_classes=dataset.NUM_CLASSES, in_channels=dataset.CHANNELS_NUM, is_attention=True,
-                    is_recurrent_residual=True)
-elif args.model == 'nested_unet':
-    net = nested_unet.NestedUNet(num_classes=dataset.NUM_CLASSES, in_channels=dataset.CHANNELS_NUM)
-elif args.model == 'xnet':
-    net = xnet.XNet(num_classes=dataset.NUM_CLASSES, in_channels=dataset.CHANNELS_NUM)
-elif args.model == 'nfn_plus':
-    net = nfn_plus.NFNPlus(num_classes=dataset.NUM_CLASSES, in_channels=dataset.CHANNELS_NUM)
-elif args.model == 'unet2tiny':
-    net = unet_tiny.UNet2tiny(num_classes=dataset.NUM_CLASSES, in_channels=dataset.CHANNELS_NUM)
-elif args.model == 'unet3tiny':
-    net = unet_tiny.UNet3tiny(num_classes=dataset.NUM_CLASSES, in_channels=dataset.CHANNELS_NUM)
-elif args.model == 'unet1tiny':
-    net = unet_tiny.UNet1tiny(num_classes=dataset.NUM_CLASSES, in_channels=dataset.CHANNELS_NUM)
-
-# 设置优化方法和损失函数
-# optimizer = torch.optim.SGD(net.parameters(), lr=args.lr, momentum=0.9, weight_decay=args.weight_decay)
-optimizer = torch.optim.Adam(net.parameters(), lr=args.lr)
-
-
-# 选择损失函数
-if args.loss == 'DiceBCE':
-    criterion = loss_function.DiceAndBCELoss(dataset.NUM_CLASSES)
-elif args.loss == 'CE':
-    criterion = torch.nn.CrossEntropyLoss()
-elif args.loss == 'SCE':
-    criterion = loss_function.SCELoss(dataset.NUM_CLASSES, alpha=1, beta=1)
-elif args.loss == 'Dice':
-    criterion = loss_function.SoftDiceLoss(dataset.NUM_CLASSES)
-elif args.loss == 'Lovasz':
-    criterion = loss_function.LovaszLoss()
-
-
-print('<================== Parameters ==================>')
-print('model: {}'.format(net))
-print('dataset: {}(training={}, validation={})'.format(train_data, len(train_data), len(val_data)))
-print('batch_size: {}'.format(args.batch_size))
-print('batch_num: {}'.format(len(train_loader)))
-print('epoch: {}'.format(args.epoch))
-print('loss_function: {}'.format(criterion))
-print('optimizer: {}'.format(optimizer))
-print('tensorboard_log_path: {}'.format(log_path))
-print('<================================================>')
-
-
-# 判断是否使用多GPU运行
-if args.parallel == 'True':
-    print('Use DataParallel.')
-    net = torch.nn.DataParallel(net)
-# GPU or CPU
-net = net.to(device)
-
-start_epoch = 0
-temp = 0
-# 加载模型
-if args.checkpoint is not None:
-    checkpoint_data = torch.load(args.checkpoint)
-    print('**** Load model and optimizer data from {} ****'.format(args.checkpoint))
-
-    # 加载模型和优化器的数据
-    net.load_state_dict(checkpoint_data['model_state_dict'])
-    optimizer.load_state_dict(checkpoint_data['optimizer_state_dict'])
-
-    # 加载上次训练的最后一个epoch和打印的最后一个temp，这里作为起点，需要在之前的基础上加1
-    start_epoch = checkpoint_data['epoch'] + 1
-    temp = checkpoint_data['temp'] + 1
-
-    args.epoch += start_epoch
-    # temp = (len(train_loader) // args.print_frequency) * start_epoch + 1
-
-else:
-    # 如果重新开始训练，则删除原来的log并新建
-    create_directory(log_path)
-
-writer = SummaryWriter(log_dir=log_path, flush_secs=30)
-# 训练与验证的过程
-print('Start training...')
-for epoch in range(start_epoch, args.epoch):
-    # 训练
-    loss_all = []
-    predictions_all = []
-    labels_all = []
-    print('-------------------------------------- Training {} --------------------------------------'.format(epoch + 1))
-    net.train()
-    for index, data in enumerate(train_loader):
-        inputs, labels = data
-        inputs, labels = inputs.to(device), labels.to(device)
-        optimizer.zero_grad()
-        outputs = net(inputs)   #
-        loss = 0
-        # 如果使用deep supervision，返回1个list（包含多个输出），计算每个输出的loss，最后求平均
-        if isinstance(outputs, list):
-            for out in outputs:
-                loss += criterion(out, labels.long())
-            loss /= len(outputs)
-        else:
-            loss = criterion(outputs, labels.long())
-        # 计算在该批次上的平均损失函数
-        loss /= inputs.size(0)
-
-        # 更新网络参数
-        loss.backward()
-        optimizer.step()
-
-        loss_all.append(loss.item())
-
-        if isinstance(outputs, list):
-            # 若使用deep supervision，用最后的输出来进行预测
-            predictions = torch.max(outputs[-1], dim=1)[1].cpu().numpy().astype(np.int)
-        else:
-            # 将概率最大的类别作为预测的类别
-            predictions = torch.max(outputs, dim=1)[1].cpu().numpy().astype(np.int)
-
-        labels = labels.cpu().numpy().astype(np.int)
-
-        predictions_all.append(predictions)
-        labels_all.append(labels)
-
-        if (index + 1) % args.print_frequency == 0:
-            # 计算打印间隔的平均损失函数
-            avg_loss = np.mean(loss_all)
-            loss_all = []
-
-            writer.add_scalar('train/loss', avg_loss, temp)
-            temp += 1
-
-            print("Training: Epoch[{:0>3}/{:0>3}] Iteration[{:0>3}/{:0>3}] Loss: {:.4f} ".format(
-                epoch + 1, args.epoch, index + 1, len(train_loader), avg_loss))
-
-    # 使用混淆矩阵计算语义分割中的指标
-    iou, miou, dsc, mdsc, ac, pc, mpc, se, mse, sp, msp, f1, mf1 = compute_metrics(predictions_all, labels_all,
-                                                                                   dataset.NUM_CLASSES)
-
-    writer.add_scalars('train/metrics', dict(miou=miou, mdsc=mdsc, mpc=mpc, ac=ac, mse=mse, msp=msp, mf1=mf1), epoch)
-
-    print('Training: MIoU: {:.4f}, MDSC: {:.4f}, MPC: {:.4f}, AC: {:.4f}, MSE: {:.4f}, MSP: {:.4f}, MF1: {:.4f}'.format(
-        miou, mdsc, mpc, ac, mse, msp, mf1
-    ))
-
-    # 验证
-    loss_all = []
-    predictions_all = []
-    labels_all = []
-
-    print('-------------------------------------- Validation {} ------------------------------------'.format(epoch + 1))
-
-    net.eval()
-    with torch.no_grad():
-        for _, data in enumerate(val_loader):
-            inputs, labels = data
-            inputs, labels = inputs.to(device), labels.to(device)
-
-            outputs = net(inputs)
-
-            loss = 0
-            # 如果使用deep supervision，返回1个list（包含多个输出），计算每个输出的loss，最后求平均
-            if isinstance(outputs, list):
-                for out in outputs:
-                    loss += criterion(out, labels.long())
-                loss /= len(outputs)
-            else:
-                loss = criterion(outputs, labels.long())
-            # 计算在该批次上的平均损失函数
-            loss /= inputs.size(0)
-
-            loss_all.append(loss.item())
-
-            if isinstance(outputs, list):
-                # 若使用deep supervision，用最后一个输出来进行预测
-                predictions = torch.max(outputs[-1], dim=1)[1].cpu().numpy().astype(np.int)
-            else:
-                # 将概率最大的类别作为预测的类别
-                predictions = torch.max(outputs, dim=1)[1].cpu().numpy().astype(np.int)
-            labels = labels.cpu().numpy().astype(np.int)
-
-            predictions_all.append(predictions)
-            labels_all.append(labels)
-
-    # 使用混淆矩阵计算语义分割中的指标
-    iou, miou, dsc, mdsc, ac, pc, mpc, se, mse, sp, msp, f1, mf1 = compute_metrics(predictions_all, labels_all,
-                                                                                   dataset.NUM_CLASSES)
-    avg_loss = np.mean(loss_all)
-
-    writer.add_scalar('val/loss', avg_loss, epoch)
-    writer.add_scalars('val/metrics', dict(miou=miou, mdsc=mdsc, mpc=mpc, ac=ac, mse=mse, msp=msp, mf1=mf1), epoch)
-
-    # 绘制每个类别的IoU
-    temp_dict = {'miou': miou}
-    for i in range(dataset.NUM_CLASSES):
-        temp_dict['class{}'.format(i)] = iou[i]
-    writer.add_scalars('val/class_iou', temp_dict, epoch)
-
-    print('Training: MIoU: {:.4f}, MDSC: {:.4f}, MPC: {:.4f}, AC: {:.4f}, MSE: {:.4f}, MSP: {:.4f}, MF1: {:.4f}'.format(
-        miou, mdsc, mpc, ac, mse, msp, mf1
-    ))
-
-    # 保存模型参数和优化器参数
-    if (epoch + 1) % args.save_frequency == 0:
-        checkpoint_path = '{}_{}_{}.pkl'.format(format, time.strftime('%m%d_%H%M', time.localtime()), epoch)
-        # save_checkpoint_path = checkpoint_path_prefix + '/' + checkpoint_path
-        save_checkpoint_path = os.path.join(checkpoint_path_prefix, checkpoint_path)
-        torch.save({
-            'is_parallel': args.parallel,
-            'epoch': epoch,
-            'temp': temp,
-            'model_state_dict': net.state_dict(),
-            'optimizer_state_dict': optimizer.state_dict()},
-            save_checkpoint_path)
-        print('Save model at {}.'.format(save_checkpoint_path))
diff --git a/LogicImplement/utils/__init__.py b/LogicImplement/utils/__init__.py
deleted file mode 100644
index 67f768c..0000000
--- a/LogicImplement/utils/__init__.py
+++ /dev/null
@@ -1,2 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
diff --git a/LogicImplement/utils/__pycache__/__init__.cpython-38.pyc b/LogicImplement/utils/__pycache__/__init__.cpython-38.pyc
deleted file mode 100644
index 4d3d8f5..0000000
Binary files a/LogicImplement/utils/__pycache__/__init__.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/utils/__pycache__/metrics.cpython-38.pyc b/LogicImplement/utils/__pycache__/metrics.cpython-38.pyc
deleted file mode 100644
index 82e4f0c..0000000
Binary files a/LogicImplement/utils/__pycache__/metrics.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/utils/__pycache__/tools.cpython-38.pyc b/LogicImplement/utils/__pycache__/tools.cpython-38.pyc
deleted file mode 100644
index 4d75706..0000000
Binary files a/LogicImplement/utils/__pycache__/tools.cpython-38.pyc and /dev/null differ
diff --git a/LogicImplement/utils/metrics.py b/LogicImplement/utils/metrics.py
deleted file mode 100644
index 5b1ed7a..0000000
--- a/LogicImplement/utils/metrics.py
+++ /dev/null
@@ -1,76 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-import numpy as np
-
-
-def compute_confusion_matrix(predictions_all, labels_all, num_classes):
-    """根据每批数据的预测和原始的label，计算混淆矩阵"""
-
-    confusion_matrix = np.zeros((num_classes, num_classes), dtype=np.int)
-    for predictions, labels in zip(predictions_all, labels_all):
-        predictions = predictions.flatten()
-        labels = labels.flatten()
-
-        # np.bincount统计每个值的出现次数
-        temp = np.bincount(num_classes * labels + predictions, minlength=num_classes ** 2).reshape(
-            (num_classes, num_classes))
-
-        confusion_matrix += temp
-
-    return confusion_matrix
-
-
-def compute_metrics(predictions_all, labels_all, num_classes):
-    """计算语义分割中的指标
-    https://github.com/ZijunDeng/pytorch-semantic-segmentation/blob/master/utils/misc.py，
-    https://github.com/LeeJunHyun/Image_Segmentation,
-    https://www.cnblogs.com/Trevo/p/11795503.html，
-    https://www.aiuai.cn/aifarm1330.html"""
-
-    confusion_matrix = compute_confusion_matrix(predictions_all, labels_all, num_classes)
-
-    # 每个类别的交并比和平均的交并比
-    # IoU=TP/(TP+FP+FN)
-    iou = np.diag(confusion_matrix) / (
-            confusion_matrix.sum(axis=1) + confusion_matrix.sum(axis=0) - np.diag(confusion_matrix))
-    miou = np.nanmean(iou)
-
-    # 每个类别的Dice相似系数和平均的Dice相似系数
-    # DSC=2*TP/(TP+FN+TP+FP)
-    dsc = 2 * np.diag(confusion_matrix) / (confusion_matrix.sum(axis=1) + confusion_matrix.sum(axis=0))
-    mdsc = np.nanmean(dsc)
-
-    # 所有像素总的准确率、每个类别的像素准确率、平均的像素准确率
-    # AC=(TP+TN)/(TP+TN+FP+FN)
-    # PC=TP/(TP+FP)
-    ac = np.diag(confusion_matrix).sum() / confusion_matrix.sum()
-    pc = np.diag(confusion_matrix) / confusion_matrix.sum(axis=0)
-    mpc = np.nanmean(pc)
-
-    # 敏感性，Sensitivity
-    # SE=TP/(TP+FN)
-    se = np.diag(confusion_matrix) / confusion_matrix.sum(axis=1)
-    mse = np.nanmean(se)
-
-    # 特异性，Specificity
-    # SP=TN/(TN+FP)
-    sp = [(np.diag(confusion_matrix).sum() - confusion_matrix[i, i]) /
-          (np.diag(confusion_matrix).sum() + confusion_matrix.sum(axis=0)[i] - 2 * confusion_matrix[i, i])
-          for i in range(len(confusion_matrix))]
-    msp = np.nanmean(sp)
-
-    # F1分数，F1 score
-    # f1=2*PC*SE/(PC+SE)
-    f1 = 2 * pc * se / (pc + se)
-    mf1 = np.nanmean(f1)
-
-    print('Confusion Matrix:')
-    print(confusion_matrix)
-    print('IoU: {}'.format(iou))
-    print('DSC: {}'.format(dsc))
-    print('PC: {}'.format(pc))
-    print('SE: {}'.format(se))
-    print('SP: {}'.format(sp))
-    print('F1: {}'.format(f1))
-
-    return iou, miou, dsc, mdsc, ac, pc, mpc, se, mse, sp, msp, f1, mf1
diff --git a/LogicImplement/utils/tools.py b/LogicImplement/utils/tools.py
deleted file mode 100644
index 40befbd..0000000
--- a/LogicImplement/utils/tools.py
+++ /dev/null
@@ -1,160 +0,0 @@
-#!/usr/bin/python3
-# -*- coding: utf-8 -*
-from torch import nn
-import json
-import os
-import torch
-import shutil
-import numpy as np
-from PIL import ImageDraw, Image
-import random
-
-
-def generate_noisy_indexes(seed, dataset_size, noisy_rate):
-    """生成需要加噪的图像的index"""
-
-    num = int(dataset_size * noisy_rate)
-    random.seed(seed)
-    indexes = random.sample(range(0, dataset_size), num)
-    return indexes
-
-
-def add_noisy_triangle_label(mask):
-    """在mask上添加噪声标注"""
-
-    while True:
-        noisy_image = Image.fromarray(np.zeros(mask.size, dtype=np.uint8))
-        # 生成随机的三角形的顶点坐标
-        noisy_points = np.random.randint(0, 256, (3, 2))
-        # 三角形的跨度不能太大
-        if (noisy_points.max(axis=0) - noisy_points.min(axis=0)).max() > 75:
-            continue
-
-        noisy_points = noisy_points.flatten()
-        noisy_points = tuple(noisy_points)
-        ImageDraw.Draw(noisy_image).polygon(noisy_points, fill=255)
-
-        # 噪声标签的面积不能太大
-        noisy_array = np.asarray(noisy_image) / 255
-        if noisy_array.sum() <= 1500 and noisy_array.sum() >= 1000:
-            # 在原图上添加噪声
-            ImageDraw.Draw(mask).polygon(noisy_points, fill=255)
-            break
-
-    return mask
-
-
-def create_directory(dir_path):
-    """创建目录，若存在，则删除后再创建"""
-
-    if os.path.exists(dir_path):
-        shutil.rmtree(dir_path)
-        print('Remove {}'.format(dir_path))
-    os.mkdir(dir_path)
-    print('Create {}'.format(dir_path))
-
-
-def process_binary_mask_tensor(mask):
-    """对生成的Tensor类型的mask进行处理"""
-
-    assert isinstance(mask, torch.Tensor), '输入的不是Tensor类型'
-
-    # ToTensor()会自动给数据添加一个channel维度，因为是类别标签，不需要这个维度
-    if len(mask.shape) == 3:
-        mask = torch.squeeze(mask, dim=0)
-
-    mask = mask.to(torch.uint8)
-    mask[mask != 0] = 1
-
-    return mask
-
-
-def process_multiple_mask_tensor(mask, color_map):
-    """对生成的Tensor类型的mask进行处理"""
-
-    assert isinstance(mask, torch.Tensor), '输入的不是Tensor类型'
-
-    # ToTensor()会自动给数据添加一个channel维度，因为是类别标签，不需要这个维度
-    if len(mask.shape) == 3:
-        mask = torch.squeeze(mask, dim=0)
-
-    # 如果mask被Resize，修改其中生成的噪声标签，并将数据类型转换为uint8，将其中的255替换为1
-    if torch.max(mask).item() <= 1:
-        # ToTensor()会将数据进行max-min归一化，这里进行还原
-        mask = (mask * 255).to(torch.uint8)
-
-        # 记录噪声标记的坐标
-        noise_indexes = torch.ones_like(mask, dtype=torch.bool)
-
-        # 去除噪声标记
-        for color_value, class_id in color_map.items():
-            noise_indexes = noise_indexes & (mask != color_value)
-        mask[noise_indexes] = 0
-
-        # 将对应的类别的颜色值替换为其类别id
-        for color_value, class_id in color_map.items():
-            mask[mask == color_value] = class_id
-
-        # 因为Resize使用线性插值的缘故，会产生一些少量的现有类别的标记，但是其数量很少，将其过滤掉
-        for class_id in color_map.values():
-            if torch.sum(mask == class_id).item() <= 10:
-                mask[mask == class_id] = 0
-
-    return mask
-
-
-def is_augmented(path):
-    """从路径判断是否是被增广的数据"""
-
-    augmented_types = ['VerticalFlip', 'HorizontalFlip', 'Transpose', 'RandomRotate90']
-    for t in augmented_types:
-        if path.__contains__(t):
-            return True
-    return False
-
-
-def remove_augmented_item(old_items):
-    """删除被数据增强的数据"""
-
-    new_items = []
-    for item in old_items:
-        if not is_augmented(item['image_path']):
-            new_items.append(item)
-    return new_items
-
-
-def make_dataset(mode, base_path, is_contain_augmented_data):
-    """从指定目录下读取用于training、validation的json文件"""
-
-    assert mode in ['train', 'val']
-
-    path = os.path.join(base_path, mode)
-    image_path = os.path.join(path, "image")
-    mask_path = os.path.join(path, "mask")
-
-    image_list = os.listdir(image_path)
-    mask_list = os.listdir(mask_path)
-    return image_list, mask_list
-
-    # json_path = os.path.join(base_path, '{}.json'.format(mode))
-    #
-    # with open(json_path, 'r') as f:
-    #     items = json.load(f)
-    #
-    # if not is_contain_augmented_data and mode == 'train':
-    #     items = remove_augmented_item(items)
-    #
-    # return items
-
-
-def initialize_weights(model):
-    """初始化神经网络中的参数"""
-
-    for module in model.modules():
-        if isinstance(module, nn.Conv2d) or isinstance(module, nn.Linear):
-            nn.init.kaiming_normal_(module.weight)
-            if module.bias is not None:
-                module.bias.data.zero_()
-        elif isinstance(module, nn.BatchNorm2d):
-            module.weight.data.fill_(1)
-            module.bias.data.zero_()