dice_loss_bce_loss.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Matplotlib created a temporary config/cache directory at /tmp/matplotlib-mtv3jcwm because the default path (/home/ruthvik/.config/matplotlib) is not a writable directory; it is highly recommended to set the MPLCONFIGDIR environment variable to a writable directory, in particular to speed up the import of Matplotlib and to better support multiprocessing.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import os\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = '-1'\n",
    "import torch\n",
    "from copy import deepcopy\n",
    "from torch.nn import functional as F\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Make a sample \"GT mask\"\n",
    "```\n",
    " [[1. 1. 1. 1. 1. 1. 1.]\n",
    "  [1. 0. 0. 1. 0. 0. 1.]\n",
    "  [1. 0. 0. 1. 0. 0. 1.]\n",
    "  [1. 1. 1. 1. 1. 1. 1.]\n",
    "  [1. 0. 0. 1. 0. 0. 1.]\n",
    "  [1. 0. 0. 1. 0. 0. 1.]\n",
    "  [1. 1. 1. 1. 1. 1. 1.]]\n",
    " ```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1. 1. 1. 1. 1. 1. 1.]\n",
      " [1. 0. 0. 1. 0. 0. 1.]\n",
      " [1. 0. 0. 0. 0. 0. 1.]\n",
      " [1. 0. 0. 0. 0. 0. 1.]\n",
      " [1. 0. 0. 0. 0. 0. 1.]\n",
      " [1. 0. 0. 1. 0. 0. 1.]\n",
      " [1. 1. 1. 1. 1. 1. 1.]]\n"
     ]
    }
   ],
   "source": [
    "def get_gt_mask():\n",
    "    gt = np.zeros(shape=(7,7))\n",
    "    gt[:,0:2] = 1\n",
    "    gt[0:2,:] = 1\n",
    "    gt[:,-2:] = 1\n",
    "    gt[-2:,:] = 1\n",
    "    gt[2:5, 1:6] = 0\n",
    "    #gt[:,3] = 1\n",
    "    gt[1,1:3] = 0\n",
    "    gt[1,4:6] = 0\n",
    "    gt[5,1:3] = 0\n",
    "    gt[5,4:6] = 0\n",
    "    #gt[3,:] = 1\n",
    "\n",
    "    return gt\n",
    "\n",
    "gt = get_gt_mask()\n",
    "print(gt)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Make a bunch of \"predicted\" masks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 186,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0 0 0 0 0 0 0 1 1 1 2 2 3 3 4 4 5 5 5 6 6 6 6 6 6 6]\n",
      "[0 1 2 3 4 5 6 0 3 6 0 6 0 6 0 6 0 3 6 0 1 2 3 4 5 6]\n"
     ]
    }
   ],
   "source": [
    "predicted_masks = []\n",
    "row, col = np.where(gt>=1)\n",
    "print(row)\n",
    "print(col)\n",
    "new_gt = deepcopy(gt)\n",
    "for i in range(len(row)):\n",
    "    new_gt = deepcopy(new_gt)\n",
    "    new_gt[row[i], col[i]] = 0\n",
    "    predicted_masks.append(new_gt)\n",
    "predicted_masks.append(np.logical_not(gt).astype(np.float32))  ##completely oppsite case"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Define dice loss\n",
    "* [See](https://arxiv.org/pdf/2007.08921.pdf) Equation 3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![alt text](dice_loss.png \"Dice Loss\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 187,
   "metadata": {},
   "outputs": [],
   "source": [
    "eps = 1.0#1e-9\n",
    "def dice_loss(gt_mask, pred_mask):\n",
    "    num = 2*(gt_mask*pred_mask).sum() + eps\n",
    "    den = pred_mask.sum() + gt_mask.sum() + eps\n",
    "    D = 1 - (num / den)\n",
    "    return D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 188,
   "metadata": {},
   "outputs": [],
   "source": [
    "dice_losses = list(map(dice_loss, [gt]*len(predicted_masks), predicted_masks))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Define bce loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 189,
   "metadata": {},
   "outputs": [],
   "source": [
    "def bce_loss(gt_mask, pred_mask):\n",
    "    #print(gt_mask)\n",
    "    #print(pred_mask)\n",
    "    bce_loss = F.binary_cross_entropy_with_logits(torch.Tensor(pred_mask), torch.Tensor(gt_mask))\n",
    "    return float(bce_loss.numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "metadata": {},
   "outputs": [],
   "source": [
    "bce_losses = list(map(bce_loss, [gt]*len(predicted_masks), predicted_masks))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 191,
   "metadata": {},
   "outputs": [],
   "source": [
    "lmbda = 1.0\n",
    "combined_losses = np.array(bce_losses) + lmbda*np.array(dice_losses)\n",
    "combined_losses = combined_losses.tolist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 192,
   "metadata": {},
   "outputs": [
    {
     "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": [
    "fig, ax = plt.subplots(1,1)\n",
    "ax.plot(range(len(dice_losses)), dice_losses, label='dice_losses', color='green', linewidth=3)\n",
    "ax.plot(range(len(bce_losses)), bce_losses, label='bce_losses', color='red', linewidth=3)\n",
    "ax.plot(range(len(bce_losses)), combined_losses, label='bce + dice', color='blue', linewidth=3)\n",
    "ax.set_ylabel('Loss')\n",
    "ax.set_xlabel('Predicted masks in increasing mismatch order')\n",
    "ax.set_yticks(np.array(list(range(0,200,10)))/100.0)\n",
    "ax.legend()\n",
    "ax.grid()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "pytorch_1121_cu113_unet",
   "language": "python",
   "name": "pytorch_1121_cu113_unet"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.6"
  },
  "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": 2
}