LeaveOut_result_processing.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 82,
   "id": "electoral-memory",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "emerging-international",
   "metadata": {},
   "source": [
    "# test torch.utils.data.Subset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "paperback-mailman",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "---------------------------------------------------------------------------\n",
      "[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11] torch.Size([1, 512, 512])\n",
      "[0] torch.Size([1, 512, 512])\n",
      "---------------------------------------------------------------------------\n",
      "[0, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11] torch.Size([1, 512, 512])\n",
      "[1] torch.Size([1, 512, 512])\n",
      "---------------------------------------------------------------------------\n",
      "[0, 1, 3, 4, 5, 6, 7, 8, 9, 10, 11] torch.Size([1, 512, 512])\n",
      "[2] torch.Size([1, 512, 512])\n",
      "---------------------------------------------------------------------------\n",
      "[0, 1, 2, 4, 5, 6, 7, 8, 9, 10, 11] torch.Size([1, 512, 512])\n",
      "[3] torch.Size([1, 512, 512])\n",
      "---------------------------------------------------------------------------\n",
      "[0, 1, 2, 3, 5, 6, 7, 8, 9, 10, 11] torch.Size([1, 512, 512])\n",
      "[4] torch.Size([1, 512, 512])\n",
      "---------------------------------------------------------------------------\n",
      "[0, 1, 2, 3, 4, 6, 7, 8, 9, 10, 11] torch.Size([1, 512, 512])\n",
      "[5] torch.Size([1, 512, 512])\n",
      "---------------------------------------------------------------------------\n",
      "[0, 1, 2, 3, 4, 5, 7, 8, 9, 10, 11] torch.Size([1, 512, 512])\n",
      "[6] torch.Size([1, 512, 512])\n",
      "---------------------------------------------------------------------------\n",
      "[0, 1, 2, 3, 4, 5, 6, 8, 9, 10, 11] torch.Size([1, 512, 512])\n",
      "[7] torch.Size([1, 512, 512])\n",
      "---------------------------------------------------------------------------\n",
      "[0, 1, 2, 3, 4, 5, 6, 7, 9, 10, 11] torch.Size([1, 512, 512])\n",
      "[8] torch.Size([1, 512, 512])\n",
      "---------------------------------------------------------------------------\n",
      "[0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 11] torch.Size([1, 512, 512])\n",
      "[9] torch.Size([1, 512, 512])\n",
      "---------------------------------------------------------------------------\n",
      "[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 11] torch.Size([1, 512, 512])\n",
      "[10] torch.Size([1, 512, 512])\n",
      "---------------------------------------------------------------------------\n",
      "[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10] torch.Size([1, 512, 512])\n",
      "[11] torch.Size([1, 512, 512])\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'\\n    从源码来看：h%killbgscriptstps://pytorch.org/docs/master/_modules/torch/utils/data/dataset.html#Subset\\n    上面循环中的得到的各个subset都是共用同一个“父亲”dataset对象，但是这并%%sh任何的影响和干扰(从源码易懂)！\\n'"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from torch.utils.data import Subset\n",
    "\n",
    "from utils.dataset import BasicDataset\n",
    "dataset = BasicDataset('./data/imgs/', './data/masks/', 1.0,mask_suffix='_mask')\n",
    "size = len(dataset)\n",
    "\n",
    "sets = []\n",
    "leave_out_subsets = []\n",
    "\n",
    "for i in range(size):\n",
    "    #sets.append(BasicDataset('./data/imgs/', './data/masks/', 1.0,mask_suffix='_mask'))\n",
    "    \n",
    "    print('-'*75)\n",
    "    \n",
    "    train_indexs = list(range(size))\n",
    "    train_indexs.remove(i)\n",
    "    \n",
    "    train = Subset(dataset,train_indexs)\n",
    "    val = Subset(dataset,[i,])\n",
    "    \n",
    "    print(train.indices,train[0]['image'].shape)\n",
    "    print(val.indices,val[0]['image'].shape)\n",
    "    \n",
    "    leave_out_subsets.append([train,val])\n",
    "    \n",
    "    \n",
    "'''\n",
    "    从源码来看：h%killbgscriptstps://pytorch.org/docs/master/_modules/torch/utils/data/dataset.html#Subset\n",
    "    上面循环中的得到的各个subset都是共用同一个“父亲”dataset对象，但是这并%%sh任何的影响和干扰(从源码易懂)！\n",
    "'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "oriental-crack",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "1\n",
      "0\n",
      "1\n"
     ]
    }
   ],
   "source": [
    "for i in range(4):\n",
    "    print(i%2)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "authentic-coaching",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "12"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import os\n",
    "\n",
    "max_size = len(os.listdir('/GPUFS/sysu_rxwang_1/defeng/Pytorch-UNet/data/imgs/'))\n",
    "\n",
    "max_size"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "combined-density",
   "metadata": {},
   "source": [
    "# show diff and predict on  train set"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "pleased-messenger",
   "metadata": {},
   "source": [
    "显示在训练集上的train和val的结果。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "piano-crack",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import logging \n",
    "\n",
    "import torch\n",
    "import torch.nn.functional as F\n",
    "from torch.utils.data import Dataset,DataLoader\n",
    "\n",
    "#from tqdm import tqdm\n",
    "from tqdm.notebook import tqdm\n",
    "\n",
    "from dice_loss import dice_coeff\n",
    "\n",
    "from PIL import Image\n",
    "import cv2\n",
    "import numpy as np\n",
    "\n",
    "from torchvision import transforms\n",
    "\n",
    "from unet import UNet\n",
    "from utils.data_vis import plot_img_and_mask\n",
    "from utils.dataset import BasicDataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "downtown-shareware",
   "metadata": {},
   "outputs": [],
   "source": [
    "def leave_out_eval(net, loader, device,save_path,imgs_name):\n",
    "    \"\"\"Evaluation without the densecrf with the dice coefficient\"\"\"\n",
    "    mask_type = torch.float32 if net.n_classes == 1 else torch.long\n",
    "    \n",
    "    flag=0\n",
    "    with tqdm(total=len(loader), desc='Prediction round', unit='batch', leave=True) as pbar:\n",
    "        for idx,batch in enumerate(loader):\n",
    "            imgs, true_masks = batch['image'], batch['mask']\n",
    "            imgs = imgs.to(device=device, dtype=torch.float32)\n",
    "            #true_masks = true_masks.to(device=device, dtype=mask_type)\n",
    "            \n",
    "            \n",
    "            with torch.no_grad():\n",
    "                mask_pred = net(imgs)\n",
    "            \n",
    "            probs = torch.sigmoid(mask_pred)\n",
    "            pred = probs>0.5 #pred:[8,512,512,1]\n",
    "            pred = pred.cpu().numpy().astype(np.uint8).transpose(0,2,3,1) #pred:[8,512,512,1]\n",
    "            if flag==0:\n",
    "                flag=1\n",
    "                np.save('probs.npy',probs.cpu().numpy())\n",
    "            \n",
    "            diff = pred - true_masks.cpu().numpy().astype(np.uint8).transpose(0,2,3,1) #diff:[8,512,512,1]\n",
    "            diff = np.squeeze(diff,axis=-1) #diff:[8,512,512]\n",
    "            #diff = diff.repeat([3],axis=-1)\n",
    "            pred = pred*255\n",
    "            blue = [255,0,0]\n",
    "            red = [0,0,255]\n",
    "            \n",
    "            for i,pic in enumerate(pred):\n",
    "                cv2.imwrite(os.path.join(save_path,imgs_name[i]),pic)\n",
    "                \n",
    "                color_diff = np.ones((*diff.shape[1:],3))\n",
    "                color_diff[diff[i]==1] = blue #over estimate\n",
    "                color_diff[diff[i]==-1] = red #under estimate\n",
    "                cv2.imwrite(os.path.join(save_path,'diff_'+imgs_name[i]),color_diff)\n",
    "                \n",
    "                i+=1\n",
    "            \n",
    "            pbar.update(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "atlantic-height",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['5', '4', '10', '6', '12', '13', '1', '7', '0', '2', '8', '3']\n",
      "Group:exp3_4 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ee452c8c721d4ef585c31f84aa5245bd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Group:exp3_6 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a1154bd65eed439cabded3b7eae9eed3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Group:exp3_9 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "00d7055b70e743a6b674f10a503f9565",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Group:exp3_0 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "da74e130676f41a0be03d1eab6898647",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Group:exp3_1 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1d0d54a6f2384a41b8871cc1d91af9d5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Group:exp3_3 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5ec7bba835794c249e2924a074119369",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Group:exp3_7 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "772ffd17458d447491a805a1ac7aad4a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Group:exp3_10 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b52a58f5bb434a85b54bbb8700fe20b6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Group:exp3_8 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "09cccc8c0c9f48639acd9addae5bb055",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Group:exp3_2 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "cf10fff42a634f4587fb008011fd15ad",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Group:exp3_11 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "83b4f913a9a644258a5b2ded295e0905",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Group:exp3_5 Model loaded !\n",
      "deleting... Done!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c166e95a0be1475cb1c0fc90596e0981",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Prediction round:   0%|          | 0/2 [00:00<?, ?batch/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if __name__ == \"__main__\":\n",
    "    dir_img = 'data/imgs/'\n",
    "    dir_mask = 'data/masks/'\n",
    "    dataset = BasicDataset(dir_img, dir_mask, 1.0,mask_suffix='_mask')\n",
    "    loader = DataLoader(dataset, batch_size=8, shuffle=False, num_workers=8, pin_memory=False, drop_last=False)\n",
    "\n",
    "    net = UNet(n_channels=1, n_classes=1)\n",
    "    net.eval()\n",
    "\n",
    "    device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')\n",
    "    logging.info(f'Using device {device}')\n",
    "    net.to(device=device)\n",
    "    \n",
    "    #name = 'backup4-DICE-UNet'\n",
    "    name = 'backup3-BCE-UNet'\n",
    "    \n",
    "    cpkts_path = '/GPUFS/sysu_rxwang_1/defeng/%s/checkpoints/'%(name)\n",
    "    cpkts_name = os.listdir(cpkts_path)\n",
    "    \n",
    "    save_path = '/GPUFS/sysu_rxwang_1/defeng/%s/data/val_result'%(name)\n",
    "    imgs_name = os.listdir('/GPUFS/sysu_rxwang_1/defeng/%s/data/imgs/'%(name))\n",
    "    \n",
    "    for idx,group in enumerate(cpkts_name):\n",
    "        net.load_state_dict(torch.load(os.path.join(cpkts_path,group,'best.pth'), map_location=device))\n",
    "        print(\"Group:%s Model loaded !\"%(group))\n",
    "        \n",
    "        path = os.path.join(save_path,group)\n",
    "        if os.path.exists(path):\n",
    "            import shutil\n",
    "            shutil.rmtree(path, ignore_errors=True)\n",
    "            print(\"deleting... Done!\")\n",
    "        os.mkdir(path)\n",
    "        \n",
    "        leave_out_eval(net,loader,device,path,imgs_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "joint-safety",
   "metadata": {},
   "source": [
    "# color map (can be exploited in the future)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "id": "still-right",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(512, 512)"
      ]
     },
     "execution_count": 204,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img = cv2.imread('../backup4-UNet/data/val_result/exp3_0/diff_1.jpg',0)\n",
    "img.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "id": "southwest-disability",
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "def colormap(n):\n",
    "    '''\n",
    "    这里有19个颜色，根据自己数据来使用，也可以根据自己的喜爱修改颜色的值。n就是num_classes.\n",
    "    np.uint8意味着array的值为0-255中间。\n",
    "    cmap就是一个映射，当标签为1的时候，这个像素转换成RGB就是[244, 35,232]的三通道值。\n",
    "    '''\n",
    "    cmap=np.zeros([n, 3]).astype(np.uint8)\n",
    "    cmap[0,:] = np.array([ 0,  0,  0])\n",
    "    cmap[1,:] = np.array([244, 35,232])\n",
    "    cmap[2,:] = np.array([ 70, 70, 70])\n",
    "    cmap[3,:] = np.array([ 102,102,156])\n",
    "    cmap[4,:] = np.array([ 190,153,153])\n",
    "    cmap[5,:] = np.array([ 153,153,153])\n",
    " \n",
    "    cmap[6,:] = np.array([ 250,170, 30])\n",
    "    cmap[7,:] = np.array([ 220,220,  0])\n",
    "    cmap[8,:] = np.array([ 107,142, 35])\n",
    "    cmap[9,:] = np.array([ 152,251,152])\n",
    "    cmap[10,:] = np.array([ 70,130,180])\n",
    " \n",
    "    cmap[11,:] = np.array([ 220, 20, 60])\n",
    "    cmap[12,:] = np.array([ 119, 11, 32])\n",
    "    cmap[13,:] = np.array([ 0,  0,142])\n",
    "    cmap[14,:] = np.array([  0,  0, 70])\n",
    "    cmap[15,:] = np.array([  0, 60,100])\n",
    " \n",
    "    cmap[16,:] = np.array([  0, 80,100])\n",
    "    cmap[17,:] = np.array([  0,  0,230])\n",
    "    cmap[18,:] = np.array([ 255,  0,  0])\n",
    "    \n",
    "    return cmap\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 206,
   "id": "charitable-tractor",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Colorize:\n",
    "    def __init__(self, n=2):\n",
    "        self.cmap = colormap(2)\n",
    "        self.cmap[n] = self.cmap[-1]\n",
    "        self.cmap = torch.from_numpy(self.cmap[:n])#array->tensor\n",
    " \n",
    "    def __call__(self, gray_image):\n",
    "        size = gray_image.size()#这里就是上文的output\n",
    "        color_image = torch.ByteTensor(3, size[0], size[1]).fill_(0)\n",
    " \n",
    "        for label in range(0, len(self.cmap)):\n",
    "            mask = gray_image == label\n",
    "            color_image[0][mask] = self.cmap[label][0]\n",
    "            color_image[1][mask] = self.cmap[label][1]\n",
    "            color_image[2][mask] = self.cmap[label][2]\n",
    " \n",
    "        return color_image\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "alien-interpretation",
   "metadata": {},
   "source": [
    "# np.where is a disaster"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 214,
   "id": "located-syntax",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[-0.04584008, -1.10596792,  2.0051606 ],\n",
       "        [ 0.59737921,  1.63828467, -0.29461262]],\n",
       "\n",
       "       [[-1.45298915,  0.40394504,  1.3919785 ],\n",
       "        [-0.4557085 ,  0.92478412, -0.48064994]]])"
      ]
     },
     "execution_count": 214,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "temp = np.random.randn(2,2,3)\n",
    "temp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 217,
   "id": "serial-privacy",
   "metadata": {},
   "outputs": [],
   "source": [
    "temp1 = np.where(temp==[-0.04584008, -1.10596792,  2.0051606],[255.0]*3,[1.0]*3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 218,
   "id": "abroad-garbage",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[1., 1., 1.],\n",
       "        [1., 1., 1.]],\n",
       "\n",
       "       [[1., 1., 1.],\n",
       "        [1., 1., 1.]]])"
      ]
     },
     "execution_count": 218,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "temp1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 212,
   "id": "dutch-negative",
   "metadata": {},
   "outputs": [],
   "source": [
    "temp[0] = np.array([2.2,3.2,5.1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 213,
   "id": "civilian-metropolitan",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[ 2.2       ,  3.2       ,  5.1       ],\n",
       "        [ 2.2       ,  3.2       ,  5.1       ]],\n",
       "\n",
       "       [[ 1.83185683,  0.09220815, -0.75080739],\n",
       "        [ 0.04100745,  0.96471119,  1.06192293]]])"
      ]
     },
     "execution_count": 213,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "temp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 233,
   "id": "interested-talent",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0, 1, 0, 1],\n",
       "       [0, 0, 1, 1],\n",
       "       [0, 0, 0, 1],\n",
       "       [0, 1, 0, 0]])"
      ]
     },
     "execution_count": 233,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mask = np.random.randint(0,2,(4,4))\n",
    "mask"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "id": "israeli-algeria",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.]],\n",
       "\n",
       "       [[1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.]],\n",
       "\n",
       "       [[1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.]],\n",
       "\n",
       "       [[1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.]]])"
      ]
     },
     "execution_count": 241,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "color_mask = np.ones((*mask.shape,3))\n",
    "color_mask"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 242,
   "id": "innocent-weekly",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[255.,   0.,   0.],\n",
       "        [  1.,   1.,   1.],\n",
       "        [255.,   0.,   0.],\n",
       "        [  1.,   1.,   1.]],\n",
       "\n",
       "       [[255.,   0.,   0.],\n",
       "        [255.,   0.,   0.],\n",
       "        [  1.,   1.,   1.],\n",
       "        [  1.,   1.,   1.]],\n",
       "\n",
       "       [[255.,   0.,   0.],\n",
       "        [255.,   0.,   0.],\n",
       "        [255.,   0.,   0.],\n",
       "        [  1.,   1.,   1.]],\n",
       "\n",
       "       [[255.,   0.,   0.],\n",
       "        [  1.,   1.,   1.],\n",
       "        [255.,   0.,   0.],\n",
       "        [255.,   0.,   0.]]])"
      ]
     },
     "execution_count": 242,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "blue = [255,0,0]\n",
    "red = [0,0,255]\n",
    "\n",
    "color_mask[mask==0] = blue\n",
    "color_mask"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 255,
   "id": "offshore-shaft",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x2b923480d510>"
      ]
     },
     "execution_count": 255,
     "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": [
    "import matplotlib.pyplot as plt\n",
    "plt.imshow(color_mask)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 245,
   "id": "sacred-antarctica",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(4,)"
      ]
     },
     "execution_count": 245,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "color_mask.shape[:-2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 243,
   "id": "breathing-macro",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "ename": "ValueError",
     "evalue": "operands could not be broadcast together with shapes (4,4) (3,) (4,4,3) ",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-243-ae3d317ac164>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwhere\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmask\u001b[0m\u001b[0;34m==\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mblue\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mcolor_mask\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m<__array_function__ internals>\u001b[0m in \u001b[0;36mwhere\u001b[0;34m(*args, **kwargs)\u001b[0m\n",
      "\u001b[0;31mValueError\u001b[0m: operands could not be broadcast together with shapes (4,4) (3,) (4,4,3) "
     ]
    }
   ],
   "source": [
    "np.where(mask==1,blue,color_mask)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "opening-factory",
   "metadata": {},
   "source": [
    "# simple boolean index matrix can work."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "flush-webster",
   "metadata": {},
   "source": [
    "<a color='red'>worked!</a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 267,
   "id": "powerful-revision",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        ...,\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.]],\n",
       "\n",
       "       [[1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        ...,\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.]],\n",
       "\n",
       "       [[1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        ...,\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.]],\n",
       "\n",
       "       ...,\n",
       "\n",
       "       [[1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        ...,\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.]],\n",
       "\n",
       "       [[1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        ...,\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.]],\n",
       "\n",
       "       [[1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        ...,\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.],\n",
       "        [1., 1., 1.]]])"
      ]
     },
     "execution_count": 267,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mask = np.random.randint(0,2,(512,512))\n",
    "diff = np.ones((512,512,3))\n",
    "diff"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 268,
   "id": "owned-technique",
   "metadata": {},
   "outputs": [],
   "source": [
    "diff[mask==1]=[255,0,0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 262,
   "id": "normal-angola",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[  1.,   1.,   1.],\n",
       "        [  1.,   1.,   1.],\n",
       "        [255.,   0.,   0.],\n",
       "        [255.,   0.,   0.]],\n",
       "\n",
       "       [[255.,   0.,   0.],\n",
       "        [  1.,   1.,   1.],\n",
       "        [255.,   0.,   0.],\n",
       "        [  1.,   1.,   1.]],\n",
       "\n",
       "       [[255.,   0.,   0.],\n",
       "        [  1.,   1.,   1.],\n",
       "        [255.,   0.,   0.],\n",
       "        [255.,   0.,   0.]],\n",
       "\n",
       "       [[  1.,   1.,   1.],\n",
       "        [255.,   0.,   0.],\n",
       "        [  1.,   1.,   1.],\n",
       "        [255.,   0.,   0.]]])"
      ]
     },
     "execution_count": 262,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "diff"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 269,
   "id": "documented-literature",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([255.,   0.,   0.])"
      ]
     },
     "execution_count": 269,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "diff[0,3]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "determined-trail",
   "metadata": {},
   "source": [
    "# border analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "daily-tourism",
   "metadata": {},
   "source": [
    "1. 取flat的序列，首先取出>0.5的数，然后进行排序，得到>0.5中的后10%的value值\n",
    "2. 将1.中的value值作为threshold，然后去未flat的prob中筛选出对应的空间点，即 pred>0.5 and pred < value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "equal-swaziland",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x1080 with 8 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "rows = 2\n",
    "cols = 4\n",
    "fig,ax = plt.subplots(rows,cols,figsize=[15,15])\n",
    "start_with = 0\n",
    "\n",
    "probs = np.load('/GPUFS/sysu_rxwang_1/defeng/backup4-DICE-UNet/probs.npy').transpose(0,2,3,1)\n",
    "\n",
    "for i in range(rows*cols):\n",
    "    #ind = start_with + i*show_every\n",
    "    \n",
    "    pred = probs[i]\n",
    "\n",
    "    mask = pred>0.5\n",
    "    pred_flat = pred[mask].flatten()\n",
    "    pred_flat.sort()\n",
    "    percent = int(pred_flat.shape[0]*0.1)\n",
    "    value = pred_flat[percent]\n",
    "\n",
    "    points = (pred>0.5) & (pred<value)\n",
    "    result = pred * points\n",
    "    \n",
    "    #print(int(i/cols),int(i%cols))\n",
    "    ax[int(i/cols),int(i % cols)].set_title('slice %d,value:%f' % (i,value))\n",
    "    ax[int(i/cols),int(i % cols)].imshow(result,cmap='gray')\n",
    "    ax[int(i/cols),int(i % cols)].axis('off')\n",
    "\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "shaped-chorus",
   "metadata": {},
   "source": [
    "<mark>pytorch version</mark>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "id": "still-employee",
   "metadata": {},
   "outputs": [],
   "source": [
    "def cal_low_conf_pixels(probs,device):\n",
    "    bs = probs.shape[0]\n",
    "    results = torch.zeros(probs.shape).to(device)\n",
    "    \n",
    "    for i in range(bs):\n",
    "        prob = probs[i]\n",
    "        mask = prob>0.5\n",
    "        pred_flat = prob[mask].flatten()\n",
    "        pred_flat = pred_flat.sort()[0]\n",
    "        percent = int(pred_flat.shape[0]*0.1)\n",
    "        value = pred_flat[percent]\n",
    "\n",
    "        points = (prob>0.5) & (prob<value)\n",
    "        result = prob * points\n",
    "        \n",
    "        results[i] += result\n",
    "        \n",
    "    return results "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "id": "recognized-haiti",
   "metadata": {},
   "outputs": [],
   "source": [
    "device = torch.device('cpu')\n",
    "\n",
    "probs_torch = torch.from_numpy(probs)\n",
    "results = cal_low_conf_pixels(probs_torch,device)\n",
    "\n",
    "results_numpy = results.numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "id": "joined-imaging",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x1080 with 8 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "rows = 2\n",
    "cols = 4\n",
    "fig,ax = plt.subplots(rows,cols,figsize=[15,15])\n",
    "start_with = 0\n",
    "\n",
    "for i in range(rows*cols):\n",
    "    pred = results_numpy[i]\n",
    "\n",
    "#     mask = pred>0.5\n",
    "#     pred_flat = pred[mask].flatten()\n",
    "#     pred_flat.sort()\n",
    "#     percent = int(pred_flat.shape[0]*0.1)\n",
    "#     value = pred_flat[percent]\n",
    "\n",
    "#     points = (pred>0.5) & (pred<value)\n",
    "#     result = pred * points\n",
    "    \n",
    "    #print(int(i/cols),int(i%cols))\n",
    "    ax[int(i/cols),int(i % cols)].set_title('slice %d,value:%f' % (i,value))\n",
    "    ax[int(i/cols),int(i % cols)].imshow(pred,cmap='gray')\n",
    "    ax[int(i/cols),int(i % cols)].axis('off')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "defined-kazakhstan",
   "metadata": {},
   "source": [
    "可以看出，低置信度的区域都集中在边缘，所谓低置信度，其实就对应难样本，\n",
    "所以有可能通过focal loss来改善边缘的预测"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "instant-arkansas",
   "metadata": {},
   "source": [
    "找到group3 outline的原因"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "visible-conversation",
   "metadata": {},
   "source": [
    "focal loss"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "applied-lingerie",
   "metadata": {},
   "source": [
    "分析为什么区域比true_mask大的原因，是不是可能医生的标注比较保守？"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "weighted-estimate",
   "metadata": {},
   "source": [
    "仰文 UNet 分割 数据集 第六附属医院"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "curious-avatar",
   "metadata": {},
   "source": [
    "可解释性，终身学习，异常检测，小样本，样本均衡"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "loaded-given",
   "metadata": {},
   "source": [
    "<mark>numpy/torch 对应sort()中axis参数的理解<br>具体见OneNote</mark>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "academic-composite",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[1, 1, 0, 1],\n",
       "        [1, 0, 0, 1],\n",
       "        [0, 1, 1, 0]])"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "d = torch.randint(0,2,(3,4))\n",
    "d"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "psychological-market",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(8, 1, 512, 512)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probs = np.load('/GPUFS/sysu_rxwang_1/defeng/backup4-DICE-UNet/probs.npy')\n",
    "probs.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "female-embassy",
   "metadata": {},
   "outputs": [],
   "source": [
    "probs.sort(axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "level-edgar",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[-1.5205, -0.1559,  0.2058],\n",
       "        [ 0.3902,  0.9185,  2.5421],\n",
       "        [ 0.0195, -0.4215, -1.1138],\n",
       "        [ 0.7582, -1.0176, -1.1639]])"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import torch\n",
    "a = torch.randn((4,3))\n",
    "a"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "raising-warren",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[[1, 0, 2],\n",
       "         [0, 1, 2]],\n",
       "\n",
       "        [[0, 2, 3],\n",
       "         [1, 0, 1]],\n",
       "\n",
       "        [[2, 1, 0],\n",
       "         [2, 3, 3]],\n",
       "\n",
       "        [[3, 3, 1],\n",
       "         [3, 2, 0]]])"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test = torch.randn((4,2,3))\n",
    "sorted_tensor,sorted_index = test.sort(axis=0)\n",
    "sorted_index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "accepting-wildlife",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([[-1.5205, -1.0176, -1.1639],\n",
       "         [ 0.0195, -0.4215, -1.1138],\n",
       "         [ 0.3902, -0.1559,  0.2058],\n",
       "         [ 0.7582,  0.9185,  2.5421]]),\n",
       " tensor([[0, 3, 3],\n",
       "         [2, 2, 2],\n",
       "         [1, 0, 0],\n",
       "         [3, 1, 1]]))"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sorted_tensor,sorted_index = a.sort(axis=0)\n",
    "sorted_tensor,sorted_index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "amino-header",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([[-1.5205, -0.1559,  0.2058],\n",
       "         [ 0.3902,  0.9185,  2.5421],\n",
       "         [-1.1138, -0.4215,  0.0195],\n",
       "         [-1.1639, -1.0176,  0.7582]]),\n",
       " tensor([[0, 1, 2],\n",
       "         [0, 1, 2],\n",
       "         [2, 1, 0],\n",
       "         [2, 1, 0]]))"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sorted_tensor,sorted_index = a.sort(axis=1)\n",
    "sorted_tensor,sorted_index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "technological-greensboro",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[-0.20514859, -0.82950229, -1.54462652],\n",
       "       [-1.33997226, -0.6520511 , -0.47617723],\n",
       "       [ 1.16392028,  1.11151317,  0.29175784],\n",
       "       [-1.45206545,  0.1465196 , -0.60927089]])"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np \n",
    "b = np.random.randn(4,3)\n",
    "b "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "mental-timeline",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[-1.45206545, -0.82950229, -1.54462652],\n",
       "       [-1.33997226, -0.6520511 , -0.60927089],\n",
       "       [-0.20514859,  0.1465196 , -0.47617723],\n",
       "       [ 1.16392028,  1.11151317,  0.29175784]])"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "b.sort(axis=0)\n",
    "b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "initial-female",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([[-0.093291  , -1.99267183],\n",
       "        [ 1.30498286, -1.39662903],\n",
       "        [ 1.98327035,  0.39252396]]),\n",
       " tensor([[-0.0933, -1.9927],\n",
       "         [ 1.3050, -1.3966],\n",
       "         [ 1.9833,  0.3925]], dtype=torch.float64))"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "c = np.random.randn(3,2)\n",
    "c_torch = torch.from_numpy(c)\n",
    "\n",
    "c.sort(axis=0)\n",
    "\n",
    "c,c_torch.sort(axis=0)[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "liked-marketing",
   "metadata": {},
   "source": [
    "# DICE+FOCAL LOSS"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "alien-ocean",
   "metadata": {},
   "source": [
    "<mark>下面的代码是错误的❌，正确的代码见dice_loss.ipynb</mark>"
   ]
  },
  {
   "attachments": {
    "image.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "id": "approximate-employer",
   "metadata": {},
   "source": [
    "![image.png](attachment:image.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "chief-spank",
   "metadata": {},
   "outputs": [],
   "source": [
    "probs = np.load('./probs.npy')\n",
    "#probs = probs.transpose(0,3,1,2)\n",
    "probs = torch.from_numpy(probs)\n",
    "probs.requires_grad = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "opponent-chancellor",
   "metadata": {},
   "outputs": [],
   "source": [
    "true_mask = torch.randint(0,2,probs.shape,dtype=torch.float32,requires_grad=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "derived-thought",
   "metadata": {},
   "outputs": [],
   "source": [
    "pred = probs>0.5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "initial-chart",
   "metadata": {},
   "outputs": [],
   "source": [
    "num_classes = 2 #2 includes background\n",
    "classes_points = []\n",
    "\n",
    "for i in range(num_classes):\n",
    "    classes_points.append(pred==i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 223,
   "id": "breathing-winning",
   "metadata": {},
   "outputs": [],
   "source": [
    "TP = torch.zeros((num_classes,probs.shape[0]),dtype=torch.float32)\n",
    "FP = torch.zeros((num_classes,probs.shape[0]),dtype=torch.float32)\n",
    "FN = torch.zeros((num_classes,probs.shape[0]),dtype=torch.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 224,
   "id": "verified-series",
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(num_classes):\n",
    "    TP[i] = ((true_mask*probs)*classes_points[i]).sum(axis=(1,2,3))\n",
    "    FN[i] = ((true_mask*(1.0-probs))*classes_points[i]).sum(axis=(1,2,3))\n",
    "    FP[i] = (((1.0-true_mask)*probs)*classes_points[i]).sum(axis=(1,2,3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 225,
   "id": "prescribed-gnome",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([[33371.2930, 33512.5742, 33726.1211, 34226.5469, 33598.1719, 33896.4102,\n",
       "          33506.8242, 33285.5312],\n",
       "         [ 1742.9943,  1125.8137,   351.3155,   632.7398,  1130.0732,   967.6789,\n",
       "           1548.3789,  1843.3257]], grad_fn=<CopySlices>),\n",
       " tensor([[33398.7344, 33764.6758, 33756.8594, 33990.7539, 33657.3047, 33972.7188,\n",
       "          33602.2383, 33476.4375],\n",
       "         [ 1707.4255,  1110.6901,   354.6981,   662.2122,  1102.6460,   933.7745,\n",
       "           1610.3724,  1760.8536]], grad_fn=<CopySlices>),\n",
       " tensor([[94956.6875, 95207.3594, 96700.8203, 96236.4297, 95571.8281, 95467.6094,\n",
       "          94972.1484, 94630.4688],\n",
       "         [ 1038.0059,   692.1862,   206.6845,   373.2602,   668.9267,   581.3210,\n",
       "            900.6210,  1056.6743]], grad_fn=<CopySlices>))"
      ]
     },
     "execution_count": 225,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "TP,FP,FN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 226,
   "id": "nuclear-cabinet",
   "metadata": {},
   "outputs": [],
   "source": [
    "alpha = beta = 0.5\n",
    "lamb = 0.5\n",
    "\n",
    "loss_dice = num_classes - (TP / (TP+alpha*FN+beta*FP)).sum(axis=0)\n",
    "loss_focal = ((((true_mask-probs)**2)*torch.log(probs)).mean(axis=(1,2,3))) * lamb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "id": "imposed-infrared",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([True, True, True, True, True, True, True, True])"
      ]
     },
     "execution_count": 201,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "((((true_mask-probs)**2)*torch.log(probs)).sum(axis=(1,2,3))) / (probs.shape[2]*probs.shape[3]) == ((((true_mask-probs)**2)*torch.log(probs)).mean(axis=(1,2,3)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 227,
   "id": "binary-portfolio",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss = (loss_dice.mean()-loss_focal.mean())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "anonymous-bathroom",
   "metadata": {},
   "outputs": [],
   "source": [
    "def DICEandfocalLoss(probs,true_mask,device,num_classes=2):\n",
    "\t\n",
    "\tpred = probs>0.5\n",
    "\tclasses_points = []\n",
    "\tfor i in range(num_classes):\n",
    "\t\tclasses_points.append(pred==i)\n",
    "\n",
    "\t\n",
    "\tTP = torch.zeros((num_classes),dtype=torch.float32).to(device)\n",
    "\tFP = torch.zeros((num_classes),dtype=torch.float32).to(device)\n",
    "\tFN = torch.zeros((num_classes),dtype=torch.float32).to(device)\n",
    "\n",
    "\tfor i in range(num_classes):\n",
    "\t\tTP[i] = (probs[:,i] * true_mask[:,i]).sum(axis=(1,2)).mean()\n",
    "\t\tFN[i] = ((1.0-probs[:,i]) * true_mask[:,i]).sum(axis=(1,2)).mean()\n",
    "\t\tFP[i] = (probs[:,i] * (1-true_mask[:,i])).sum(axis=(1,2)).mean()\n",
    "\t\t\n",
    "\t\n",
    "\talpha = beta = 0.5\n",
    "\tlamb = 0.5\n",
    "\n",
    "\tloss_dice = num_classes - (TP / (TP+alpha*FN+beta*FP)).sum()\n",
    "\tloss_focal = loss_focal = ((((true_mask-probs)**2)*torch.log(probs)).mean(axis=(1,2,3))) * lamb\n",
    "\t\n",
    "\tloss = loss_dice.mean()-loss_focal.mean()\n",
    "\t\n",
    "\treturn loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "indirect-pakistan",
   "metadata": {},
   "outputs": [],
   "source": [
    "if torch.cuda.is_available():\n",
    "    device = torch.device('cuda:0')\n",
    "else:\n",
    "    device = torch.device('cpu')\n",
    "    \n",
    "probs = probs.to(device)\n",
    "true_mask = true_mask.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "orange-progressive",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss = DICEandfocalLoss(probs,true_mask,device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "recorded-thing",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss.backward()"
   ]
  }
 ],
 "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": "3.7.9"
  },
  "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": 5
}