create_seg-dataset_with-aug.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 188,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 189,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import json "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "====================================================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 基本配置"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 191,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "deleting... Done!\n"
     ]
    }
   ],
   "source": [
    "dataset_path = '../pancreas/backup1-seg-dataset/'\n",
    "new_dataset_path = '../pancreas/tumor-seg-dataset/'\n",
    "\n",
    "if os.path.exists(new_dataset_path):\n",
    "    import shutil\n",
    "    shutil.rmtree(new_dataset_path, ignore_errors=True)\n",
    "    print(\"deleting... Done!\")\n",
    "    \n",
    "os.mkdir(new_dataset_path)\n",
    "os.mkdir(os.path.join(new_dataset_path,'imgs'))\n",
    "os.mkdir(os.path.join(new_dataset_path,'masks'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 192,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['AILIKEMU_AP_1.CT.0003.0023.2018.01.06.08.58.26.453125.106165917',\n",
       " 'AILIKEMU_AP_1.CT.0003.0033.2018.01.06.08.58.26.453125.106166368',\n",
       " 'AILIKEMU_AP_1.CT.0003.0029.2018.01.06.08.58.26.453125.106166188',\n",
       " 'AILIKEMU_AP_1.CT.0003.0022.2018.01.06.08.58.26.453125.106165872',\n",
       " 'AILIKEMU_AP_1.CT.0003.0032.2018.01.06.08.58.26.453125.106166323',\n",
       " 'AILIKEMU_AP_1.CT.0003.0024.2018.01.06.08.58.26.453125.106165962',\n",
       " 'AILIKEMU_AP_1.CT.0003.0028.2018.01.06.08.58.26.453125.106166143',\n",
       " 'AILIKEMU_AP_1.CT.0003.0034.2018.01.06.08.58.26.453125.106166413',\n",
       " 'AILIKEMU_AP_1.CT.0003.0031.2018.01.06.08.58.26.453125.106166278',\n",
       " 'AILIKEMU_AP_1.CT.0003.0030.2018.01.06.08.58.26.453125.106166233',\n",
       " 'AILIKEMU_AP_1.CT.0003.0026.2018.01.06.08.58.26.453125.106166053',\n",
       " 'AILIKEMU_AP_1.CT.0003.0027.2018.01.06.08.58.26.453125.106166098',\n",
       " 'AILIKEMU_AP_1.CT.0003.0025.2018.01.06.08.58.26.453125.106166008']"
      ]
     },
     "execution_count": 192,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img_names = [name[:-5] for name in os.listdir(os.path.join(dataset_path,'tumor'))]\n",
    "img_names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# imgs_path = os.path.join(dataset_path,'imgs')\n",
    "# names = [name[:-4] for name in os.listdir(imgs_path)]\n",
    "\n",
    "# command = 'rm -f '\n",
    "# for name in names:\n",
    "#     #print(name)\n",
    "#     if name not in img_names:\n",
    "#         print('deleting: ',name)\n",
    "#         os.popen(command + os.path.join(imgs_path,name)+'.jpg')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 从labelme的json中制作mask并生成数据集"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>最后用于训练的数据集中，index和原来文件名的对应关系<br>\n",
    "    json_names数组依次对应index:0~13"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(['AILIKEMU_AP_1.CT.0003.0023.2018.01.06.08.58.26.453125.106165917.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0033.2018.01.06.08.58.26.453125.106166368.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0029.2018.01.06.08.58.26.453125.106166188.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0022.2018.01.06.08.58.26.453125.106165872.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0032.2018.01.06.08.58.26.453125.106166323.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0024.2018.01.06.08.58.26.453125.106165962.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0028.2018.01.06.08.58.26.453125.106166143.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0034.2018.01.06.08.58.26.453125.106166413.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0031.2018.01.06.08.58.26.453125.106166278.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0030.2018.01.06.08.58.26.453125.106166233.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0026.2018.01.06.08.58.26.453125.106166053.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0027.2018.01.06.08.58.26.453125.106166098.json',\n",
       "  'AILIKEMU_AP_1.CT.0003.0025.2018.01.06.08.58.26.453125.106166008.json'],\n",
       " '~~~',\n",
       " '0-12')"
      ]
     },
     "execution_count": 193,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "json_path = os.path.join(dataset_path,'tumor')\n",
    "json_names = os.listdir(json_path)\n",
    "json_names,'~~~','0-%d'%(len(json_names)-1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>而index和训练顺序的关系为 nm\n",
    "   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "metadata": {},
   "outputs": [],
   "source": [
    "#os.listdir('../pancreas/backup1-seg-dataset/imgs/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0023.2018.01.06.08.58.26.453125.106165917.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0033.2018.01.06.08.58.26.453125.106166368.json\n",
      "mask: 1\n",
      "AILIKEMU_AP_1.CT.0003.0033.2018.01.06.08.58.26.453125.106166368.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0029.2018.01.06.08.58.26.453125.106166188.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0022.2018.01.06.08.58.26.453125.106165872.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0032.2018.01.06.08.58.26.453125.106166323.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0024.2018.01.06.08.58.26.453125.106165962.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0028.2018.01.06.08.58.26.453125.106166143.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0034.2018.01.06.08.58.26.453125.106166413.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0031.2018.01.06.08.58.26.453125.106166278.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0030.2018.01.06.08.58.26.453125.106166233.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0026.2018.01.06.08.58.26.453125.106166053.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0027.2018.01.06.08.58.26.453125.106166098.json\n",
      "mask: 0\n",
      "AILIKEMU_AP_1.CT.0003.0025.2018.01.06.08.58.26.453125.106166008.json\n"
     ]
    }
   ],
   "source": [
    "# https://blog.csdn.net/weixin_39975744/article/details/111515090\n",
    "imgs_path = os.path.join(dataset_path,'imgs')\n",
    "json_path = os.path.join(dataset_path,'tumor')\n",
    "json_names = os.listdir(json_path)\n",
    "\n",
    "index = 0\n",
    "for name in json_names:\n",
    "    with open(os.path.join(json_path,name), \"r\") as f:\n",
    "        label = json.load(f)\n",
    "        img = cv2.imread(os.path.join(imgs_path,name[:-5]+'.jpg'),0)\n",
    "        img[0:33,0:245] = 0\n",
    "        h, w = img.shape[:2]\n",
    "        \n",
    "        shapes = label[\"shapes\"]\n",
    "        mask = np.zeros([h, w, 1], np.uint8) # 创建一个大小和原图相同的空白图像\n",
    "        for i,shape in enumerate(shapes):\n",
    "            '''\n",
    "                就应该一个json的所有区域 一起 作为一个mask，\n",
    "                因为它们可能都是标注，只是不连通而已。\n",
    "            '''\n",
    "            print('mask: %d' %i)\n",
    "            print(name)\n",
    "            \n",
    "            category = shape[\"label\"]\n",
    "\n",
    "            points = shape[\"points\"]\n",
    "            points_array = np.array(points, dtype=np.int32)\n",
    "            mask = cv2.fillPoly(mask, [points_array], 255) #category_types.index(category)\n",
    "            \n",
    "            #print(mask.shape,img.shape)\n",
    "        cv2.imwrite(os.path.join(os.path.join(new_dataset_path,'imgs'),str(index)+'.jpg'),img)\n",
    "        cv2.imwrite(os.path.join(os.path.join(new_dataset_path,'masks'),str(index)+'_mask'+'.jpg'),mask)\n",
    "        index+=1\n",
    "            #plt.imshow(mask,cmap='gray')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# 移动生成的数据集到所有UNet代码目录(即移动到各个衍生目录)\n",
    "import glob\n",
    "UNet_path = glob.glob('/GPUFS/sysu_rxwang_1/defeng/*UNet*')\n",
    "\n",
    "for path in UNet_path:\n",
    "    if os.path.exists(os.path.join(path,'data','imgs')):\n",
    "        import shutil\n",
    "        shutil.rmtree(os.path.join(path,'data','imgs'), ignore_errors=True)\n",
    "        print(\"deleting... Done!\")\n",
    "\n",
    "    if os.path.exists(os.path.join(path,'data','masks')):\n",
    "        import shutil\n",
    "        shutil.rmtree(os.path.join(path,'data','masks'), ignore_errors=True)\n",
    "        print(\"deleting... Done!\")\n",
    "        \n",
    "# for path in UNet_path:\n",
    "#     if os.path.exists(os.path.join(path,'imgs')):\n",
    "#         import shutil\n",
    "#         shutil.rmtree(os.path.join(path,'imgs'), ignore_errors=True)\n",
    "#         print(\"deleting... Done!\")\n",
    "\n",
    "#     if os.path.exists(os.path.join(path,'masks')):\n",
    "#         import shutil\n",
    "#         shutil.rmtree(os.path.join(path,'masks'), ignore_errors=True)\n",
    "#         print(\"deleting... Done!\")\n",
    "\n",
    "\n",
    "    os.popen('cp -r %s %s'%(os.path.join(new_dataset_path,'imgs'),os.path.join(path,'data','imgs')))\n",
    "    os.popen('cp -r %s %s'%(os.path.join(new_dataset_path,'masks'),os.path.join(path,'data','masks')))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[]"
      ]
     },
     "execution_count": 197,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import glob\n",
    "\n",
    "glob.glob('/GPUFS/sysu_rxwang_1/defeng/*UNet*')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 对2得到的数据集进行data augmentation(elastic&affine)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "metadata": {},
   "outputs": [],
   "source": [
    "def affine_transform(image, alpha_affine=10, random_state=None):\n",
    "    if random_state is None:\n",
    "        random_state = np.random.RandomState(None)\n",
    "\n",
    "    shape = image.shape\n",
    "    shape_size = shape[:2]\n",
    "\n",
    "    center_square = np.float32(shape_size) // 2\n",
    "    square_size = min(shape_size) // 3\n",
    "    pts1 = np.float32([  #以图片中心为原点，长度为min(shape_size)//3 采集了3个采样点。\n",
    "        center_square + square_size,  \n",
    "        [center_square[0] + square_size, center_square[1] - square_size],\n",
    "        center_square - square_size])\n",
    "    #print(pts1)\n",
    "    pts2 = pts1 + random_state.uniform(-alpha_affine, alpha_affine, size=pts1.shape).astype(np.float32)\n",
    "\n",
    "    M = cv2.getAffineTransform(pts1, pts2)\n",
    "    distorted_image = cv2.warpAffine(\n",
    "        image, M, shape_size[::-1], borderMode=cv2.BORDER_REPLICATE) #cv2.BORDER_REPLICATE / cv2.BORDER_REFLECT_101)\n",
    "\n",
    "    return distorted_image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {},
   "outputs": [],
   "source": [
    "from scipy.ndimage.interpolation import map_coordinates\n",
    "from scipy.ndimage.filters import gaussian_filter\n",
    "\n",
    "def elastic_transform(image, alpha, sigma, affine_flag=False,alpha_affine=10, random_state=None):\n",
    "    \"\"\"Elastic deformation of images as described in [Simard2003]_ (with modifications).\n",
    "    .. [Simard2003] Simard, Steinkraus and Platt, \"Best Practices for\n",
    "         Convolutional Neural Networks applied to Visual Document Analysis\", in\n",
    "         Proc. of the International Conference on Document Analysis and\n",
    "         Recognition, 2003.\n",
    "\n",
    "     Based on https://gist.github.com/erniejunior/601cdf56d2b424757de5\n",
    "    \"\"\"\n",
    "    if random_state is None:\n",
    "        random_state = np.random.RandomState(None)\n",
    "\n",
    "    shape = image.shape\n",
    "    shape_size = shape[:2]\n",
    "    \n",
    "    # Random affine\n",
    "#     if affine_flag:\n",
    "#         center_square = np.float32(shape_size) // 2\n",
    "#         square_size = min(shape_size) // 3\n",
    "#         pts1 = np.float32([center_square + square_size, [center_square[0]+square_size, center_square[1]-square_size], center_square - square_size])\n",
    "#         pts2 = pts1 + random_state.uniform(-alpha_affine, alpha_affine, size=pts1.shape).astype(np.float32)\n",
    "#         M = cv2.getAffineTransform(pts1, pts2)\n",
    "#         image = cv2.warpAffine(image, M, shape_size[::-1], borderMode=cv2.BORDER_REFLECT_101)\n",
    "\n",
    "    dx = gaussian_filter((random_state.rand(*shape) * 2 - 1), sigma) * alpha\n",
    "    dy = gaussian_filter((random_state.rand(*shape) * 2 - 1), sigma) * alpha\n",
    "    dz = np.zeros_like(dx)\n",
    "\n",
    "    x, y, z = np.meshgrid(np.arange(shape[1]), np.arange(shape[0]), np.arange(shape[2]))\n",
    "    indices = np.reshape(y+dy, (-1, 1)), np.reshape(x+dx, (-1, 1)), np.reshape(z, (-1, 1))\n",
    "\n",
    "    return map_coordinates(image, indices, order=1, mode='reflect').reshape(shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<os._wrap_close at 0x7ff78810ad50>"
      ]
     },
     "execution_count": 200,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "aug_dataset_path = '../pancreas/aug-tumor-seg-dataset/'\n",
    "\n",
    "if os.path.exists(aug_dataset_path):\n",
    "    import shutil\n",
    "    shutil.rmtree(aug_dataset_path, ignore_errors=True)\n",
    "    print(\"deleting... Done!\")\n",
    "    \n",
    "os.mkdir(aug_dataset_path)\n",
    "os.mkdir(os.path.join(aug_dataset_path,'imgs'))\n",
    "os.mkdir(os.path.join(aug_dataset_path,'masks'))\n",
    "\n",
    "os.popen('cp -r %s %s'%(os.path.join(new_dataset_path,'imgs'),aug_dataset_path))\n",
    "os.popen('cp -r %s %s'%(os.path.join(new_dataset_path,'masks'),aug_dataset_path))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np \n",
    "\n",
    "img_path = os.path.join(aug_dataset_path,'imgs')\n",
    "mask_path = os.path.join(aug_dataset_path,'masks')\n",
    "\n",
    "for name in os.listdir(img_path):\n",
    "    img = cv2.imread(os.path.join(img_path,name))\n",
    "    mask = cv2.imread(os.path.join(mask_path,name[:-4]+'_mask.jpg'))\n",
    "    \n",
    "    rand_nums = np.random.randint(0,99,(2,))\n",
    "    \n",
    "    affine_transformed_img = affine_transform(img,img.shape[1] * 0.09,np.random.RandomState(rand_nums[0]))\n",
    "    affine_transformed_mask = affine_transform(mask,img.shape[1] * 0.09,np.random.RandomState(rand_nums[0]))\n",
    "    \n",
    "    elatic_transformed_img = elastic_transform(img,img.shape[1]*3,img.shape[1]*0.07,np.random.RandomState(rand_nums[1]))\n",
    "    elatic_transformed_mask = elastic_transform(mask,img.shape[1]*3,img.shape[1]*0.07,np.random.RandomState(rand_nums[1]))\n",
    "    \n",
    "    cv2.imwrite(os.path.join(img_path,name[:-4]+'a.jpg'),affine_transformed_img[...,0])\n",
    "    cv2.imwrite(os.path.join(mask_path,name[:-4]+'a_mask.jpg'),affine_transformed_mask[...,0])\n",
    "    \n",
    "    cv2.imwrite(os.path.join(img_path,name[:-4]+'e.jpg'),elatic_transformed_img[...,0])\n",
    "    cv2.imwrite(os.path.join(mask_path,name[:-4]+'e_mask.jpg'),elatic_transformed_mask[...,0])\n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "39"
      ]
     },
     "execution_count": 202,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(os.listdir(img_path))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# 移动生成的数据集到所有UNet代码目录(即移动到各个衍生目录)\n",
    "import glob\n",
    "UNet_path = glob.glob('/GPUFS/sysu_rxwang_1/defeng/*UNet*')\n",
    "\n",
    "for path in UNet_path:\n",
    "    if os.path.exists(os.path.join(path,'data','imgs')):\n",
    "        import shutil\n",
    "        shutil.rmtree(os.path.join(path,'data','imgs'), ignore_errors=True)\n",
    "        print(\"deleting... Done!\")\n",
    "\n",
    "    if os.path.exists(os.path.join(path,'data','masks')):\n",
    "        import shutil\n",
    "        shutil.rmtree(os.path.join(path,'data','masks'), ignore_errors=True)\n",
    "        print(\"deleting... Done!\")\n",
    "        \n",
    "# for path in UNet_path:\n",
    "#     if os.path.exists(os.path.join(path,'imgs')):\n",
    "#         import shutil\n",
    "#         shutil.rmtree(os.path.join(path,'imgs'), ignore_errors=True)\n",
    "#         print(\"deleting... Done!\")\n",
    "\n",
    "#     if os.path.exists(os.path.join(path,'masks')):\n",
    "#         import shutil\n",
    "#         shutil.rmtree(os.path.join(path,'masks'), ignore_errors=True)\n",
    "#         print(\"deleting... Done!\")\n",
    "\n",
    "\n",
    "    os.popen('cp -r %s %s'%(os.path.join(aug_dataset_path,'imgs'),os.path.join(path,'data','imgs')))\n",
    "    os.popen('cp -r %s %s'%(os.path.join(aug_dataset_path,'masks'),os.path.join(path,'data','masks')))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "rand_nums = np.random.randint(0,99,(2,))\n",
    "a1 = np.random.RandomState(rand_nums[0])\n",
    "a2 = np.random.RandomState(rand_nums[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([0.09310829]), array([0.09310829]), array([0.97165592]))"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a2.rand(1),a1.rand(1),a1.rand(1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 制作完全没有pancreas的数据集"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "metadata": {},
   "outputs": [],
   "source": [
    "img_path = '../pancreas/backup1-seg-dataset/imgs/'\n",
    "tumor_path = '../pancreas/backup1-seg-dataset/tumor/'\n",
    "pancreas_path = '../pancreas/backup1-seg-dataset/pancreas/'\n",
    "# result_path = '/GPUFS/sysu_rxwang_1/defeng/pancreas/backup1-seg-dataset/results'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "deleting... Done!\n"
     ]
    }
   ],
   "source": [
    "without_pancreas_dataset_path = '../pancreas/without-tumor-seg-dataset/'\n",
    "\n",
    "if os.path.exists(without_pancreas_dataset_path):\n",
    "    import shutil\n",
    "    shutil.rmtree(without_pancreas_dataset_path, ignore_errors=True)\n",
    "    print(\"deleting... Done!\")\n",
    "    \n",
    "os.mkdir(without_pancreas_dataset_path)\n",
    "os.mkdir(os.path.join(without_pancreas_dataset_path,'imgs'))\n",
    "os.mkdir(os.path.join(without_pancreas_dataset_path,'masks'))\n",
    "\n",
    "# os.popen('cp -r %s %s'%(os.path.join(new_dataset_path,'imgs'),without_pancreas_dataset_path))\n",
    "# os.popen('cp -r %s %s'%(os.path.join(new_dataset_path,'masks'),without_pancreas_dataset_path))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 205,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "44.882301330566406\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "44.733333587646484\n",
      "35.69512939453125\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "34.92687225341797\n",
      "41.60724639892578\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "38.43376541137695\n",
      "46.12549591064453\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "46.12549591064453\n",
      "37.39714431762695\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "36.54237747192383\n",
      "44.525909423828125\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "44.31172561645508\n",
      "41.91610336303711\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "39.49075698852539\n",
      "34.121490478515625\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "33.316551208496094\n",
      "39.14832305908203\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "37.52403259277344\n",
      "40.87907791137695\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "38.283687591552734\n",
      "41.847049713134766\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "40.39061737060547\n",
      "41.50886535644531\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "39.78419876098633\n",
      "43.04283905029297\n",
      "(512, 512, 512) (512, 512) (512, 512)\n",
      "42.374305725097656\n"
     ]
    }
   ],
   "source": [
    "tumor_names = [name[:-5] for name in os.listdir(tumor_path)]\n",
    "\n",
    "\n",
    "for index,name in enumerate(tumor_names):\n",
    "    img = cv2.imread(os.path.join(img_path,name+'.jpg'),0)\n",
    "    img[0:33,0:245] = 0\n",
    "    h,w = img.shape[:2]\n",
    "    \n",
    "    flag = 0\n",
    "    #color_mask = np.ones((*img.shape,3))\n",
    "    tumor_mask = np.zeros([h, w, 1], np.uint8)\n",
    "    pancreas_mask = np.zeros([h, w, 1], np.uint8)\n",
    "    if os.path.exists(os.path.join(tumor_path,name+'.json')):\n",
    "        #flag = 1\n",
    "        if os.path.exists(os.path.join(pancreas_path,name+'.json')):\n",
    "            flag = 1\n",
    "            with open(os.path.join(pancreas_path,name+'.json'), \"r\") as f:\n",
    "                annos = json.load(f)\n",
    "                shapes = annos['shapes']\n",
    "                for shape in shapes:\n",
    "                    label = shape['label']\n",
    "\n",
    "                    points = shape['points']\n",
    "                    points_array = np.array(points, dtype=np.int32)\n",
    "                    pancreas_mask = cv2.fillPoly(pancreas_mask, [points_array], 255)\n",
    "        with open(os.path.join(tumor_path,name+'.json'), \"r\") as f:\n",
    "            annos = json.load(f)\n",
    "            shapes = annos['shapes']\n",
    "            for shape in shapes:\n",
    "                label = shape['label']\n",
    "                \n",
    "                points = shape['points']\n",
    "                points_array = np.array(points, dtype=np.int32)\n",
    "                tumor_mask = cv2.fillPoly(tumor_mask, [points_array], 255)\n",
    "    \n",
    "#     pancreas_mask = np.where(pancreas_mask==255,1,0)\n",
    "#     tumor_mask = np.where(tumor_mask==255,1,0)\n",
    "#     intersect = pancreas_mask & tumor_mask\n",
    "#     pure_pancreas = pancreas_mask & (1-intersect)\n",
    "#     pure_pancreas = pure_pancreas.squeeze(axis=-1)\n",
    "#     print(pure_pancreas.shape)\n",
    "    \n",
    "    pure_pancreas1 = np.where(((pancreas_mask==255) & (tumor_mask==0)),1,0).squeeze(axis=-1)\n",
    "    \n",
    "    #print(pure_pancreas==pure_pancreas1) true\n",
    "    \n",
    "    print(img.mean())\n",
    "    img = img*(1-pure_pancreas1) #✔\n",
    "    #img[pure_pancreas1] = 0 #❌ 见5.test\n",
    "    print(img[pure_pancreas1].shape,img.shape,pure_pancreas1.shape)\n",
    "    print(img.mean())\n",
    "\n",
    "    #print(tumor_mask)\n",
    "    \n",
    "    \n",
    "#         if flag == 1:\n",
    "#             color_mask = cv2.addWeighted(color_mask1,1-alpha,color_mask2,alpha,0) #https://blog.csdn.net/zh_jessica/article/details/77992578\n",
    "#         else:\n",
    "#             print('img:%s(index:%d) has no pancreas'%(name,index))\n",
    "#             color_mask = color_mask2\n",
    "\n",
    "    #if flag == 0: continue\n",
    "\n",
    "    cv2.imwrite(os.path.join(without_pancreas_dataset_path,'imgs',str(index)+'.jpg'),img)\n",
    "    cv2.imwrite(os.path.join(without_pancreas_dataset_path,'masks',str(index)+'_mask'+'.jpg'),tumor_mask)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 对4得到的数据集进行data augmentation(elastic&affine)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 206,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<os._wrap_close at 0x7ff788057590>"
      ]
     },
     "execution_count": 206,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "aug_dataset_path = '../pancreas/aug-without-tumor-seg-dataset/'\n",
    "\n",
    "if os.path.exists(aug_dataset_path):\n",
    "    import shutil\n",
    "    shutil.rmtree(aug_dataset_path, ignore_errors=True)\n",
    "    print(\"deleting... Done!\")\n",
    "    \n",
    "os.mkdir(aug_dataset_path)\n",
    "os.mkdir(os.path.join(aug_dataset_path,'imgs'))\n",
    "os.mkdir(os.path.join(aug_dataset_path,'masks'))\n",
    "\n",
    "os.popen('cp -r %s %s'%(os.path.join(without_pancreas_dataset_path,'imgs'),aug_dataset_path))\n",
    "os.popen('cp -r %s %s'%(os.path.join(without_pancreas_dataset_path,'masks'),aug_dataset_path))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 207,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n",
      "(512, 3) (512, 512)\n"
     ]
    }
   ],
   "source": [
    "import numpy as np \n",
    "\n",
    "img_path = os.path.join(aug_dataset_path,'imgs')\n",
    "mask_path = os.path.join(aug_dataset_path,'masks')\n",
    "\n",
    "for name in os.listdir(img_path):\n",
    "    img = cv2.imread(os.path.join(img_path,name))\n",
    "    mask = cv2.imread(os.path.join(mask_path,name[:-4]+'_mask.jpg'))\n",
    "    \n",
    "    rand_nums = np.random.randint(0,99,(2,))\n",
    "    \n",
    "    affine_transformed_img = affine_transform(img,img.shape[1] * 0.09,np.random.RandomState(rand_nums[0]))\n",
    "    affine_transformed_mask = affine_transform(mask,img.shape[1] * 0.09,np.random.RandomState(rand_nums[0]))\n",
    "    \n",
    "    elatic_transformed_img = elastic_transform(img,img.shape[1]*3,img.shape[1]*0.07,np.random.RandomState(rand_nums[1]))\n",
    "    elatic_transformed_mask = elastic_transform(mask,img.shape[1]*3,img.shape[1]*0.07,np.random.RandomState(rand_nums[1]))\n",
    "    \n",
    "    print(affine_transformed_img[0].shape,elatic_transformed_img[...,0].shape)\n",
    "    cv2.imwrite(os.path.join(img_path,name[:-4]+'a.jpg'),affine_transformed_img[...,0])\n",
    "    cv2.imwrite(os.path.join(mask_path,name[:-4]+'a_mask.jpg'),affine_transformed_mask[...,0]) \n",
    "    \n",
    "    cv2.imwrite(os.path.join(img_path,name[:-4]+'e.jpg'),elatic_transformed_img[...,0])\n",
    "    cv2.imwrite(os.path.join(mask_path,name[:-4]+'e_mask.jpg'),elatic_transformed_mask[...,0])\n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 把aug后的数据集(with/out pancreas)移动到backup8的data目录中，并做好训练集、验证集和测试集的分类"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 348,
   "metadata": {},
   "outputs": [],
   "source": [
    "dir_with_pancreas_dataset = '../pancreas/aug-tumor-seg-dataset/'\n",
    "dir_without_pancreas_dataset = '../pancreas/aug-without-tumor-seg-dataset/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 349,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "deleting... Done!\n"
     ]
    }
   ],
   "source": [
    "# 创建文件夹\n",
    "data = '../backup8-ablation-UNet/data'\n",
    "With = '../backup8-ablation-UNet/data/with'\n",
    "without = '../backup8-ablation-UNet/data/without'\n",
    "\n",
    "if os.path.exists(data):\n",
    "    import shutil\n",
    "    shutil.rmtree(data, ignore_errors=True)\n",
    "    print(\"deleting... Done!\")\n",
    "os.mkdir(data)\n",
    "\n",
    "if os.path.exists(With):\n",
    "    import shutil\n",
    "    shutil.rmtree(With, ignore_errors=True)\n",
    "    print(\"deleting... Done!\")\n",
    "os.mkdir(With)\n",
    "os.mkdir(os.path.join(With,'train'))\n",
    "os.mkdir(os.path.join(With,'train','imgs'))\n",
    "os.mkdir(os.path.join(With,'train','masks'))\n",
    "os.mkdir(os.path.join(With,'val'))\n",
    "os.mkdir(os.path.join(With,'val','imgs'))\n",
    "os.mkdir(os.path.join(With,'val','masks'))\n",
    "os.mkdir(os.path.join(With,'test'))\n",
    "os.mkdir(os.path.join(With,'test','imgs'))\n",
    "os.mkdir(os.path.join(With,'test','masks'))\n",
    "\n",
    "if os.path.exists(without):\n",
    "    import shutil\n",
    "    shutil.rmtree(without, ignore_errors=True)\n",
    "    print(\"deleting... Done!\")\n",
    "os.mkdir(without)\n",
    "os.mkdir(os.path.join(without,'train'))\n",
    "os.mkdir(os.path.join(without,'train','imgs'))\n",
    "os.mkdir(os.path.join(without,'train','masks'))\n",
    "os.mkdir(os.path.join(without,'val'))\n",
    "os.mkdir(os.path.join(without,'val','imgs'))\n",
    "os.mkdir(os.path.join(without,'val','masks'))\n",
    "os.mkdir(os.path.join(without,'test'))\n",
    "os.mkdir(os.path.join(without,'test','imgs'))\n",
    "os.mkdir(os.path.join(without,'test','masks'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 350,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['9', '4', '3', '12', '8'] \n",
      "\n",
      "['3e', '5e', '9e', '7e', '12e'] \n",
      "\n",
      "['7a', '0a', '10a', '2a', '8a']\n"
     ]
    }
   ],
   "source": [
    "import glob\n",
    "\n",
    "original_img_names = glob.glob(os.path.join(dir_with_pancreas_dataset,'imgs','*[!ea].jpg'))\n",
    "original_img_names = [name.split('/')[-1][:-4] for name in original_img_names]\n",
    "elastic_img_names = glob.glob(os.path.join(dir_with_pancreas_dataset,'imgs','*e.jpg'))\n",
    "elastic_img_names = [name.split('/')[-1][:-4] for name in elastic_img_names]\n",
    "affine_img_names = glob.glob(os.path.join(dir_with_pancreas_dataset,'imgs','*a.jpg'))\n",
    "affine_img_names = [name.split('/')[-1][:-4] for name in affine_img_names]\n",
    "\n",
    "print(original_img_names[:5],'\\n')\n",
    "print(elastic_img_names[:5],'\\n')\n",
    "print(affine_img_names[:5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 351,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(13, 13)"
      ]
     },
     "execution_count": 351,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(original_img_names),len(elastic_img_names)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 352,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(['1', '11', '7'],\n",
       " '**************************************************',\n",
       " ['9', '4e', '4a'],\n",
       " 9)"
      ]
     },
     "execution_count": 352,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import random\n",
    "random.seed(8081)\n",
    "\n",
    "\n",
    "# 采样set-B and set-C\n",
    "original_sample = random.sample(original_img_names,4)\n",
    "aug_sample = random.sample(elastic_img_names,1) + random.sample(affine_img_names,1)\n",
    "for idx,name in enumerate(original_sample):\n",
    "    original_img_names.remove(name)\n",
    "    \n",
    "    #elastic_img_names.remove(name+'e')\n",
    "    #affine_img_names.remove(name+'a')\n",
    "    #add\n",
    "#     if idx != 3:\n",
    "#         elastic_img_names.remove(name+'e')\n",
    "#         affine_img_names.remove(name+'a')\n",
    "\n",
    "\n",
    "elastic_img_names.remove(aug_sample[0])\n",
    "# if aug_sample[0][0] in original_img_names:\n",
    "#     original_img_names.remove(aug_sample[0][0])\n",
    "# if aug_sample[0][0]+'e' in affine_img_names:\n",
    "#     affine_img_names.remove(aug_sample[0][0]+'a')\n",
    "\n",
    "affine_img_names.remove(aug_sample[1])\n",
    "# if aug_sample[1][0] in original_img_names:\n",
    "#     original_img_names.remove(aug_sample[1][0])\n",
    "# if aug_sample[1][0]+'e' in elastic_img_names:\n",
    "#     elastic_img_names.remove(aug_sample[1][0]+'e')\n",
    "    \n",
    "\n",
    "B = original_sample[:3]\n",
    "#B = [name.split('/')[-1][:-4] for name in B]\n",
    "C = [original_sample[-1]] + aug_sample\n",
    "#C = [name.split('/')[-1][:-4] for name in C]\n",
    "\n",
    "B,'*'*50,C,len(original_img_names)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 353,
   "metadata": {},
   "outputs": [],
   "source": [
    "suffix = '_mask'\n",
    "jpg = '.jpg'\n",
    "\n",
    "# 移动val\n",
    "# val1 = \n",
    "# val2 = \n",
    "for name in B:\n",
    "    #with - imgs\n",
    "    src_path_with = os.path.join(dir_with_pancreas_dataset,'imgs',name+jpg)\n",
    "    dst_path_with = os.path.join(With,'test','imgs',name+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_with,dst_path_with))\n",
    "    #with - masks\n",
    "    src_path_with = os.path.join(dir_with_pancreas_dataset,'masks',name+suffix+jpg)\n",
    "    dst_path_with = os.path.join(With,'test','masks',name+suffix+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_with,dst_path_with))\n",
    "    \n",
    "    #-------------- 对于B--测试集来说，都是使用没有mask掉胰腺区域的图片\n",
    "    #without - imgs\n",
    "    src_path_without = os.path.join(dir_with_pancreas_dataset,'imgs',name+jpg)\n",
    "    dst_path_without = os.path.join(without,'test','imgs',name+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_without,dst_path_without))\n",
    "    #without - masks\n",
    "    src_path_without = os.path.join(dir_with_pancreas_dataset,'masks',name+suffix+jpg)\n",
    "    dst_path_without = os.path.join(without,'test','masks',name+suffix+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_without,dst_path_without))\n",
    "    \n",
    "for name in C:\n",
    "    #with - imgs\n",
    "    src_path_with = os.path.join(dir_with_pancreas_dataset,'imgs',name+jpg)\n",
    "    dst_path_with = os.path.join(With,'val','imgs',name+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_with,dst_path_with))\n",
    "    #with - masks\n",
    "    src_path_with = os.path.join(dir_with_pancreas_dataset,'masks',name+suffix+jpg)\n",
    "    dst_path_with = os.path.join(With,'val','masks',name+suffix+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_with,dst_path_with))\n",
    "    \n",
    "    #-------------- 对于C--验证集来说，都是使用没有mask掉胰腺区域的图片\n",
    "    #without - imgs\n",
    "    src_path_without = os.path.join(dir_with_pancreas_dataset,'imgs',name+jpg)\n",
    "    dst_path_without = os.path.join(without,'val','imgs',name+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_without,dst_path_without))\n",
    "    #without - masks\n",
    "    src_path_without = os.path.join(dir_with_pancreas_dataset,'masks',name+suffix+jpg)\n",
    "    dst_path_without = os.path.join(without,'val','masks',name+suffix+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_without,dst_path_without))\n",
    "\n",
    "for name in original_img_names:\n",
    "    #with - imgs\n",
    "    src_path_with = os.path.join(dir_with_pancreas_dataset,'imgs',name+jpg)\n",
    "    dst_path_with = os.path.join(With,'train','imgs',name+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_with,dst_path_with))\n",
    "    #with - masks\n",
    "    src_path_with = os.path.join(dir_with_pancreas_dataset,'masks',name+suffix+jpg)\n",
    "    dst_path_with = os.path.join(With,'train','masks',name+suffix+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_with,dst_path_with))\n",
    "    \n",
    "    #without - imgs\n",
    "    src_path_without = os.path.join(dir_without_pancreas_dataset,'imgs',name+jpg)\n",
    "    dst_path_without = os.path.join(without,'train','imgs',name+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_without,dst_path_without))\n",
    "    #without - masks\n",
    "    src_path_without = os.path.join(dir_without_pancreas_dataset,'masks',name+suffix+jpg)\n",
    "    dst_path_without = os.path.join(without,'train','masks',name+suffix+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_without,dst_path_without))\n",
    "    \n",
    "for name in elastic_img_names:\n",
    "    #with - imgs\n",
    "    src_path_with = os.path.join(dir_with_pancreas_dataset,'imgs',name+jpg)\n",
    "    dst_path_with = os.path.join(With,'train','imgs',name+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_with,dst_path_with))\n",
    "    #with - masks\n",
    "    src_path_with = os.path.join(dir_with_pancreas_dataset,'masks',name+suffix+jpg)\n",
    "    dst_path_with = os.path.join(With,'train','masks',name+suffix+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_with,dst_path_with))\n",
    "    \n",
    "    #without - imgs\n",
    "    src_path_without = os.path.join(dir_without_pancreas_dataset,'imgs',name+jpg)\n",
    "    dst_path_without = os.path.join(without,'train','imgs',name+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_without,dst_path_without))\n",
    "    #without - masks\n",
    "    src_path_without = os.path.join(dir_without_pancreas_dataset,'masks',name+suffix+jpg)\n",
    "    dst_path_without = os.path.join(without,'train','masks',name+suffix+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_without,dst_path_without))\n",
    "    \n",
    "for name in affine_img_names:\n",
    "    #with - imgs\n",
    "    src_path_with = os.path.join(dir_with_pancreas_dataset,'imgs',name+jpg)\n",
    "    dst_path_with = os.path.join(With,'train','imgs',name+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_with,dst_path_with))\n",
    "    #with - masks\n",
    "    src_path_with = os.path.join(dir_with_pancreas_dataset,'masks',name+suffix+jpg)\n",
    "    dst_path_with = os.path.join(With,'train','masks',name+suffix+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_with,dst_path_with))\n",
    "    \n",
    "    #without - imgs\n",
    "    src_path_without = os.path.join(dir_without_pancreas_dataset,'imgs',name+jpg)\n",
    "    dst_path_without = os.path.join(without,'train','imgs',name+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_without,dst_path_without))\n",
    "    #without - masks\n",
    "    src_path_without = os.path.join(dir_without_pancreas_dataset,'masks',name+suffix+jpg)\n",
    "    dst_path_without = os.path.join(without,'train','masks',name+suffix+jpg)\n",
    "    os.popen('cp %s %s'%(src_path_without,dst_path_without))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 343,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(33, 3, 3)"
      ]
     },
     "execution_count": 343,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(os.listdir(os.path.join(without,'train','imgs'))),len(os.listdir(os.path.join(without,'val','imgs'))),len(os.listdir(os.path.join(without,'test','imgs')))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 345,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(33, 3, 3)"
      ]
     },
     "execution_count": 345,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(os.listdir(os.path.join(without,'train','masks'))),len(os.listdir(os.path.join(without,'val','masks'))),len(os.listdir(os.path.join(without,'test','masks')))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<mark>==========================================================================================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 180,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['../backup8-ablation-UNet/data/with/train/masks/0a_mask.jpg']"
      ]
     },
     "execution_count": 180,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mask_file = glob.glob('../backup8-ablation-UNet/data/with/train/masks/0a_mask.*')\n",
    "mask_file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 178,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['/data/defeng/backup8-ablation-UNet/data/with/train/masks/0a_mask.jpg']"
      ]
     },
     "execution_count": 178,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mask_file = glob.glob('/data/defeng/backup8-ablation-UNet/data/with/train/masks/0a_mask.*')\n",
    "mask_file"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\\\\sshfs\\defeng@172.18.167.97"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['5.jpg',\n",
       " '9.jpg',\n",
       " '4.jpg',\n",
       " '10.jpg',\n",
       " '6.jpg',\n",
       " '11.jpg',\n",
       " '12.jpg',\n",
       " '1.jpg',\n",
       " '7.jpg',\n",
       " '0.jpg',\n",
       " '2.jpg',\n",
       " '8.jpg',\n",
       " '3.jpg']"
      ]
     },
     "execution_count": 134,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import os\n",
    "os.listdir('../backup5-fusion-UNet/data/imgs/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 0.73298536 -0.3466885 ]\n",
      " [-0.35624475  0.67056095]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(array([[0., 0.],\n",
       "        [0., 0.]]),\n",
       " (2, 2))"
      ]
     },
     "execution_count": 125,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a = np.random.randint(0,2,(2,2))\n",
    "b = np.random.randn(2,2)\n",
    "\n",
    "print(b)\n",
    "b[a] = 0\n",
    "b,b.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1, 1],\n",
       "       [1, 0]])"
      ]
     },
     "execution_count": 133,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[0., 0.],\n",
       "        [0., 0.]],\n",
       "\n",
       "       [[0., 0.],\n",
       "        [0., 0.]]])"
      ]
     },
     "execution_count": 131,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "b[a]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x2ae58f2d9b10>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# '../Pytorch-UNet/data/masks/0_mask.jpg'\n",
    "# https://blog.csdn.net/qq295456059/article/details/51378313\n",
    "# 灰度图训练时注意图片读入方式，必须加0，反则就变为三通道的了。\n",
    "img = cv2.imread('../pancreas/backup2-seg-dataset/imgs/0.jpg',0)\n",
    "plt.imshow(img,cmap='gray')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'L'"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from PIL import Image\n",
    "img = Image.open('../Pytorch-UNet/data/imgs/0.jpg')\n",
    "img.mode"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(512, 512, 3)"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img = np.array(img)\n",
    "img.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(512, 512, 1)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img = np.expand_dims(img, axis=2)\n",
    "img.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "255"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x2ae58e87f990>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "img[0:33,0:245] = 0\n",
    "plt.imshow(img,cmap='gray')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import args\n",
    "\n",
    "parser = argparse.ArgumentParser(description='Train the UNet on images and target masks',\n",
    "                                 formatter_class=argparse.ArgumentDefaultsHelpFormatter)\n",
    "parser.add_argument('-e', '--epochs', metavar='E', type=int, default=-1,\n",
    "                    help='Number of epochs', dest='epochs')\n",
    "parser.add_argument('-t', '--threshold', metavar='E', type=float, default=0.5,\n",
    "                    help='threshold for classification', dest='threshold')\n",
    "\n",
    "parser.parse_args()\n",
    "\n",
    "path = './data/'\n",
    "\n",
    "EPOCH = parser.epochs\n",
    "if EPOCH == -1:\n",
    "    \n",
    "\n",
    "\n",
    "command = 'python predict.py -m checkpoints/CP_epoch300.pth --scale 1.0 -i data/imgs/0.jpg -o data/val_result/val_0.jpg -t 0.5'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import re\n",
    "\n",
    "regex = re.compile(r'\\d+')\n",
    "epochs_index = [int(regex.findall(name[:-4])[0]) for name in os.listdir('../Pytorch-UNet/checkpoints/')]\n",
    "\n",
    "\n",
    "# num = int(max(regex.findall(epochs_index)))\n",
    "# print(num)\n",
    "max(epochs_index)\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:torch]",
   "language": "python",
   "name": "conda-env-torch-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.10"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}