visualization.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os.path as osp\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn import linear_model\n",
    "\n",
    "root_dir = 'MY_ROOT_DIRECTORY'\n",
    "scores_dir = osp.join(root_dir, 'scores')\n",
    "tasks = ['full', 'scarce', 'reynolds', 'aoa'] # Task for which you want to visualize the results\n",
    "for task in tasks:\n",
    "    true_coefs = np.load(osp.join(scores_dir, task, 'true_coefs.npy'))\n",
    "    pred_coefs_mean = np.load(osp.join(scores_dir, task, 'pred_coefs_mean.npy'))\n",
    "    pred_coefs_std = np.load(osp.join(scores_dir, task, 'pred_coefs_std.npy'))\n",
    "\n",
    "    sns.set()\n",
    "\n",
    "    fig, ax = plt.subplots(2, 2, figsize = (20, 10))\n",
    "    # ax[2].scatter(true_coefs[:, 1], true_coefs[:, 0], label = 'True', color = 'black', marker = 's')\n",
    "    model_name = ['MLP', 'GraphSAGE', 'PointNet', 'GUNet'] # Model included in the results\n",
    "    for l, model in enumerate(model_name):\n",
    "        reg = linear_model.LinearRegression()\n",
    "        reg.fit(true_coefs[:, :1], pred_coefs_mean[:, l, :1])\n",
    "        x_zero = np.array([true_coefs[:, 0].min(), true_coefs[:, 0].max()])\n",
    "        y = reg.predict(x_zero[:, None])\n",
    "        ax[0, 0].plot(x_zero, y)\n",
    "        ax[0, 0].scatter(true_coefs[:, 0], pred_coefs_mean[:, l, 0], alpha = 0.5, linewidths = 1, marker = 'x', s = 20*np.ones_like(true_coefs[:, 0]), label = model)\n",
    "        \n",
    "        reg = linear_model.LinearRegression()\n",
    "        reg.fit(true_coefs[:, 1:2], pred_coefs_mean[:, l, 1:2])\n",
    "        x_one = np.array([true_coefs[:, 1].min(), true_coefs[:, 1].max()])\n",
    "        y = reg.predict(x_one[:, None])\n",
    "        ax[0, 1].plot(x_one, y)\n",
    "        ax[0, 1].scatter(true_coefs[:, 1], pred_coefs_mean[:, l, 1], alpha = 0.5, linewidths = 1, marker = 'x', s = 20*np.ones_like(true_coefs[:, 1]), label = model)\n",
    "\n",
    "        reg = linear_model.LinearRegression()\n",
    "        reg.fit(true_coefs[:, :1], pred_coefs_std[:, l, :1])\n",
    "        x = np.array([true_coefs[:, 0].min(), true_coefs[:, 0].max()])\n",
    "        y = reg.predict(x[:, None])\n",
    "        ax[1, 0].plot(x, y)\n",
    "        ax[1, 0].scatter(true_coefs[:, 0], pred_coefs_std[:, l, 0], alpha = 0.5, linewidths = 1, marker = 'x', s = 20*np.ones_like(true_coefs[:, 0]), label = model)\n",
    "\n",
    "        reg = linear_model.LinearRegression()\n",
    "        reg.fit(true_coefs[:, 1:2], pred_coefs_std[:, l, 1:2])\n",
    "        x = np.array([true_coefs[:, 1].min(), true_coefs[:, 1].max()])\n",
    "        y = reg.predict(x[:, None])\n",
    "        ax[1, 1].plot(x, y)\n",
    "        ax[1, 1].scatter(true_coefs[:, 1], pred_coefs_std[:, l, 1], alpha = 0.5, linewidths = 1, marker = 'x', s = 20*np.ones_like(true_coefs[:, 1]), label = model)\n",
    "        # ax[2].errorbar(pred_coefs_mean[:, l, 1], pred_coefs_mean[:, l, 0], xerr = pred_coefs_std[:, l, 1], yerr = pred_coefs_std[:, l, 0], fmt = 'x', capsize = 3, label = model)\n",
    "    ax[0, 0].plot(x_zero, x_zero, c = 'black', label = 'Identity')\n",
    "    ax[0, 0].set_xlabel('True ' + r'$C_D$')\n",
    "    ax[0, 0].set_ylabel('Mean predicted ' + r'$C_D$')\n",
    "    ax[0, 1].plot(x_one, x_one, c = 'black', label = 'Identity')\n",
    "    ax[0, 1].set_xlabel('True ' + r'$C_L$')\n",
    "    ax[0, 1].set_ylabel('Mean predicted ' + r'$C_L$')\n",
    "    ax[0, 0].legend(loc = 'best')\n",
    "    ax[0, 1].legend(loc = 'best')\n",
    "\n",
    "    ax[1, 0].set_xlabel('True ' + r'$C_D$')\n",
    "    ax[1, 0].set_ylabel('Std predicted ' + r'$C_D$')\n",
    "    ax[1, 1].set_xlabel('True ' + r'$C_L$')\n",
    "    ax[1, 1].set_ylabel('Std predicted ' + r'$C_L$')\n",
    "    ax[1, 0].legend(loc = 'best')\n",
    "    ax[1, 1].legend(loc = 'best');\n",
    "\n",
    "    fig.savefig(osp.join(scores_dir, task, 'coefs.png'), bbox_inches = 'tight', dpi = 150)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "sns.set()\n",
    "\n",
    "root_dir = 'MY_ROOT_DIRECTORY'\n",
    "scores_dir = osp.join(root_dir, 'scores')\n",
    "tasks = ['full', 'scarce', 'reynolds', 'aoa'] # Task for which you want to visualize the results\n",
    "\n",
    "for task in tasks:\n",
    "    model_name = ['MLP', 'GraphSAGE', 'PointNet', 'GUNet'] # Model included in the results\n",
    "    xs = [str(0.2), str(0.4), str(0.6), str(0.8)]\n",
    "\n",
    "    true_surf_coefs = []\n",
    "    for i in range (3):\n",
    "        true_surf_coefs.append(np.load(osp.join(scores_dir, task, 'true_surf_coefs_' + str(i) + '.npy')))\n",
    "    surf_coefs = []\n",
    "    for i in range (3):\n",
    "        surf_coefs.append(np.load(osp.join(scores_dir, task, 'surf_coefs_' + str(i) + '.npy')))\n",
    "    true_bls = np.load(osp.join(scores_dir, task, 'true_bls.npy'))\n",
    "    bls = np.load(osp.join(scores_dir, task, 'bls.npy'))\n",
    "\n",
    "    fig, ax = plt.subplots(len(surf_coefs), 2, figsize = (int(len(surf_coefs)*10), 30))\n",
    "    for i in range(len(surf_coefs)):\n",
    "        ax[i, 0].scatter(true_surf_coefs[i][0, :, 0], true_surf_coefs[i][0, :, 1], label = 'True', linewidths = .5, marker = 'x', s = 5*np.ones_like(true_surf_coefs[i][0, :, 0]))\n",
    "        ax[i, 1].scatter(true_surf_coefs[i][1, :, 0], true_surf_coefs[i][1, :, 1], label = 'True', linewidths = .5, marker = 'x', s = 5*np.ones_like(true_surf_coefs[i][1, :, 0]))\n",
    "        for j in range(len(surf_coefs[0])):\n",
    "            ax[i, 0].scatter(surf_coefs[i][j, 0, :, 0], surf_coefs[i][j, 0, :, 1], label = model_name[j], linewidths = .5, marker = 'x', s = 3*np.ones_like(surf_coefs[i][j, 0, :, 0]))\n",
    "            ax[i, 1].scatter(surf_coefs[i][j, 1, :, 0], surf_coefs[i][j, 1, :, 1], label = model_name[j], linewidths = .5, marker = 'x', s = 3*np.ones_like(surf_coefs[i][j, 1, :, 0]))\n",
    "        ax[i, 0].set_xlabel(r'$x/c$')\n",
    "        ax[i, 1].set_xlabel(r'$x/c$')\n",
    "        ax[i, 0].set_ylabel(r'$c_p$')\n",
    "        ax[i, 1].set_ylabel(r'$c_\\tau$')\n",
    "        ax[i, 0].set_title('Pressure coefficient')\n",
    "        ax[i, 1].set_title('Skin friction coefficient')\n",
    "        ax[i, 0].legend(loc = 'upper right')\n",
    "        ax[i, 1].legend(loc = 'upper right')\n",
    "        ax[i, 0].invert_yaxis()\n",
    "        ax[i, 1].set_yscale('log')\n",
    "    fig.savefig(osp.join(scores_dir, task, 'surf.png'), bbox_inches = 'tight', dpi = 150)\n",
    "\n",
    "    fig, ax = plt.subplots(4, len(bls), figsize = (int(len(bls)*10), 40))\n",
    "    for i in range(len(bls)):\n",
    "        j = 0\n",
    "        ax[j, i].plot(true_bls[i, j, 1], true_bls[i, j, 0], label = 'True')\n",
    "        for k in range(len(bls[0])):\n",
    "            ax[j, i].plot(bls[i, k, j, 1], bls[i, k, j, 0], label = model_name[k])\n",
    "        ax[j, i].set_xlabel(r'$u/U_\\infty$')\n",
    "        ax[j, i].set_ylabel(r'y/c')\n",
    "        ax[j, i].set_title(r'$x = $' + xs[j])\n",
    "        ax[j, i].legend(loc = 'best')\n",
    "\n",
    "        j = 1\n",
    "        ax[j, i].plot(true_bls[i, j, 2], true_bls[i, j, 0], label = 'True')\n",
    "        for k in range(len(bls[0])):\n",
    "            ax[j, i].plot(bls[i, k, j, 2], bls[i, k, j, 0], label = model_name[k])\n",
    "        ax[j, i].set_xlabel(r'$v/U_\\infty$')\n",
    "        ax[j, i].set_ylabel(r'y/c')\n",
    "        ax[j, i].set_title(r'$x = $' + xs[j])\n",
    "        ax[j, i].legend(loc = 'best')\n",
    "\n",
    "        j = 2\n",
    "        ax[j, i].plot(true_bls[i, j, 3], true_bls[i, j, 0], label = 'True')\n",
    "        for k in range(len(bls[0])):\n",
    "            ax[j, i].plot(bls[i, k, j, 3], bls[i, k, j, 0], label = model_name[k])\n",
    "        ax[j, i].set_xlabel(r'$\\nu_t/\\nu$')\n",
    "        ax[j, i].set_ylabel(r'y/c')\n",
    "        ax[j, i].set_title(r'$x = $' + xs[j])\n",
    "        ax[j, i].legend(loc = 'best')\n",
    "\n",
    "        j = 3\n",
    "        ax[j, i].plot(true_bls[i, j, 1], true_bls[i, j, 0], label = 'True')\n",
    "        for k in range(len(bls[0])):\n",
    "            ax[j, i].plot(bls[i, k, j, 1], bls[i, k, j, 0], label = model_name[k])\n",
    "        ax[j, i].set_xlabel(r'$u/U_\\infty$')\n",
    "        ax[j, i].set_ylabel(r'y/c')\n",
    "        ax[j, i].set_title(r'$x = $' + xs[j])\n",
    "        ax[j, i].legend(loc = 'best')\n",
    "    fig.savefig(osp.join(scores_dir, task, 'bl.png'), bbox_inches = 'tight', dpi = 150)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.9.12 ('airfrans': conda)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.11"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "6addcc3e3096a2fd73dc6d001c61ac497c8143e1cb3b3cdf61764a4f6e624233"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}