qual_eval.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_ = \"disa\"  # \"disa\", \"isa\", or \"sa\"\n",
    "if model_ == \"disa\":\n",
    "    from models.disa import *\n",
    "elif model_ == \"isa\":\n",
    "    from models.isa import *\n",
    "elif model_ == \"sa\":\n",
    "    from models.sa import *\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "from utils.dataset import *\n",
    "\n",
    "device = torch.device(\"cuda\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = \"tetr\"  # \"tetr\", \"mds\", or \"clevr\"\n",
    "\n",
    "if dataset == \"tetr\":\n",
    "    test_set = Dataset(\"tetrominoes\", \"data/tetrominoes/\", \"test\", noise=False, crop=False, resize=False, proppred=False)\n",
    "    num_slots = 4\n",
    "    num_iterations = 3\n",
    "    slots_dim = 32 if model_ == \"disa\" else 64\n",
    "    resolution = (35, 35)\n",
    "    encdec_dim = 32 \n",
    "    smallarch = True\n",
    "    if model_ == \"isa\":\n",
    "        learned_factors = False\n",
    "        scale_inv = False\n",
    "elif dataset == \"mds\":\n",
    "    test_set = Dataset(\"multidsprites\", \"data/multidsprites/\", \"test\", noise=False, crop=False, resize=False, proppred=False)\n",
    "    num_slots = 6\n",
    "    num_iterations = 3\n",
    "    slots_dim = 32 if model_ == \"disa\" else 64\n",
    "    resolution = (64, 64)\n",
    "    encdec_dim = 32 \n",
    "    smallarch = True\n",
    "    if model_ == \"isa\":\n",
    "        learned_factors = True\n",
    "        scale_inv = True\n",
    "elif dataset == \"clevr\":\n",
    "    test_set = Dataset(\"clevr\", \"data/clevr6/\", \"test\", noise=False, crop=\"center\", resize=True, proppred=False)\n",
    "    num_slots = 7\n",
    "    num_iterations = 3\n",
    "    slots_dim = 32 if model_ == \"disa\" else 64\n",
    "    resolution = (128, 128)\n",
    "    encdec_dim = 64\n",
    "    smallarch = False\n",
    "    if model_ == \"isa\":\n",
    "        learned_factors = False\n",
    "        scale_inv = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1fa2057a980>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sample_idx = 0\n",
    "sample = test_set.__getitem__(sample_idx)\n",
    "image = sample[\"image\"].unsqueeze(0).to(device)\n",
    "masks = sample[\"mask\"].unsqueeze(0).to(device)\n",
    "plt.grid(False); plt.axis(\"off\")\n",
    "plt.imshow(image.squeeze(0).cpu().permute((1,2,0)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "if model_ == \"sa\":\n",
    "    model = SlotAttentionAutoEncoder(resolution, num_slots, num_iterations, slots_dim, encdec_dim, smallarch, learned_slots=False).to(device)\n",
    "    model.eval()\n",
    "    model.encoder_cnn.encoder_pos.grid = model.encoder_cnn.encoder_pos.grid.to(device)  # model.to(device) do not move\n",
    "    model.decoder_cnn.decoder_pos.grid = model.decoder_cnn.decoder_pos.grid.to(device)  # these tensors automatically\n",
    "    if dataset == \"tetr\":\n",
    "        ckpt = \"checkpoints/tetrominoes/tetr_sa_106ep.ckpt\"\n",
    "    elif dataset == \"mds\":\n",
    "        ckpt = \"checkpoints/multidsprites/mds_sa_266ep.ckpt\"\n",
    "    elif dataset == \"clevr\":\n",
    "        ckpt = \"checkpoints/clevr6/clevr6_sa_273ep.ckpt\"\n",
    "    model.load_state_dict(torch.load(ckpt, map_location=device)[\"model_state_dict\"], strict=True)\n",
    "    texturestruction, textures, masks, slots, attn = model(image)\n",
    "elif model_ == \"isa\":\n",
    "    model = ISA(resolution, num_slots, num_iterations, slots_dim, encdec_dim, smallarch, learned_slots=True, bilevel=True, learned_factors=learned_factors, scale_inv=scale_inv).to(device)\n",
    "    model.eval()\n",
    "    model.slot_attention.pos_emb.grid = model.slot_attention.pos_emb.grid.to(device)\n",
    "    model.dec_pos_emb.grid = model.dec_pos_emb.grid.to(device)\n",
    "    if dataset == \"tetr\":\n",
    "        ckpt = \"checkpoints/tetrominoes/tetr_isa_106ep.ckpt\"\n",
    "    elif dataset == \"mds\":\n",
    "        ckpt = \"checkpoints/multidsprites/mds_isa_266ep.ckpt\"\n",
    "    elif dataset == \"clevr\":\n",
    "        ckpt = \"checkpoints/clevr6/clevr6_isa_273ep.ckpt\"\n",
    "    model.load_state_dict(torch.load(ckpt, map_location=device)['model_state_dict'], strict=True)\n",
    "    reconstruction, textures, masks, slots, attn = model(image)\n",
    "elif model_ == \"disa\":\n",
    "    model = DISA(resolution, num_slots, num_iterations, slots_dim, encdec_dim, smallarch, learned_slots=True, bilevel=True, learned_factors=True, scale_inv=True).to(device)\n",
    "    model.eval()\n",
    "    model.slot_attention.pos_emb.grid = model.slot_attention.pos_emb.grid.to(device)\n",
    "    model.enc_pos_emb.grid = model.enc_pos_emb.grid.to(device)\n",
    "    model.dec_pos_emb.grid = model.dec_pos_emb.grid.to(device)\n",
    "    if dataset == \"tetr\":\n",
    "        ckpt = \"checkpoints/tetrominoes/tetr_disa_run2_106ep.ckpt\"\n",
    "    elif dataset == \"mds\":\n",
    "        ckpt = \"checkpoints/multidsprites/mds_disa_266ep.ckpt\"\n",
    "    elif dataset == \"clevr\":\n",
    "        ckpt = \"checkpoints/clevr6/clevr6_disa_273ep.ckpt\"\n",
    "    model.load_state_dict(torch.load(ckpt, map_location=device)['model_state_dict'], strict=True)\n",
    "    reconstruction, textures, masks, slots, edge_mask, attn = model(image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA0YAAAI+CAYAAACYIhoAAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/NK7nSAAAACXBIWXMAAAsTAAALEwEAmpwYAABnBklEQVR4nO3debxkV1nv/+9Tw5l6zDxA0h0IoCCXqAyKkDRjLiCCPycURByuqD/w4k9RFAfgonK9egUvcBEvMgrKVRkUNECSTgAFmUIQCUPSp7uTTic9nZ7OVMP6/VHVtZ61+lR1nXPqTLU/79erX732WXvv2lX11K69aj1rbQshCAAAAACKrLTWBwAAAAAAa42GEQAAAIDCo2EEAAAAoPBoGAEAAAAoPBpGAAAAAAqPhhEAAACAwhuahpGZnTKzBw1gP68ys/cM4piAYWBmu8zs7rU+DqBfxCw2IuIWG80wxuyGaxiZ2aSZzbQbQmf+XR5C2BxCuGuFH3uXmTXbj3nSzL5uZj+9ko+JjSmL04Nm9g4z27zWx9WNmT3CzD5mZsfMbMrMvmBmz1zCft5hZq89xzr/zcy+YmZ1M3vVkg8aA0XMdq2/2MzeZ2YHzOy4mX3azB63vKPHoBC3Pde52cwOmdkJM/uymT1n6UeOQSFm+1r3OjML/a4/SBuuYdT27HZD6My/A6v42AdCCJslbZX0K5L+wswetoqPj43j2e1YuUbSd0r6zbU9nJ7+QdLHJV0i6WJJvyzpxAo91rck/bqkj6zQ/rF0xOzZNkv6nKTvlnS+pHdK+sh6vpApIOJ2Yf9V0mUhhK2Sfl7Se8zsshV6LCwOMduFmVUlvUHSZ1fqMXrZqA2js7Rblle3y+8wszeZ2UfaPTufNbMHu3XfYGb727+ifMHMnrjYxwstH5V0VNJ/au+3ZGavMLM7zeyImb3fzM53j/sEM/uXdot7v5m9qP33bWb2rvYvO3vN7LfNrNSue5GZfcrM/rjdWt9jZs9Y1ouFVRVCOCjpBrVOgJIkM/seFwtfNrNdru58M3t7+xfqY2b2QVf3X8zsW2Z21Mw+bGaXu7pgZr9gZt9sb/cmM7NzHZ+ZXSjpKkl/EUKYb//7dAjhU13W/3Yz290+9q+a2Q+0//7zkp4v6dfbv4T9Q5fX450hhH+SdPJcx4a1Qcwmr8VdIYT/GUK4N4TQCCG8VdKIJH4QW2eI27Nej9tDCPUzi5Kqkq4413Fi9RCzC/pVSR+TdMe5jm9FhBA21D9Jk5KeusDfg6Sr2+V3qNVgeaykiqS/kvTXbt0XSLqgXferkg5KGmvXvUrSe7o89i5Jd7fLJUk/IKkp6Tvbf3uZpM9IeqCkUUl/Lul97bor1boQ/HG1Tk4XSLqmXfcuSR+StEXSTknfkPSz7boXSapJ+i+SypJ+UdIBSbbW7wX/+ovTdjx8RdIb2ssPkHRE0jPbcfS09vJF7fqPSPobSee1Y+W69t+fLOmwpO9qx9f/knRr9hn4R0nb2/F2SNJ/7uNYTdI329s+V9IlPeK+qlaPz2+pdXH45HZcP6xd/w5Jr+3zNXqPpFet9XvFP2K235htr3+NpFlJ29b6PeMfcdvHY/5jO16DpH+WVFrr96zo/4jZno+3Q61r4M39xvjA35+1DpAlBtQpSVPtfx90b7pvGP0ft80zJd3RY5/HJD2qXX6VejeMmu3HnZPUkPQyV/81SU9xy5ep1aipqNVN+oEF9llu7+vh7m8vlrS7XX6RpG+5uon2c710rd8L/vUVpyfb79eNkra3635D0ruz9W+Q9FPtmGlKOm+Bfb5N0h+55c3t+NrZXg6SnuDq3y/pFX0e7wMlvVHSne3Hv1XSQ9p1/sT3RLV+SCi5bd+ndgNnMScy0TBaV/+I2b4ec6taFzG/udbvF/+I20XEbVXSMyT9ylq/X/wjZs/xWB+S9GP9rr8S/zZqKt1zQwjb2/+e22Wdg648rVaQSJLM7FfN7GvWGkg7JWmbpAv7fOwDIYTtan1B/plaLeIzdkj6QLsLcUqthlJDrbzMK9QKqtyFarWs97q/7VXrV4OznksIYbpdJL99/XtuCGGLWieOb1OMsR2SfuRMnLRj5QlqnfSukHQ0hHBsgf1dLhcnIYRTav2StGCsKIv7XkIId4cQXhJCeHD7+E6r1ZO50DHsDyE03d/yeMXGRcx2YWbjauXafyaE8IeL2RYrjrjt/Zi10Epfvv5MahPWHDGbMbNnS9oSQvibftZfKRu1YbRk1hpP9BuSflStVvd2ScfV6i7sWwhhrr2fR5rZc9t/3i/pGa7Rtj2EMBZCuKdd9+AFdnVYrVb9Dve3KyXds5jjwfoVQrhFrV8+/rj9p/1q/SLk42RTCOF17brzzWz7Ars6IBcnZrZJrZTMgcZKCGG/pDdJ+o4ux3CFtcfAtfl4DYM8FqwNYjZlZqOSPtje5sXLOlisGOL2nCpa+DoEa4SYTTxF0qOtNVPfQUk/JullZvahZR72ohSuYaTWOJ66WvmVFTP7XbV6fxYthDAv6U8k/W77T2+R9PtmtkOSzOwii9Nj/pWkp5rZj5pZxcwuMLNrQggNtbo0f9/MtrS3/f/USjPC8Hi9pKeZ2TVqvbfPNrPrzaxsZmPWmgr+gSGEeyX9k6Q3m9l5ZlY1s2vb+3ivpJ82s2vaF2p/IOmzIYTJ5RxY+3FebWZXW2sCkQsl/Yxa4+Vyn1Xr16Jfbx/bLknPlvTX7fr7JPW8n1h7uzG1zj+V9vMvL+c5YEW8XsTsmRmS/lbSjKQXZr+GYv15vYhbmdm3mdkzzGy8vf0LJF0r6ZblPAesiNeLmJWk35H0ULXGcV4j6cOS/kLSqt4Wp4gNoxvUCqxvqNXFN6tWK3yp/lLSle0uwDeo9UZ+zMxOqhU4j5OkEMI+tcY6/apaE0PcJulR7X28VK1gukvSp9QK8L/s58HN7C1m9pZlHD9WQQjhkFpdz7/T/sXlOWoNUDykVvy9XPHz+JNq9SLeIel+tSb1UAjhRrVOHH8n6V61fvl7Xj+Pb2ZPNLNTXarn1Zr04xNqTcH572qNe3vRAs9jXq1JR56hVm/nm9W6WDwze8zbJD3cWikAH+zyeH+h1kXmj0t6Zbv8k/08D6weYrbj8ZK+X9LTJU1ZvH/eomczxcojbuOhqDVm+n61nvt/VWvsxhf7eR5YPcRsZ/uTIYSDZ/6pdW1wOoRwtJ/nMSgWApkvAAAAAIqtiD1GAAAAAJCgYQQAAACg8GgYAQAAACg8GkYAAAAACo+GEQAAAIDCqwx6h2bGNHdDKoSwqJvgbhgv/ViM2Tden1Ve0yl9SvcmNU/UxW5pxJXzl8l61HWTrxd61NVc+bArn+ixj/xjOu3KdVfOfzsZ6VKWpPEuddUe+3hgVvc0V/5uV351tt4n1Y+hjVlxrh1mQxu3v3ddjNnXfD2r9LcTvD+p+Z8a7ZRf68rHzjo/9XrZ+v249NqHv92aP6/VsvX8OW42q/OXXee5cj4j8YQr58/zdJfHym8Hd2eXbaTWvT/POKLlGtaY/dkLdnQC53HT6Xfe1pEtnfL9jUZSd2B0U6e8rxZf+xDmkvUabrke0n2cdHdMu6cZF/bU0/VmK+6lr2XxZu57uekfezxbz38+6mld8LE+78oz6XruM1HN9j/mrp/GFJ/L5mwPmzdv65TL56Wf2arF7cZL8b24OKSfv1rFvd5T6W3n9j4ufkZu+0h9wZilxwgAAABA4Q38Pkb8ijm8hvUXoX9xPyW+M6v7Xlf+zqzuoa6c/fayAflfXE66cv6r7r+48t9ldf/aKfm+qn/O1tp9Ryx/82Rad8djYvlu9/fr/z5d74ZXxvKzrkzrDt4Qy5/XcMasxLl2mA3rufYFMx/uxOwF//atpO6q77qvUy5vuSipm1D8ZX6b4q/e5awXqOp6eypZQkwpWY7rNZT++t5Iyulvx3W33HA9NXPZcdSTX9zTX/DNnR3H3fMqZ7/SV3S5W0pfj7rivTiD224i6zEa0dc65c26I6m7WA/ulD+gP+mUdz96KlnvI19wC9nP+9d8KZa/dPVwxuyjn/i9nTf35N1pT2Zz06FOuaz0y+w8F25bLPakjI2kGRTlalyxWk1jdmQkxlvVdVbVK+n7XJuLvVPzM2mvVhhz5fEYK83s26NcilcxoZz2slgpble1+NhjzU3JeqdPxe0qc+k+5txrd9riZ6K2Iz3eubnYY5R1oEnjsRe1tCW+jqeaaWDOb4qhWDp6PKmrPDL2XN3+uv30GAEAAADAQmgYAQAAACg8GkYAAAAACm/gs9IBG81rXfmfPp7WvXW7W5hOZ5PRhS6nePb8WD6RrTfl0lgbWUpr2S+73yks+2j61fKRJcHl8o7vieWHZLPSXXFZLL89m03m0wdj+XKXr3vnV9L1dsTcdj31xWndY94by9t2xnJ+lvm2WLw2q/p1V/5BV35gtuKbPhtncLr1/31vUvcFvVQA1p/P/Pc4tuDOPT+WVv6vOF5Dc9n56YGTsdxw57u5bDY48+fCdIyDxtyJ8zx3Qr0iOyef78Y81LKZN0/GGfF02g3emM5m3jzgnsvhbCBlxY1OvTCOMVK4O12v7vYxMp/WnXIzzE2518rvT5K+w2+XPZcvbI/lD7l9pMNokq+eq06ldbd91i1craFUPxFn79s+uiOpm6nH7+x7ZkeTuq+P+AEy7n3IJyn043ksi9lkQtu4v7HR9LEuPs/NbHfBPUldYyTGfd2NRarPphcSFtwYt2yyttJMfJ5jEzGOtm6bStar7Yv7CFNpsIxeGI/Z7ov7qN6RDiSyahzrNFJNx1JVRuLsuduPxP1f0kw/O3s3ubFPl6fP5d77zt0fRI8RAAAAgMKjYQQAAACg8EilQ+FN/lMsX/xHad2z3Sypj39x2n29aSwuH3b3ep0aSyfv9lkWp7KfIo67nnN/H7LjWY/6CdfrfSLLMplxv280R+MUrBfMpV3lF94Tu5SPPSPdx6VXXdopP2f/73XKP/S6NDXjYv18p7xflyV1t7nyp105v/WhS0A56/aubrZuNXV7XNjyrmS92p2f75T/Zn86De37BWA9uvJf39cpv+nwl5O6Bz4i5hjNfOxwUvedV8U5+ctbXUrYxdmNEqoubSZPb6u5db8WT9hh+vxkta9PxJPvZy9LT8RfuCym/Ry4NJ5fRy/bmqy3Y8cDOuUtI19L6k6fiOfUo+5GD1NKj2PW3Uz0uC5M6qYsphSV6u4LIZu2uLI5pnCd/uZtSd2JRz2yU770JR/ulJ9/Z5oCddHf/G2nvPXE+5K6n8u+R4ZR/UicRn7TpjTt66qJmLL1mAvS64Ntbhr16mz8FhyfSW+0u30uvkcTIU2ZrDfidne4G8F+vp7G5Yl7LumUK83ssn573M7cNO/zpXQftdm4XLL0QsXN5K0Jl+Y6diiNldMuDbW+Ld3/+ET8PM6MxPXCwXS90VJ8DUqWptKV5uJrN1uLr82Ds+uqx26LB3zJpduSur/dee5mDz1GAAAAAAqPhhEAAACAwqNhBAAAAKDwGGOEwptx02fef0la97YLXPlL2YZ+Vkw/FCf/VPlU93JWV+qjLKVTdOeDdvwsrJVvdIr3NrN5vc97WCx/6aak6suPiHm+Nzw9DrT6JS3Cgdti+Z6vd4pXTqfjBXbuOd4pP8KOJHV7LorHXD0Wp7n93k/dm6z3Kw+JOfd/+rCnZwfy7n6PGMAqOhTiCfZ3L7gmqdt3dRwfdPAZ6ViL0kPjibjp5zsey85xJbdczqbhbrrbKJzeF8uHs1sSHIon6a370ymuL/nyRNzFoTgm6MBF2TTZD3OPNTGd1pXcOJXgtqtOpeuNuy+OkH2plNw4lZqbDryczadddrdh+NdsHu5Hx3OqzcfX7Y7P3Jas9vo98buitCep0r1+eOfjNZTK8/E1bZxI4/KwG/dzyOpJXakSX1OrxvFHmyppTPm6k5V0zNzsRFwuWYyH2slDyXrjJ10MnEinh68d9BcTwZXSi5ERN+aoNJE+lxk3tfdsI9bNhnQ9c2Ot6+clVTpdjTF8shb3Vx5PP6fNkXgcI9vSsXsqx89cw42D+vKWiWS1rz4oLjezC7K7Jr+oc6HHCAAAAEDh0TACAAAAUHik0qHwKn72zGNZpe/ZPp7V+VkmfbrcBdl6vuc8T6XLpuXui3VfPn9znD705In0Fts1P833jz853Ue/ZwLfhX/7rekuTkx1yg8tx/J33D+VrHdZOBq3yab0nDl6oFOemo+pH98YP5Csp3qcXvz4t/IXFcB6NHdislMeL/9HUnf+kZh7c/Bgek+CsYl4kp4ecbnEY9mJy6fPVbLfff1y1aWwXb49Xe/KuN6JbB8nKu7xDrkUudu+me7joNv/BSey43CHW45fDo3ZLEd63j2X0XSaaFXcurWY3vUAy16P0/H8On3pdyRVOx99ead8hUsJGzuWpjfXQ3ztj2WZi5rS0NvqUrYmyukLYCHWhWY6XXeou1Q6l3E2bul33qlGjKP5Rpp22XTvbcO9t6dH55L1wgPiY81cmV7EzLop7KuleCCn59J4q7sUufFsFvwTccZylS6PMTthaVzOjbvXo5y+HpVt7vEsXqdUT6cPZqfjPpvbppK60nj8XDVL8aIrbEvjfmYk7nP0aDptf6WcpectgB4jAAAAAIVHwwgAAABA4ZFKh8Irz7qUs9JYWul7tvOfEfyyn/AnT7nzPbl5KkK/qXR+uyyrwqfnHd0e09R0cTpjjB7hyvknPzkuN/XQt/4tXe8eN5vTifQO3peE+MQvOhT73uun0lly7ou96Bo5mL1YM3EGoFK94spZutxMTCW4sJFOf3NcDxCA9ad50qX51O5KK/fFGadGjqQnufGJeJ5oVl2qzViaJtOolBcsS8pS6Vxdvp6vq+bpeL7OpQpdkeUe7XOpTvPpTGYKMbWu0dgc/57PKFdzOdin0lm3VHXr1uOX1D2bsi+HE+4LbHRzUrV9Pr6mpf+IM/PVDqbn/OPH4mNtz7L9Tuap58OoEVPRSiH9wjafx255jrtfL6pn6825795mKb1AmHOpb3UXH6e3pd/tsy610o6nb8rJy2K56jarZZMlNly4zWSXQZXt8bvYQtxwbib97Mwec6l0l2bXH1vd5+BkLFfKaWz7VMPRkex5uten7D6b85adBxpxudxI6+b6uOiixwgAAABA4dEwAgAAAFB4NIwAAAAAFB5jjHKfcOUnd10rdWe2vHcJ+9iXLe/sczssm836vPfL0kqf4jqTTiErn//adNOzZnnYZ03R7XVLd+2ernz2GKOKm0J73I3ZqV+erueH+qRpt9IR9xrMfT2WD9yTrnd/3P/FtXQa2h0zcaDV9qNxvXopzbFvTsUxXaGRvqZhLr4gVXfn8Fo9e9K1+JtOuZq9L8pymwGsC6Va/Gw2p9OxEHY0nig3N9PPe/W0m0q4Ek+OzVp6CVN302k3yumJ1y83/Vik6hLHGI24cUTN7Bzkxzw0soGlwY8lcufQajbGKDlJb0qrau5k3nDbVbJz34wbB3oq/bKp3h0vOuqzh105vc1DyQ1N2pQOF5Wme31RDQkXipYPEg75oOFu4usUsm0ayT7TuqYLv7lqvLCYr6RxH7471p18ePrItcfE8vxT+jzcb6WL9b1urFOyjx7ftXuz5RvddcDNMcbqs9uS1UIpjt2rZ9dSVopvxrwbb1TNxxG5afBLlfxiJ18+Gz1GAAAAAAqPhhEAAACAwiOVLnerK/uuwLzH2Pf07+xzH3mvq9/nVf0cHFaET6Wr56l0rm42S5fwdyOvu7kuS1naV9n1B4fst4jg++ldgFi+nivnqXQj/vjdHJzTWT+0nxk7j+fDLpez6fZ3f5o7seVU3P8VzaNJ3QWHYkpHczauZ2PpcZTqMQWlMZM+GXN39zaXk1jLn3MtPoGRkOcj8nsPsB5V3XknT6O1YzG9ZixkU1zXY8rOWDlOjR3m03NLveZS5Mrp5Y1PpWu4VKTmfJZy59Ln8jS7ZLnmy9mXe9MtN7LnUnapak13fm1mqXSl6sLrSVJwKXLB1c1mx3Eq7nO8kZ4Xt546End/bzyX21x6G4aRsXg7hIkt2bl2ZvjTli34oO11EddjH67cM/ku+95vui0bbprvuS1pXI67jMzaf2T79B+ztbym9RmaZfdg1TSGGiPxsxlCdpDu2qpei5/9kKWJBp/yeiq7HiifO/2RKwgAAAAAhUfDCAAAAEDh0TACAAAAUHiMMcr1O/ukf+Xy5mW32Rd77bvbtM1Ycel03em4GdWnYnk+G2M06/Jkyy4fPB9j5KdQ7TnGyAdBj6CqZ4HkxwTNu/zwWhZUfnbtfMzO0QNuuzgYqXIizY/fUY+vz2XZ+KNwcioubIofkPKpdBxAcFPvhmwcQLni8oZdTnWtkT7naj2+HqP5683vPcC6NOKmri5Zem6puNshbCmPJnUzJ+MAhVrdndfGRpL1qpX42R/JzrWNuhs7NB/P141s6uNGbeFpvaV0jJEv17JxjqHm51nOzsP+fFVy5+5mOk22zL0GIRtjZG7AqBtvtHk+PU+emoljjCbm0tdq06mpTnna3ZZhNJu9fLx0fqc8VknP5ZVsPBK66DnIqMeYFzemyUJ8Y0rZFPClcnyAsfSjo9n1ck3rhxL5MMo+p8ENKG4ojdlkXHctxl6Yz6aYH4nrNefSuubWbMzfAriCAAAAAFB4NIwAAAAAFB6pdLl+m4q9uhPLXeqWNtMjVlhpdiou1I6klTWXfzaXpdJVu6TS5XlqZdeHnE/DbX5d3/ecBUfokUrn0zFOuVTA7Vlftu9BrmfP5b77O8WxWtzfZXOHk9WuvD+mcDRPpHeu10jsw6+edKkes2nXdbPp0uyyO7UHN2V506XSzWcv6bhLrRstMV03sBE05+O5YHQ+PS+ULJ6TquPpOW6zu3v97Gz8vNez89iMSymqVdM0nJJLfSv76YJrWcqdn/K7lKXZuX00Xblq6Xns9Hy8JYHm88ssdy4f9al0WYqPbXKbbErrSi6FzWJ5a5ZidWou5ixtOp59bxy+r1Mcm4mvaSmcl6y2qTHWKVeVpvSNz8wJ5+a/vkOv1Ln8mtCn0rkp4EeyfdhUjLHq1nQXs+vlmtZ/VH1afz17Lm65lk3XPTYXl4NLjbXsQJr1GPclyy4eJvLU+7NxBQEAAACg8GgYAQAAACg8Uuly33LlT7hyrybkE7Pl/a58c499+N6/7zv3oWFllJJZ6fJUOnc38rnptC5JpXMfJctSuxo+zSILgpKrK7vt7Kw+9VjMU+kqLm1txqXS5flnp93sLDP3pXVuRrnLTt/bKV91/6H0KNxMRpaloFTqLs1iPj6v/M7ypST1LZ+xKT5Pn3JQy57KWNOl0pXzVLpsWiUA64KVY+rVbDrBmSou7asSJpK6ajOmxU1UN3fK8/WxZL3xajwP1+ppmlfNLU6X3Qx1vVLuQpoip6af2S6W6+V0H/X5eE6e65XLFGJO0Uglm9nO5T6Xmuml2px7HTeX4/fSppl0Bq6LNsXH3houTOpG9k3F/Yf42s/ObkvWG53b5NZLU+lGZjjX9sN/l1nPC8GUudkOXZa5rJJ+ITZdSno1nyhwvVzTjruy/+xnqXR+JMJZV0E1n5Po0uosjUO/HErp/hulc8csPUYAAAAACo+GEQAAAIDCo2EEAAAAoPAYY5T7GVe+zpV7DPnQZFb3oCU87i1L2AYDUZp1Y3Rq6fTUqruc7blTaZ0fY1Ry03Vnd0GXn8q1mU/D7XPY3XalHvNenjWkxs19OeqO8XiWbFxxicIzx5OqCTeWaudUHHO1+XiaUz5/Kh5vqObTcPskfjf9azblrVw+dJ7CH5IxRu5xs3FVI+41qORjjObc7z3ZXcABrJ3xUhyLU8lOC43ZeK6ZP5aO52y4MUEjE/FDPT6xOVlv01gcDzOvdPzRfCluN1GN+5uvzWbrxfJMOR2P0ByJx1+qxLrxkI75GHdzEx9upmOdmrV4HBbieXd8NB105WcqLjfTuoY75W1qxuMfr6frjbrbTUw00pNhfV98fS68Or4ZU6fSk/JoJe5zvJGeazdN89t6P0KPJes1x3VYuNxopB+e8ub4ntXy68//7Mq7uj9Uz2vaPa785B778L6VLX+oy2Nl1zqh2X3Mla8L/nXLhg0FN445ZOOxQh/D4ohqAAAAAIVHwwgAAABA4ZFKl/MpbXe5cq8m5BUrdCxYFaXZqbgwn91hvO7SCuZPpHUzfnptl6Zg2Ty0fnrqZhZISYqEn6cy7153y3ksunQJTbsUlFPp8Y673VebaarKVSNxmu8LDsXtTh1N9+EPv5pNhxtm4gOYuVNLOUsR8eU8s7Dk14tr5ql04y4lsZpNc6spt5NLBGCdqPhp9i29/CiNxM97s5Ge5Oohpu3On3RTfp9MU33Lburt6sR4Ujfh0u42NeN04PNK15t3U29vyq6Q5usxRW7epQDVKun5aZObEriRzuStej2mMc9bPP5NWQaRuSnKQyO9jUR9JD7ehMsj2lbK0vaOxPP8bC3Nbaqeiuf87S7de+74w5L15qrxu61q1aRu4nSaTo2F+ego9RyX0Z+Q7SPMumm9p7KVB3FN+0lX3ttjO39YV/ao8489nz1/NxQhZFXNZLZuN113KZu+vBTjvjGSpdJVs/z9BdBjBAAAAKDwaBgBAAAAKDwaRgAAAAAKjzFGuS7TI/bUY7ZFrH8256aurqf55qq58Su1qbSu7N744BLJK1neddkt5wnnTT8+xuW+huw3C79YzuoabhzQ6Tj9a+nwsWS1Cy6K211YThOFdx6MObnTx+OU5Y1aOrdlNcQc+zCfPc9mzD+3spuSu8fnqPcYo2g+Swsed4OdKpZWlo4xxghYl9xHtZQNDfQXIyEbY1mxeB4acVNt5zP71tyYzfmpdLzN3Il4niyPxHEzlU3pOX/zeByLFBoTSZ2fAny+HI+4VknHUW5yy6PZdOAz8/FcO+Oe9baQjt8ZLcexEfPN9MUaddNmb3XjsS4aSW8pcWQ6nq8rh9OxFhMz98fjcN9zE9X09Zgub+mUrZG+4OOnuPgZpLNGH/kxNQoLliWl453TcBvMNa1fbvZYL3RZL6/z2zWy9Sxu2MhufVJ354FQiTusV9Ix07UQP/sh/QgrZGOeF0KPEQAAAIDCo2EEAAAAoPBIpcv57j6fRUQTcnjNxFS60SzVba7uunLnsum6ky5llwZRSVM45NMsmlnfczL/pPv7Wal0rrKc9VHX3ZS1czFX5WJLU+ku3h7X23HszqRu7sRRtxADvzJ3NFmv6aYvt3r6WpVcF3iv/Dk/1Wiw7J7gPjvR1c3V0/VG3Qeykk3lff7RPm5tDWD11eLnuJSlyaRpPunn3U9xXPJpdtnptOL+MFpKz6ENd06ad+lt80dmkvXmS/H7oDw6mtT5tLvRsXhrh2YjTYMbmYuXVqNhc1K3+dTpuJ07/tFsyuFNE3G5PpK+VifdFMcT7vYQFx9Pv3vmj8UUo/HZNOV4x4SbUvxYPMap+fT8OVr2Kd7ZefjUmLA8SQifdZsO93r770pLY7vksh9r2WiAgVzT+vDzrYY8Da7fffTIwGy4z4FlKaQ1NyzBKvFAms3j6Xql+Lktn/V65Dl+Z+NyHwAAAEDh0TACAAAAUHik0uX8ZF27XTnPDPJNyutW7GiwCsJMTDfY3EzvMG5+hpR6Nt2LxRSMqkvbmM/Sw8zdtb2UdQ033SxKFtxMbllfc8nNxjJRT1MdKtNudhaX0nHhzL3Jepfsi/uvl9I0u3LddV9Px1SP+Xo2g8usf+z0uYRyf1Pe+Gd21qx0PpXOdanXs+7vkWZ8LtWQpvSdf2+a/gJgnXBZWdYj3db6nj4r1TXlTukkov583czOW0nK3XyaZjfnvitCaSrubyxNpdNofKKbRtNUuh1Vl/rsJvY8dTT9nXrcnec3b0mP8fSWeD4sT8cnNmbpefK84/FcON3cmtQ1m3G2udr+uP/TR9LHmrsoPpdSNc2dGpvjt/VBCtlnoumuFxquXArpjLA2F+sq92Q79SG82z9Ytl6va9q7XPmWLtvky9+X1fmUtnKXsqTgvuub2fd+I7n2ibEYsgMJbvbcSilNIW3kU1kugKgGAAAAUHg0jAAAAAAUHg0jAAAAAIXHGKPci1y537FD+7Ll1wzmULA6yifi2KEtIc1HHXOLzXx22UbMN6804nSRM400abbk7sqcTyTddLd/L7lcWAvpmmU3ffd4Lf09ozoXc8zPc+ns51s61fbIqZhvPt+YSuvm4pijuVMxB7d8Oh2/03R3m1Y5m6szHzB05s/5H8zlBpeyMVfmp4aN6zWzqWxHXG572JI+7rZ0aBWA9aIRP6um/IS6tHFF/bKk3N9YpEpW58/XDTc+dN6NPZKkcDrWjY2dTOouK8XxD1c04vluZlM2bqQW93F1dqk26s61R6filNkXXJaspvK8GyM6kn6nzI+675EtcR93ltIxmndU4z7uq6bfB/VTi5mvGVLPmarP+rJsuj/48Ub17Bqgckmsm96R7fPprryrz8fOr2n98hN77MPbky1/sMt62bg4uc9YPfveb7hrq5K/Pshmja+4efBny+nY8Hojn7/7bPQYAQAAACg8GkYAAAAACo9UutxuV77TlfP+T7+cd11iQzl2MKY6bJtJu11nN2/vlMey2br9TdFL1ZgSUcs+VsHdpbqcpY80ay6VztWFbMrvUojTwY5lOX2zruv5wuZEXO9wOoXstpG4fHrm/KROR0/FYzoVp3EN1TRFpDFy0pWzqbz98fu0mJB2h/vpOEM+dWYjLlvFf8jS51JtxBc/jKYpKOFYevd3AOuDJem2eepczySjVeOPopyn2bnjL7vzWCWk58kTLi36rpPpPqbDVKf8Rff3A9lxpF836Tn0oa6y5Na8J0sjPuHK20+ndZe4rLiL3Slz6vSmZL358Ys75YnR9DhK0/y2Pki9kkmb7vu1OZKm/NfcG924O9twtysv9ZrWT9H9LVfOD9gvX9Vj//4SKQuhuptKvxTSa4ymW7dhva6XYrlWTl+rmT6aPUQ1AAAAgMKjYQQAAACg8GgYAQAAACg8xhj14puNvZI/Vzg1+kZ3IE/KpzjtYa8r5+meiCZOxo/BtuqWpG7s1LZYbqT5rmNjcZxLeey8Tnm+NJGs56ejLM+n41+absrXkpuq2rJ5vUuVOMZmopT+nlHaG8f9jLi6LaV0WsoLt8Rc8WMnNyd10/e743LTcNcbM8l6NTdteCObXrdRdjm/bnxQaKbjg0LTT7lZ6lpX9euNp/uwTfH1njmWTi87evR+AVjnVnZ27gHJfjt2J2Zz5eyUrCtm4vnq20fSc9eFbkjkIzTVKT9lEUf1jVo8v97tLkCevIjrg0l3HLc14/n06zovWe8rtXh+tbn0TZvb2vfDFVryqln/debH1LixxeVaei3S8Le9yMZCJ5Z6TVvqUddtu3y90KUum5Lb/C1BmmndvJtL38/yXcpuhDLvBhmFSjoG2U+z3w09RgAAAAAKj4YRAAAAgMIjlS7XrSuwVxNyhVPpbnbd45NZne84zw9j58ocztAZd1OhXlpJp7HeOh1Tzqam0rSyzW4K6bGR453y+Zel6Xj1kZhaNzebpn2F+ZgGV2vE47C091cjrkd5+3gajNPV2I08ozgH68jp9LGaLt+jMZdOydq0mD5Rrsa0urnRNKpqFtPzmkq7pOsjLpXOd3mX0n2UKvF4y1maScXNgV4qxefSKKev6WzpgrgQ0tNYZfpzAoAlcaer/NylEM9dweUGlZrpekdG4rlrb3YuP+hynQ67v+9V/65yj32zK09m6/V7fTA9Em/XUJpPz7WVufi9UWtkadyn+W29H+kk9aFrXZ7eVir5tHP3vWnZLTCSeax7PPhSr2n9sv/az1sQ/rHrWd1xdZGlf7p9NrN0ffNTdLvp8htnPZm4XTM/kJC/QGcjqgEAAAAUHg0jAAAAAIVHKl1u0pU/6cq9et+uW5lDOaPXRB+9JhlZH/cRX//Oa8TUsauy9/mSWryjeW0kncmtOrK9Ux45Ert1Lw07k/XutzhL2vnT6UduzuLtyMdHYgpDeSxNdTt8+lg8prkHJnXbSnFqoIrFfdSz2eBOucM/Vk5nx9s/G5M6Drru65B1c5dK8firlWy2OXc6CS69LTSz9MG6m7UvpDPn1ebicq0yFh93S/pYJxSf81w97SpvnEz3CQBLctYMYu7c6NKZwlmzicX8uaD8PJnl1i38UD3T4Lptt9Trgxn/vLL0wYlaPJefOGvWuw0xteC6kmeY+egoZe9gpRRrRxvxe3R2Pk1jNzcLYnn2dFLX8Dma/pq21wSG12bL+135lh778If/+KzumCtPu2edvSA+fbWZzTYnN4ttyUV3s5nGYWj6fLzseiCk1yMLoccIAAAAQOHRMAIAAABQeDSMAAAAABQeY4xy71jCNrsHfAyZfvOLs2xMWr19um/uQKd858lvJXVfdTnUx3V3Uudf30tc2U6kY5HG3F2pv1SfTur8B7Dk7lj90PqlyXo2HXOK79GRpO6kK293AbGpcmWy3hXzD+mU57K829FaHNPUdOOIyqPp9OWhGaehrc2n+5ibjXWzlViuV9Oc3nrVTcPt1pOk4JabI35q3PRTMBW2d8rnz5xK6kphQgCwXL2GYSRTCWcrnnbjj45k564L3Mr9zpjc7zVAr7pe1wfTY/G5HEm/onQymWs6fbRQZYzRYs2FNFj8mK589EvNTdc92ojb1ZtjyXr12fg+lBvpmLbGFjce6btcRR44/mv0UFbnL2me4Mr5cLmpLmUpHad0xI1VTi+X1PTjvPPwcvFnfphV/lzm3bTetXQ8lpr5KK+zce0MAAAAoPBoGAEAAAAovEKm0u3Wxzvla/WUvrbZpz3J8k49eKDH1It/5JuyOt/TmPcm5jMuYmHXhtg3/LDRxyR1D517WKf8VO3qa393N9N7mN/ZjO/gO/rcx77pu5LlScXla/XU7hu6gNhXS2N2T+2bnfJ9I7NJ3V0uyk7U4nSf87WjyXqNekzca1bTLupGJXb7N0aabr20P7xUir/HWD7ldyWekubdevUsVcU2xzS+ykw6PWnlIgHAsp2dphb/4qdWtmzFrS7NbluPhDx/lv/EIo7rya486cpLvT4Yc/cjaWTpSxW33MieS57ijHMrhfQFnivF5WpI+ypGXGp5I8TvyvmpNIfNtsX1Jqa2JXXhc/HWHJWRmEbWrKYpZbXT8bFHTqbfy81qfN8b293392j6XBqz8boihOzeJ9Mx/W/0mEvsrKTrBZdCn09fHnwq51w8jpB9AG3eTfmdB3SvW+90HhcAAAAACo6GEQAAAIDCo2EEAAAAoPAKOcboZt3aKe/JpmAOSVZuzKvcoR1LeqzdrryYMT9+lMq+Je7D5y//wSK2K5pDuj8ujNyR1N09d08sKx3303D5r2VXvlo7k/U+rhs75buysWo+3sztY6euSta7xcXsXT1i1lz5SqXTdR9WHC80q3R8UHB5uHUX9+VsktfSrBtHNDuX1DXcPptuH01L86Eb5WlXzqYWdWOOGvV4erJmut7opVs65UrtWHqMl2fzzQLAkmTjE8yda3sMr9nizl07lJ3j3Hn+WW7/1y3iqCZdeRDXB59yT3NvM33OR0vJICMs02gWOH7a93oWb5tdrDQrF8R9nDqcrFeaits1G+n+Zw/EWKy497Y2l17+W6V7c8D8kOR9cf+V8fR4y2Pu+/vi9Hu/WopjjMpld7zZuCpzY4dkWV3DLc+551nOPozu0iTUs7osvhdCjxEAAACAwqNhBAAAAKDwCplK1++9ok1+2sD8vtH9udGV71zEEe105Vtd+S71b2nJf8VTUewnLs+PdF3PzroNc+Rr8ikmfYpcPr1pyaWchWQq2LMfvbuF75Gex2zTHWWzlKbBlSweRy3EaTxPaSZZ74ROLliWpBn3ePM+fSSkd+kOdXd/70Y2la1L6bNGzNuojKapf/VKTJebD+l03WPb0mUA6FvoUpaSU23oViFp2qXcHcjWHHHL6+X6wGdKzZez9Ch3TiaTbvDGXGrdqbOm8o7l8Ub8Tp2z0WS90lRMrSuNp6mb9Zl4a4vxufgdPdtI381aOV77jFTGkzqFmBbnp7+uj6VTfo/OxuNqZNHSLLvptetuevg8qNwxWnYNYy7tLvgXJ7tgsnn3OjbTawfVzx3F9BgBAAAAKDwaRgAAAAAKr/CpdHZW2tPC94rO11uKvBXqk4jyHvtuSVv5PbR7tWy5J3V/jip2E1craXrYprnYNZx361a6pK2VsnfF3PLZ8ebrFp6h7sxfuu8jTeTrtt58KUbPYUsjacot+1li8niz5DmnXfYldzrxr0fecZ3M/FjK7qTuZkBqdvm7JJXKsXu8nNelhwUAA9Irzy6a96nJ2XqjyfdDXG8trw8qbsX8XNvsMYuXhe516I9/Hyayr/1pn/7eiAmP4+GiZL1qKV7DnK4fSeo2TccUuc1Nd53STKNj1qV/joY0/cxqMWWu1HDDDdKJ5xRKcb1qc2tSd7oS6+Zcal5oZrP0ubxOy2el8zPYuZS7dNo8SfVYFxpp2qEa52720GMEAAAAoPBoGAEAAAAoPBpGAAAAAAqvkGOM9rh7RX9CNyd16RiKWL52UfeljnplJJd61HXbLm/J+uzMfEJxxhj1pxJiZvbc6TQ/d1Jx6udbk4lR0/TaoJiT+3Q9OVnvLhdvNyYTtKbjkYIrPzmLt70DiNmLypfH9UrpyJ8t5scmxUE6m5VOXz6rOPV2njvfTMZI+XFKaWQGtxyyU1CSV+/ykLO0d4VmHAs2Yun03GVjUlkAA9Bjum7rsdq4O59OZN/ElS5jjNby+qBZ6j421T9CfmYtB64yBqmcvfabSvFdOxHiqz/XTEeTWWlTp1wtncjq4vs3pjjOZ766KVkv6PxOebSe3s5D9am4npvOvVpOb8VRnznQKY/U04hrjJwXj9+NTctmh1dzLj63fIyRH8fcmIsblrIx06Wam268nn0qzpof/Gz0GAEAAAAoPBpGAAAAAApvVVPpPq6PdcpP0VO7rue7cg9pf1J3u77h9vGUpC50mcRynyaz5Xh/6Cf2uY+7tTdZfq1eveB6uT2ufFNW5x8pn2bTJ2P5V+CjPR4r79R+8oJrIffITQ/ulGt6aFL31NOPjmU9ra/9LSZmu9mbRE7vmO1mfxb3/17/Yqd8uJqmn+33d7Z290FvKO1Sb7rlZjZXp3WZsrzZIzVDzXQffgrRppsytFLLpuOcPxXrzk9TB2qjWRoAgPWnR5raumHdb5ug0KUs6aSb+rieJaBtc9/2/syVJlmn8uuDJ7nyIK4PrnDp5LL0jam7KZjrzTQ9ykr5kWFZstSxiou/cjOmtdd0PFlvuh6/zxvldKrtsXp8P+cbMU1+tFxP1quFmJ5emkvf16ZLx/MztJdraSrdnEtTs+l0WEJlYntcmI/bVerpcdTn3T7OSsN3x+UuHczSe3SYewmajXT/6jH9/Bn0GAEAAAAoPBpGAAAAAAqPhhEAAACAwlvVMUaf0C2d8v5sHIbPaPRTDl+tByXr7dbuTnlPNu6nnEwXHPdxpa5K1vukPtkp3+mmQW5t56dPjvvbqR1air9a0lbSZ135X135cYvYh39mr1vicRTBnfWZTnnrzN1J3W43piaP2eBizOdv947ZNN7KLvJXOmYPl+PznA7pmJ16cNN4un00s+x2c8ujZ9XZguuVlD5Ww63XzHLzg5tG3G+3qTydruf2X9k6k9Sd3JSNRwKwPrgxE1ZKP/s+9b+UDUA696iAAfKHlY238eMrkvHI2XoXzsZLq4eMpuMfynNxKuHr3e0grl3EIforn3td+XuWuI9b3XiNZj09r58ox+UjzeySkTFGg5UNBGvOxe/D89ytMypj6Tiio24a7ntDOqbtVD0uT7nv+Uotfe92urFJpeZoehxN951dddNk19K+lcb81k45ZFN+V8qH3IqxrjJ3ZbKe+e3q+bTk7gWad7GYnUtU99cY+XzgjDECAAAAgHOiYQQAAACg8FY1la501qSTkblu6ZDek/msNc8uLbTs0+pCtp5/rLzLvpmsGY9iVTvzE//syl/vsV7eyr1iBY5lGFVLMRWrmk11WWr2ukuyv2v5wnczb/Exm8eiF7r8fUAx67rYzdLpuscVn/e0YupH1XXfS1JJPi0kn8o75mOEpC49Dn8eqGRT2ZbdHb1rLpWuNJ9OuVmZia/31mqeVrAe5/0FkHyz17LzgkuzC3lqjK3d969XcsfRLPWY4no01t03m54nR136nJ+i+87ssfwrkD97nyR9qysv9fpg1s2KPJut2bD4XKrZ9MnNajYV8hDyKeIhe7qlcvfv27Nmeu/6APH1Nktf39PN+ICjbvdjlfQ9utB9n9+fXbNcFOL1zXnu78fSr03dF+Jjj5TT57LdTeceGu71yL6/G/PbOmU7mVSpNOLqwjc75cpc+lxKbrlUz6brtpjiZy5jvlROr2fUiMdfy1IGVep1Tdde5ZxrAAAAAMCQo2EEAAAAoPBWNZWu6bok60r78cyl6JRd91wjS4sp+27NPh83nJWw5LfsnpDXbdaxtZR3XK+X49rIjtdjClg1pB+Jbe63g8ZZMRvTzEoDiNle72Uas71+z+ges/MuheNwLe2iPupmbjmt2PU+p3TGt7JbrmSpdCPu9Sm5tDrLXreKq/PrtZZdWoGL9pMh7So/fSq+3jPz6at6/PAWAViHRuNZqVRKZ2truhQgC9lsVMlscCvMP1aze9pySGb/So932p1PR/PnsvBD9Xxe/T7npV4fuGw51Rrp90vd5SxVNJbW1Vb1EnJNNF3aWjl7QWsuPipZDOTXnd0kKXf5jGkuxObqMTctNE8lq826fUxnu3io+07119178uNzs9dVS2kkXTIe1512aWrZ5LYK2tQpl2ppGn7VHhHLE/HvpePpTsYa8bwwN5PGm4252fLm3GexlD7p+RNxn9VT2QtSTc87C6HHCAAAAEDh0TACAAAAUHg0jAAAAAAU3qomiH5an+6Ub00mmczvdB0P6+m6LllvnyY75Vt0U1LnRzIEN63ktfreZL17ta9TvjE7DnNtxeAydp+gp2itVLuUpWz606xufUxwuv6NuAT26caxpO5uN47mU/pUUuenqvTx9vTsHuZ3u3uM36qbk7qG285c3D8xu4f5PdrTKd+cHYeSmI2fgjxmm5UYLSOWTUte8znFMZe5qRPpPpLlNM951D32uDumSjbF66hGXV2aQzzjIvo+F8GzemCyXm0u3u997p40H/rI8WzqTgDrgh930Wymn9tkmEDIpulNvt1W75ut1yzhwY0jKpXSb99KNdaN19Lfn2tu3IivSZ9xMrzkrNEqvm4Q1wdT07F8Xz39bqi7R5vOHiHUhv+3dXNTVzdqacxW/XdvFrPJwK1e+3flkUa6zdbS5k759Gj87m3OpY9VLY13yudV0zE7lyruY64c9/+ILCIOb4rXH/eNjCZ1h9xYsgeOxBgYG92WrFc7dEmnXGnel9SNlo53ylvOi9ud3L45WS8cj9cfc3en1xilE/GY590Y76DtyXrze+M1QPnBWTMne24LGf6oBgAAAIBzoGEEAAAAoPBWNZXuU7pl0dt8egnb5P5lAPv4pD657H0slb8j9k1ZXa+kgut61CGacNNYN8N0UvcIXdMpX6enJ3XbFOecnHXd0geUpnI9Wld1ys/W9yV1DTdd9Zx7N+/OUhaudI/1TD0mqSu7+1nfr6lOebO2JuvdVzvSKR913dqtx47pEvPumBrZdNrBTdddzqbyTieL9ftLo/SEW54567eZmFhwxJXnXAqtJJ3nXtNHNjcldZfOpOl/ANYHq8Q0LWtm6XKNml9IN2y6hKOwijepyL9g3UObxcunkD2XZojPZX5zeo47ejwul9z3Rn6V0msqb5+sfY8r35yt1yuZa5ffh5tGfb6WvvbjimlPFyqtq5/KEwCHj5Xie9kspc+35OpCM4vLZn/9Dj6j9GQzvSRvWLwOmJiP5fFy+j5snnC3C7kwrStPx+/DHffGY3pgdvlfORHr9pTS7/3DzXhd5J/VJqXXSxU9pFOuHb0rqTv90Q93yv6VOqlUVY+L+8gmoK8n1yb+U5F955e3x/VG0+eiY+eOWXqMAAAAABQeDSMAAAAAhUfDCAAAAEDhreoYIyzN25e4XZ5vjIUdHouZ2NOb0ukhTV/ulLconX5ypBLHthyrxPE7hyrpPhr1OOZoazal50iI04GXXf798ZF0fFClEfNpxxppDvH25gWd8hGXfz+/Oc15rjbjNJ7lWpp3OzcT6+Zm4zHVauk4onrw26XTuvqMdp/928gmm51z5ekse96vO+3K9Wwy22M62in/+9Y0l7lWHv68d2AjCpU4TmI+G5PRLMVlK6V1Zqs4rsg/bq9KNw7KmulYi5H5ODZippZOD3zajf28WPG8++DsfNp058Yj2Wihg+58eJH7+xXZGKAJ99v3Pdm59rCbH70+Ecsjpez2B7U4CmTS0jrbPq5h1xyN7182/Eq1shsvVlpaP0OzEb8RZ2bS764TbuxMY7Mrb0m/l09fHI/x2N6kSicOxtg85N6uE+nQXE3HSxgdOC+tO+VaChe4D8W2bPjOll/4Zqdc/tXXJHWV2+OtOUZveWWnvP0fv5Qex32f7ZRLU1ckdeVGfN6NWReL49lE9RfF9Wbc2OrWuufrXOgxAgAAAFB4NIwAAAAAFJ75uzcDAAAAQBHRYwQAAACg8GgYAQAAACg8GkYAAAAACo+GEQAAAIDCo2EEAAAAoPBoGAEAAAAoPBpGAAAAAAqPhhFQYGa228x+bq2PA+gXMYuNiLjFRlPUmC1sw8jMJs3sqWt9HEAv7TidN7MLs7/fZmbBzHau4rG8xMw+b2ZzZvaO1XpcbCzrJWbNbNTM3mZme83spJl9ycyesRqPjY1nvcRt+zHfY2b3mtkJM/tGES9OcW7rKWbdYz/EzGbN7D2r/diDUtiGEbCB7JH042cWzOyRksbX4DgOSHqtpL9cg8fGxrIeYrYiab+k6yRtk/Q7kt6/FhcL2DDWQ9xK0h9K2hlC2CrpByS91sy+ew2OA+vfeonZM94k6XNr+PjLVviGkZm9yMw+bWZ/amZTZnaXmT2+/ff9Zna/mf2UW/9Z7V8eT7TrX5Xt74XtXyiPmNnv+J4pMyuZ2SvM7M52/fvN7PxVfsrYeN4t6YVu+ackvcuv0CsuzWys/QvkkXaMf87MLskfxMwuM7PbzezXFjqIEMLfhxA+KOnIIJ4Uhtqax2wI4XQI4VUhhMkQQjOE8I9qXURwgYlu1jxuJSmE8NUQwtyZxfa/By/zuWE4rYuYba/zPElTkm5c5nNaU4VvGLU9TtLtki6Q9F5Jfy3pMZKulvQCSW80s83tdU+rFYTbJT1L0i+a2XMlycweLunNkp4v6TK1fqV8gHucX5b0XLV+wbxc0jG1WtdAL5+RtNXMvt3MypJ+TFLeTd01LtU6UW6TdIVaMf4Lkmb8xu1f0W+R9MYQwh+vyLNAkay7mG1/2T9U0leX9IxQBOsmbs3szWY2LekOSfdK+uhynhiG1rqIWTPbKuk1kn512c9ojdEwatkTQnh7CKEh6W/UCpDXhBDmQggfkzSvViNJIYTdIYSvtH+BvF3S+9Rq6EjSD0v6hxDCp0II85J+V61fes54saRXhhDubv8a9CpJP2xmldV4ktjQzvwq9DS1vijv8ZXniMuaWie8q0MIjRDCF0IIJ9zmD5e0W9LvhRDeusLPA8WxbmLWzKqS/krSO0MIdyzzeWG4rYu4DSH8kqQtkp4o6e8lzfVaH4W2HmL2v0l6Wwhh/0Ce0RqiYdRynyvPSFIIIf/bZkkys8eZ2c1mdsjMjqvVuj4z8O1ytXLa1d7HtNK0ox2SPtDurpyS9DVJDUlndVsCmXdL+glJL1LWTS6dMy7fLekGSX9tZgfM7I/aF4pnPF+tE+nfruQTQOGsi5g1s1J7f/OSXrKM54NiWBdxK0ntC9VPSXqgpF9c6hPC0FvTmDWzayQ9VdKfDuC5rDkaRov3XkkflnRFCGGbpLdIsnbdvWqdwCRJZjauVkv8jP2SnhFC2O7+jYUQktY9kAsh7FVrfMQz1fr1MNc1LkMItRDCq0MID5f0eEnfrzQn+VWSDkt6b7srHli29RCzZmaS3qbWj08/FEKoLfd5Ybith7hdQEWMMUIX6yBmd0naKWmfmR2U9GuSfsjMvri8Z7Y2aBgt3hZJR0MIs2b2WLVa6Wf8raRnW2vyhhFJr1ZsNEmtYPx9M9shSWZ2kZk9Z7UOHBvez0p6cgjh9AJ1XePSzJ5kZo9sn9ROqNV13nDb1iT9iKRNkt7d/oX9LGZWMbMxSWVJ5fagTdJA0cuaxqyk/y3p2yU9O4Qw02UdILdmcWtmF5vZ88xss5mVzex6tWYdu2lgzw7DaC3PtW9Vq+F+TfvfWyR9RNL1y31Sa4GG0eL9kqTXmNlJtcYQvf9MRQjhq5JeqtbkDfdKOinpfsXc4Deo1Wr/WHv7z6g18QNwTiGEO0MIn+9S3TUuJV2qVqP9hFrpm7coG5zZHhP3/0i6WNJfdjn5/bZaaaWvUGtSkpn234AFrWXMtn+AerFaX9QHzexU+9/zl/3EMNTW+Fwb1Eqbu1utCZr+WNLLQggfWtaTwlBby5gNIUyHEA6e+SfplKTZEMKhATy1VWchhHOvhSVpz2Q3JekhIYQ9a3w4AAAAALqgx2jAzOzZZjZhZpvU+qXnK5Im1/aoAAAAAPRCw2jwniPpQPvfQyQ9L9AtBwAAAKxrpNIBAAAAKDx6jAAAAAAUHg0jAAAAAIU38HuQmBm5eUMqhGDnXmvjeVrpR4jZIfXx5v8dypiVpObBhxC3Q6p06TeHMm451w6vYT3XErPDq1vM0mMEAAAAoPBoGAEAAAAoPBpGAAAAAAqPhhEAAACAwqNhBAAAAKDwaBgBAAAAKDwaRgAAAAAKj4YRAAAAgMKjYQQAAACg8GgYAQAAACg8GkYAAAAACo+GEQAAAIDCo2EEAAAAoPBoGAEAAAAoPBpGAAAAAAqPhhEAAACAwqNhBAAAAKDwaBgBAAAAKDwaRgAAAAAKj4YRAAAAgMKjYQQAAACg8GgYAQAAACi8ylofwLpzsyvv6nObyR7LS93HVX1uBwAb0PWXX7PWhyBJuuHAbWt9CACwItbL+W29nO/7QY8RAAAAgMKjYQQAAACg8Eily+125ck+t9m5AvsAAAAAsGroMQIAAABQeDSMAAAAABQeDSMAAAAAhUfDCAAAAEDh0TACAAAAUHg0jAAAAAAUHg0jAAAAAIVHwwgAAABA4dEwAgAAAFB4lbU+gHVn0pV397nNrhXYBwAAAIBVQ48RAAAAgMKjYQQAAACg8GgYAQAAACg8xhjlXuTKu/rcZjJbvmoJj3vLErYBgAK74cBta30IADDUrr/8mrU+hFVFjxEAAACAwqNhBAAAAKDwSKXL7XblyT632TnwowAAAACwiugxAgAAAFB4NIwAAAAAFB4NIwAAAACFR8MIAAAAQOHRMAIAAABQeDSMAAAAABQeDSMAAAAAhUfDCAAAAEDh0TACAAAAUHiVtT6AdWfSlXf3uc2ugR8FAAAAgFVEjxEAAACAwqNhBAAAAKDwaBgBAAAAKDzGGOVe5Mq7+txmMlt+9SAOBAAAAMBqoccIAAAAQOHRMAIAAABQeKTS5Xa78mSf2+wc+FEAwIbXCM1OuWz8DgcAA2cWyyGs3XEMCb6pAAAAABQeDSMAAAAAhUfDCAAAAEDhMcZoA7jZlXctYrtJV75qIEcCAMPr+suvWfY+bjhw27L3AfRrEPE2iLgHhgU9RgAAAAAKj4YRAAAAgMIjlW4D2O3Kk4vYbudAjwIAFocpugFghTFF90DxrQUAAACg8GgYAQAAACg8Uulyk668u89tdg38KABguJXKsdxsdIpWHUlWC7X51ToiABgqtRDPrVWL59zpZnpenSil590io8cIAAAAQOHRMAIAAABQeDSMAAAAABQeY4xy71zCNrcM/CgAYKjdcPcX1voQAGCoff8DvnutD2HDoccIAAAAQOHRMAIAAABQeIVMpbtZN3fKu/qca3symcdbukpXDfCIzvXY0e5FbLdroEeBtXTDgduWvY/rL79m2fsAFmMQMTeI2AcAoB/0GAEAAAAoPBpGAAAAAAqPhhEAAACAwivkGKPdbqROPnaom53auaTHutmVdy1iu8ku5aXu49WL2A4bVyM0O+Wy8bsHNgizWA5hSbtgDB02GsbPAesPV04AAAAACo+GEQAAAIDCK2Qq3Wra7cqTi9hu54D3gWIgfQ4b0hLT5wAAGCSuogAAAAAUHg0jAAAAAIVHwwgAAABA4dEwAgAAAFB4NIwAAAAAFB4NIwAAAACFV8jpuifdpNe7k8mwu9ulXStyLAAAAADWHj1GAAAAAAqPhhEAAACAwlvVVLqbdXOn3G9qmk97y5fXch+v1qv73C7a3dcWLbsGvA8szQ0HblvrQwAW7frLr1nrQ1i/SuV0udlYm+MA+lQLaYxWrdxlTQDLRY8RAAAAgMKjYQQAAACg8GgYAQAAACi8VR1j5KfGzsfsdLNTO9flPvr1ziVtJd3iyje78q5F7GPSlfsbEYXc4cbpTvnC8qY1PBKgf6VNMVabp0/3WHN4DGI8IGOzsJoGEW+MgwUGix4jAAAAAIVHwwgAAABA4a1qKh2WZrcrTy5iu50DPYpiKsvW+hCAxQthrY8AAIANhx4jAAAAAIVHwwgAAABA4ZFKB/SpEZpd68rGbwxYp6xHOigpdwAAdHA1BwAAAKDwaBgBAAAAKDwaRgAAAAAKb1XHGE26yaZ3J5NQd7dLu9blPlAM133hZzvltz/qnUnd9tJ8pzzhhnFsK40k65XdGI9GNqbjvkbcx/FmtVPeUWkk602UYt3J5nxSV3Xjm5pu/1tLY0Ix3fuz13TKD/jA3qQuzM7Ghflap9icmVW6ohtTl42hK23dHKvGYpw1j02lu5iPsWqjo+n+Gy7GSy6GZ2YEAMBaoMcIAAAAQOHRMAIAAABQeKuaSvdOvfPcK2Vu0S3Lftz1so+lmnTl3YvYbtdAj6KYLnvu1zrl39Jj1/BIgP7d9ptvjgu/uXbHsRH0moYfWJd6TcEPYFnoMQIAAABQeDSMAAAAABQeDSMAAAAAhbeqY4ywNIsfmdWydqOiAGBjKLupyG84cNvaHQjQL3dbhusvv2btjgMYQvQYAQAAACg8GkYAAAAACs+C65IFAAAAgCKixwgAAABA4dEwAgAAAFB4NIwAAAAAFB4NIwAAAACFR8MIAAAAQOHRMAIAAABQeDSMAAAAABQeDSMAA2Fm7zCz1y5i/Reb2ev7XPeXzex1Sz44YAHELDYaYhYbzUaLWRpGwCKZ2aSZzZjZKTM72P7Qb17r4/LMbKeZvSr7224zC2b2qOzvH2z/fdcqHt+IpN+W9D/c395qZl83s6aZvSjb5K2SXmBmF6/WMQ4TYnYgx5fErJk91Mw+ZGaHzOyomd1gZg9zmxCzy0DMDuT48pi90Mw+bWZHzGzKzP7VzL7PbULMLgMxO5DjO+vawNX9VPt4fs79eeAxS8MIWJpnhxA2S7pG0ndK+s21PZwWM/seM3ulpEp7+Voz+y23yjckvdCtf4Gk75F0aFUPVHqOpDtCCPe4v31Z0i9J+mK+cghhVtI/yR07Fo2YXZ48ZrdL+rCkh0m6RNK/SfrQmZWJ2YEgZpcnj9lTkn5G0kWSzpP03yX9g5lVJGJ2QIjZ5Vno2kBmdp5ar+VX/d9XImZpGAHLEEI4KOkGtU6CkjonoH9p/yL3Zf9ri5mdb2ZvN7MDZnbMzD7o6v6LmX2r/evzh83sclcXzOwXzOyb7e3eZGa2wPF8RtK/S/rfkp4n6RmS/syt8leSfszMyu3lH5f0AUnz7rEe2/4lccrM7jWzN7Z/xZG1/KmZ3W9mx83sdjP7jvw4zGyLmd1sZn+20HG2j+uW7NjfFEK4UdLsAutL0m5Jz+pShz4Rs4OJ2RDCv4UQ3hZCOBpCqEn6U0kPa19QnLFbxOyyEbMDi9nZEMLXQwhNSSapoVYD6Xy3zW4Rs8tGzA7u2qDtD9vHe3iBut0aYMzSMAKWwcweqNYH+Vvt5QdI+oik16r1ZfNrkv7OzC5qb/JuSROSHiHpYrUupmRmT1brg/+jki6TtFfSX2cP9/2SHiPpUe31ru9yWMGVG9nyAUn/Ienp7eUXSnpXtn1D0q9IulDS90p6ilo9OWpvd62kh6r1i/mPSTriN25fGN4o6dMhhF8OIfjHP+ORkr7e5fi7+Zpazx3LQMyuWMxeK+lgCMHvm5gdAGJ2sDFrZrer9QPUhyX9nxDC/a6amB0AYnZwMWtmj5X0aElv6fK8BhqzNIyApfmgmZ2UtF/S/ZJ+r/33F0j6aAjhoyGEZgjh45I+L+mZZnaZWifKXwghHAsh1EIIZ34Zeb6kvwwhfDGEMKdWl/H3mtlO95ivCyFMhRD2SbpZ7peoM8zseyT9J0m/qNbJ82OS/mu22rskvdBa4yG2hxD+1VeGEL4QQvhMCKEeQpiU9OeSrmtX1yRtkfRtkiyE8LUQwr1u88vV+rXn/4YQfrvH67dd0ske9Qs5KWnbIrdBRMyuUMy2L4LeJOn/y6qI2eUhZlcgZkMI/0nSVkk/IelTWTUxuzzE7ABjtt2D9WZJL233dC5koDFLwwhYmueGELZI2qXWieDC9t93SPqRdlfzlJlNSXqCWr/0XCHpaAjh2AL7u1ytX4IkSSGEU2r92vIAt85BV56WdNagzvZJ67WS6u3lW0MIf5Ct9veSnizppWr9SpWw1qDyf7TW4NETkv7gzPMLIdwk6Y1qXQTeZ60JE7a6zZ8laVzdf9k545haJ9HF2CLp+CK3QUTMrkDMtn/x/ZikN4cQ3pdVE7PLQ8yu0Hm2nVb3PkmvsHTQPTG7PMTsYGP2lyTdnjfSMgONWRpGwDK0f9V5h6Q/bv9pv6R3hxC2u3+bQgiva9edb2bbF9jVAbVOnJIkM9sk6QJJ9yywbj/HNRlCeFWXumm1Biv+ohY4+amVg3yHpIeEELZK+i218tHPbP9nIYTvVqvL/6GSXu62/QtJ/yzpo+3n0M3t7W0X49vVmqABy0DMDi5mrTUg+GOSPhxC+P0FtiFmB4CYXdHzbFXSg9wyMTsAxOzAYvYpkn6w3Rg7KOnxkv7EzN7o1hlozNIwApbv9ZKeZmbXSHqPpGeb2fVmVjazMTPbZWYPbHcr/5OkN5vZeWZWNbNr2/t4r6SfNrNrzGxUrV9iPtvurl4JvyXpui773yLphKRTZvZtap0kJUlm9hgze5yZVSWdVitPvZFt/xK1coT/0czGuzz+RxW74M/se8TMxtQ60Vbbr50/R12n1uuH5Xu9iFlv0THb/jX0BrXy5V/RZRtidnBeL2LWW0rMfo+ZPaF9rh03s99Qa0bFz7ptiNnBeb2IWW8p1wYvUqvhc0373+clvVrSK906A41ZGkbAMoUQDqmVm/s7IYT9ak03+VtqTXO5X61fTc581n5SrVzcO9TKP35Zex83SvodSX8n6V5JD1Zr5phzMrMnmtmpRR7zgRBCnlt+xq+plXt+Uq1fef7G1W1t/+2YWt37RxR/ETuz7yDp59V67h9qN3Zy/yDp28zNrqPWL+8zav0i9NZ2+dr2cxyT9ExJ7+z/WaIbYnYgMfuDag14/mlr3bfkzL8r28+RmB0gYnYgMTuqVqrTEbV6HJ4p6VkhhAPt50jMDhAxu/yYDa2xUwfP/FNrlrwTIYTj7ec48Ji1sOCkEACwsszs5yU9PITwsj7WfamkK0IIv77iBwZ0QcxioyFmsdGsdczSMAIAAABQeKTSAQAAACg8GkYAAAAACo+GEQAAAIDCo2EEAAAAoPAqa30AwFozM2YgGVIhBDv3WhsTcTu8hjVuidnhRcxio+kWs/QYAQAAACg8GkYAAAAACo+GEQAAAIDCo2EEAAAAoPBoGAEAAAAoPBpGAAAAAAqPhhEAAACAwqNhBAAAAKDwaBgBAAAAKDwaRgAAAAAKj4YRAAAAgMKjYQQAAACg8GgYAQAAACg8GkYAAAAACo+GEQAAAIDCo2EEAAAAoPBoGAEAAAAoPBpGAAAAAAqPhhEAAACAwqNhBAAAAKDwaBgBAAAAKDwaRgAAAAAKr7LWBwAAAABgwD7hyk/uc5s7s+W9S9jHvmx5Z5/brQP0GAEAAAAoPBpGAAAAAAqPVDoAAABg2Nzqyj4lzrL1Gq68s899hGw9v8+r+jm49YkeIwAAAACFR8MIAAAAQOHRMAIAAABQeIwxAgAAAIZNPpaoG98ayLtMQpdyr303+3zcdYgeIwAAAACFR8MIAAAAQOGRSgcAAAAMm367P3qlyJW71PWarnsDo8cIAAAAQOHRMAIAAABQeKTSAQAAAMPmW678CVfu1S3yxGx5vyvf3GMfPrXu+859aOsVPUYAAAAACo+GEQAAAIDCo2EEAAAAoPAshHy+PaBYzIwPwZAKIQzJBKJnI26H17DGLTE7vIjZdcqPCbrOlfN3yz/LyazuQYM8oPWjW8zSYwQAAACg8GgYAQAAACg8pusGAAAAhs0trnyXK/fqFrlihY5lg6DHCAAAAEDh0TACAAAAUHg0jAAAAAAUHmOMAAAAgGETupR7GcqJ1/tHjxEAAACAwqNhBAAAAKDwSKUDAAAAho1Pnyu7Mt0iXfHSAAAAACg8GkYAAAAACo9UOgAAAGDY7HXl3a6cz1Dnu0muW7Gj2RDoMQIAAABQeDSMAAAAABQeDSMAAAAAhWch9HsrXGA4mRkfgiEVQhjae3gTt8NrWOOWmB1exOw6dbMr9zt2aF+2vHMwh7LedItZeowAAAAAFB4NIwAAAACFx3TdAAAAwLDZ7cp3unKeROaXd6zY0WwI9BgBAAAAKDwaRgAAAAAKj4YRAAAAgMJjjBEAAMAGdKP7fftJava93V5XvmqAx4N1zHeF9JqEfCgnXu8fPUYAAAAACo+GEQAAAIDCI5UOAABgA7rZpc9NZnU+sS7Pjtq5MoeD9ca6lHt1i5BKBwAAAADFRsMIAAAAQOGRSgcAALABdcuUknpPQlbwbKnimHTlT7pyo8c2163MoWwU9BgBAAAAKDwaRgAAAAAKj4YRAAAAgMJjjBEAAMAG1GuskK8rZ3X8Kl4Q71jCNrsHfAwbDJ8NAAAAAIVHwwgAAABA4ZFKB2wAu/XxTvlaPaWvbfZpT7K8Uw8e6DEBANaWP8vflNX5KbrzlLtrV+ZwgA2PHiMAAAAAhUfDCAAAAEDh0TACAAAAUHiMMQI2gJt1a6e8R3cndSHJJG92Sju0Y6UPCwCwRLtdeTFjfva68r4l7uMuV/6DRWwHDDt6jAAAAAAUHg0jAAAAAIVHKh2wIfR3f3Nz9zcvnXWvcwDAenGjK9/ZY7387L/TlW915bvUPxKtgYXRYwQAAACg8GgYAQAAACg8UumADcGny1lWs/D9zfP1AADrU/4rddOVQ1aXLy+0zUL79Ph2ABZGjxEAAACAwqNhBAAAAKDwaBgBAAAAKDzGGAEbwB53f/NP6Oakzo8lKrnytbpu5Q8MALAkoUtZSn+17jXGyJfzX7r9OKL85g2MMQIWRo8RAAAAgMKjYQQAAACg8CyEbhM/AsVgZl0/BB/Xxzrlp+ip3ffhEhMOaX9Sd7u+4fbxlKQudJl4dZ8ms+V4T/Mn9rmPu7U3Wb5SVy243jALIQxtxkivuMXGNqxxS8ymnu/Ke7I6/0Ll03A/2ZX9t83XezxWHlB+H6/rsV2/iFlsNN1ilh4jAAAAAIVHwwgAAABA4dEwAgAAAFB4TNcN9PAJ3dIp78/GDvm8bz9N9tV6ULLebu3ulPdk437KLpPcj1PKxwN9Up/slO90U3e3tou/b/jxRju1QwCA9emvlrjdZ135X135cYvYh/8WGcQYI2BY0GMEAAAAoPBoGAEAAAAoPFLpgB5KZ02UGplLWwvJZKj5NrZAaaFln1YXsvX8Y6V1IXm8uMdml2m8AQDD4Z9dudd03fmv4FeswLEAw4AeIwAAAACFR8MIAAAAQOGRSgf00HSpaXXVkjpTtVMuq9EpN7IEubJb7je5LZyVZOe37J6QF3qk7QEAhlc9W+Y7AFg8eowAAAAAFB4NIwAAAACFR8MIAAAAQOExxgjo4dP6dKd8q25N6krJuJ/4UXq6rkvW26fJTvkW3ZTU+VFLQeVO+Vp9b7Leve4+5Tdmx2Hu943gssyfoKcIADC8ql3KknreRIKbOQALo8cIAAAAQOHRMAIAAABQeBYCHaooNjPjQzCkQghDO2MtcTu8hjVuidnB+2lX3pvV9XqxfcL3qwZwHMQsNppuMUuPEQAAAIDCo2EEAAAAoPBoGAEAAAAoPMYYofDIIR5ew5r3LhG3w2xY45aYHV7ELDYaxhgBAAAAQBc0jAAAAAAUHql0AAAAAAqPHiMAAAAAhUfDCAAAAEDh0TACAAAAUHg0jAAAAAAUHg0jAAAAAIVHwwgAAABA4dEwAgAAAFB4NIwALIuZ7TSzYGaVPtcfNbP/MLNL+1z3DjO7ePlHCrQQs9hoiFlsNBs1ZmkYAUtgZpNm9tS1Po6FmNk7suVd7ZPT32d/f1T777tX8/gk/bykW0MIB9vH8SQzu9nMjpvZpF8xhDAn6S8l/cYqH+PQIWaXJY/Zl5vZv5vZSTPbY2YvP7MiMTs4xOyy5DH7MjO7y8xOmNkBM/vTMxesxOzgELPLksSsO56RdiPo7jN/W8mYpWEEDAkz+x9m9sh2eaL9xXdlu/qQpMeb2QVuk5+S9I3VPk5JL5b0brd8Wq0T3MsXXl3vlfRTZja60geG1bWBY9YkvVDSeZL+s6SXmNnzXD0xO6Q2cMz+g6TvCiFslfQdkh4l6ZddPTE7pDZwzJ7xckn3L/D3FYlZGkbAMpnZi8zs0+2TzVT7V7nHt/++38zuN7Ofcus/y8y+1P7lbr+ZvSrb3wvNbK+ZHTGz3/G/QJlZycxeYWZ3tuvfb2bntzf972qdWJ4k6T2SPhBC2Neum5f0QUnPa++nLOlHJf1V9thvaB/TCTP7gpk90dU91sw+3667z8z+Z5fX44fax/wdC9RdKenBkj575m8hhH8LIbxb0l0L7S+EcLekY5K+Z6F6LB4xe9brsdiY/aMQwhdDCPUQwtclfUjS97l6YnbAiNmzXo/FxuydIYSpM6tIakq62tUTswNGzJ71eiwqZtt/v0rSCyT9Yb7NSsUsDSNgMB4n6XZJF6j1K8ZfS3qMWl88L5D0RjPb3F73tFq/Nm+X9CxJv2hmz5UkM3u4pDdLer6kyyRtk/QA9zi/LOm5kq6TdLlaJ4U3ufrg/m9kx/iu9uNK0vWSvirpQLbO5yRdI+n89vP4v2Y21q57g6Q3tH9xfLCk9+cvgpn9tFon4aeGEP49r5f0SEl3hRDqC9T18jW1fuHE4BCzWn7MmplJemL72DxidvCIWS09Zs3sJ8zshKTDasXmn2fbEbODR8xqWefZ/yXptyTNLLCNtAIxS8MIGIw9IYS3hxAakv5G0hWSXhNCmAshfEytX2WulqQQwu4QwldCCM0Qwu2S3qfWyUySfljSP4QQPhVCmJf0u4onNKn1q88rQwh3t3NsXyXph62VK/4bkt4q6Wa1TnI/bLG7XCGEf5F0vpk9rF3/rvxJhBDeE0I40v4l/E8kjUp6WLu6JulqM7swhHAqhPCZbPOXqdXlvSuE8K0ur9N2SSe7vordnWxvi8EhZgcTs69S67v07dnfidnBI2aXEbMhhPe2L14fKuktku7LViFmB4+YXWLMmtkPSqqEED7QZRtpBWKWhhEwGP4LZkaSQgj53zZLkpk9zlqTDRwys+OSfkHShe31Lpe0/8xGIYRpSUfcfnZI+kC7W35KrV9LGpIuCSG8PITwlfZ2p0MIv+K6y894t6SXqNWlftbJxsx+1cy+Zq2JEKbU+lXqzLH9rFpfqHeY2efM7PuzzV8u6U3t7u1ujkna0qO+my2SppawHbojZpcZs2b2ErUuJJ7VvhjxiNnBI2YHcJ4NIXxTrV6BN2dVxOzgEbNLiFkz2yTpjyS9tMc20grELA0jYPW9V9KHJV0RQtim1i931q67V9IDz6xoZuNqdcGfsV/SM0II292/sRDCPWdWCCG8qMdjv1vSL0n6aPvE2tHOGf4NtfKLzwshbJd0/MyxhRC+GUL4cUkXq9Ul/rftk9cZT5f022b2Qz0e/3ZJD7I+p+90vl3Slxe5DQaHmM1i1sx+RtIrJD2lyxc+Mbu2iNne59mKWmlPHjG7tojZGLMPkbRT0ifN7KCkv5d0mZkdNLOdbruBxywNI2D1bZF0NIQwa2aPlfQTru5vJT3bWgM0RyS9WvHEKLVOlL9vZjskycwuMrPn9PvAIYQ9anXNv7LLcdXVmqWmYma/K2nrmUoze4GZXRRCaCr+QuNzlb+q1gxdbzKzH+jy+HdL+qakx7r9ltq5ytXWoo21n/uZ+geoldecd89j9RCzacw+X9IfSHpaCOGsSUOI2XWBmE1j9uesfc8Xa41X+U1JN7p6YnbtEbMxZv9drbTDa9r/fk6t3rdr1O45W6mYpWEErL5fkvQaMzupVp5wZ6BiCOGranUd/7VavxCdVGuayjNpOm9Q6xelj7W3/4xagzv71s5RzgdWStINkv5JrWk690qaleu6V+vE9lUzO9U+jueFEGazfX9Z0vdL+gsze0aXQ/hzST/plq9VK53go5KubJc/5up/QtI7F0hVwuohZtOYfa1av9Z+zsxOtf+9xdUTs2uPmE1j9vskfcXMTqt1rv2oWoPazyBm1x4x247Z9limg2f+SToqqdlePtPoWpGYtRDCudcCsCasNVvNlKSHtH/R2fCsdc+BL6mVgnRvH+t+WdK1IYSF7mOAdYaYJWY3GmKWmN1oiNmVi1kaRsA6Y2bPVivFwST9iVq/+nxX4MOKdYqYxUZDzGKjIWZXB6l0wPrzHLXuIXBArQGIz+PEh3WOmMVGQ8xioyFmVwE9RgAAAAAKjx4jAAAAAIVHwwgAAABA4S32JovA0DEz8kmHVAjBzr3WxkTcDq9hjVtidngRs9housUsPUYAAAAACo+GEQAAAIDCo2EEAAAAoPBoGAEAAAAoPBpGAAAAAAqPhhEAAACAwqNhBAAAAKDwaBgBAAAAKDwaRgAAAAAKj4YRAAAAgMKjYQQAAACg8GgYAQAAACg8GkYAAAAACo+GEQAAAIDCo2EEAAAAoPBoGAEAAAAoPBpGAAAAAAqPhhEAAACAwqNhBAAAAKDwaBgBAAAAKDwaRgAAAAAKj4YRAAAAgMKrrPUBAAAAABiwm1x5V5/bTPZYXuo+HtTndusAPUYAAAAACo+GEQAAAIDCI5UOAAAAGDa7XXlvn9vs7LGPySXuYwOhxwgAAABA4dEwAgAAAFB4NIwAAAAAFB5jjAAAAACczbqUc2GlD2R10GMEAAAAoPBoGAEAAAAoPFLpAAAAgGHmU916pcT12m5I0uV6occIAAAAQOHRMAIAAABQeKTSAQAAAMNm0pV391jPp9btWuI+vHwfGwg9RgAAAAAKj4YRAAAAgMKjYQQAAACg8CyEAsy9B/RgZnwIhlQIYTGTkm4oxO3wGta4JWaHFzG7Tt3kyrtcude7NZktXzWog1lfusUsPUYAAAAACo+GEQAAAIDCY7puAAAAYNjsduVJV+6VSrdzJQ5k46DHCAAAAEDh0TACAAAAUHg0jAAAAAAUHg0jAAAAAIVHwwgAAABA4dEwAgAAAFB4NIwAAAAAFB4NIwAAAACFR8MIAAAAQOFV1voAAAAAAAzYpCvvdmXrsc2ulTiQjYMeIwAAAACFR8MIAAAAQOHRMAIAAABQeBZCWOtjANaUmfEhGFIhhF6Z1BsacTu8hjVuidnhRcyuUze58q4e6/l3bzKru2pQB7O+dItZeowAAAAAFB4NIwAAAACFx3TdAAAAwLDZ7cqTPdbzSWU7V+JANg56jAAAAAAUHg0jAAAAAIVHwwgAAABA4THGCAAAYAPqdzbm3KQrP2ggR4INbWNPSj5Q9BgBAAAAKDwaRgAAAAAKj1Q6AACADWi3K08uYrudAz0KDJWCp9XRYwQAAACg8GgYAQAAACg8UukAAOuHZcsFT+sAgCWbdOXdrpyfV/15d9cKHcsGQY8RAAAAgMKjYQQAAACg8GgYAQAAACg8xhgBANYPxhQBwGC8awnb3Drwo9hQ6DECAAAAUHg0jAAAAAAUHql0wAZwk27qlHf1OZfmZHYf9AfpQQM8IgDAWtvryrsXsd2uwR4GMDToMQIAAABQeDSMAAAAABQeDSMAAAAAhccYI2AD2O2yx/OxQ93s1M4VORYAwPLd5Mq7FrHdpCv7MUaL2cceV37NIrYDhh09RgAAAAAKj4YRAAAAgMIjlQ4AAGCV7XblyUVst9OVb+lzH6HHPgBE9BgBAAAAKDwaRgAAAAAKj1Q6AACADS5PlwOwePQYAQAAACg8GkYAAAAACo+GEQAAAIDCY4wRsAFMuolYdyeTvHa3a1H3QQcAACg2eowAAAAAFB4NIwAAAACFZyEwwSOKzcy6fghu0k2dcq/UNJN1ypPZ/cf9cr6P0GWC1Xwfe7V32ft4kB604HrDLIRg515rY+oVt9jYhjVuidnUC115chHb7XLlva68J1vPB1H+wvt9vGYRj90NMYuNplvM0mMEAAAAoPBoGAEAAAAoPBpGAAAAAAqP6bqBHj6hT3TK+ZidbnZqZ7Lsp9fec1YWeH/7uEW3LHsfKI6KO7XXVV/DIwHQzbuWuN2trnyTK+9axD4mXXkQY4yAYUGPEQAAAIDCo2EEAAAAoPBIpQN6MK2PGUi7TckNLIR4AYphtytPZnW9puveuQLHAgwDeowAAAAAFB4NIwAAAACFRyodAADAkCGhFlg8eowAAAAAFB4NIwAAAACFR8MIAAAAQOExxgjo4Rbd0inflNxjXCq53xX8tN67svuP79XeTvlm3ZzUNdXslP0Uy0/Sk5L19mlf1+PoNqV4vg8Uxw7t6JT3a39S1y3mcr3qusV+vo1/rDxOu+2fqcYBAGuFHiMAAAAAhUfDCAAAAEDhWQikLaDYzIwPwZAKISycZzgEiNvhNaxxS8wO3gtdeXIR2+1y5dcM4DiIWWw03WKWHiMAAAAAhUfDCAAAAEDh0TACAAAAUHiMMULhkUM8vIY1710ibofZsMYtMTu8iFlsNIwxAgAAAIAuaBgBAAAAKDxS6QAAAAAUHj1GAAAAAAqPhhEAAACAwqNhBAAAAKDwaBgBAAAAKDwaRgAAAAAKj4YRAAAAgML7/wHNXZIOS+qDgwAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 1080x720 with 20 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def visualize(image, reconstruction, textures, masks):\n",
    "    fig, ax = plt.subplots(4, num_slots + 1, figsize=(15, 10))\n",
    "    ax[0][0].imshow(reconstruction.squeeze(0).permute((1,2,0)).detach().cpu().clip(0.,1.))\n",
    "    ax[0][0].set_title('Final Recon.')\n",
    "    ax[1][0].imshow(image.squeeze(0).cpu().permute((1,2,0)))\n",
    "    ax[1][0].set_title('Image')\n",
    "    for i in range(num_slots):\n",
    "        texture = textures.squeeze(0)[i].detach().cpu()\n",
    "        mask = masks.squeeze(0)[i].detach().cpu()\n",
    "        ax[0][i + 1].imshow(texture, vmin=0., vmax=1.)\n",
    "        ax[0][i + 1].set_title('Recon. Slot %s' % str(i + 1))\n",
    "        ax[1][i + 1].imshow(mask, vmin=0., vmax=1.)\n",
    "        ax[1][i + 1].set_title('Mask %s' % str(i + 1))\n",
    "        ax[2][i + 1].imshow(texture*mask, vmin=0., vmax=1.)\n",
    "        ax[2][i + 1].set_title('Recon.*Mask (%s)' % str(i + 1))\n",
    "        ax[3][i + 1].imshow(image.squeeze(0).permute(1,2,0).cpu()*mask, vmin=0., vmax=1.)\n",
    "        ax[3][i + 1].set_title('Image*Mask (%s)' % str(i + 1))\n",
    "    for i in range(len(ax)):\n",
    "        for j in range(len(ax[i])):\n",
    "            ax[i][j].grid(False)\n",
    "            ax[i][j].axis('off')\n",
    "    plt.rcParams[\"svg.fonttype\"] = \"none\"\n",
    "    # fig.savefig(\"comp.svg\")\n",
    "\n",
    "if model_ == \"disa\":\n",
    "    reconstruction, textures, masks, slots, edge_mask, attn = model(image)\n",
    "else:\n",
    "    reconstruction, textures, masks, slots, attn = model(image)\n",
    "    textures = textures.clip(0.,1.)\n",
    "visualize(image, reconstruction, textures, masks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 792x504 with 15 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def visualize_perm(image, reconstruction, textures, masks, reconstruction_, textures_):\n",
    "    fig, ax = plt.subplots(3, num_slots + 1, figsize=(2.2*(num_slots+1), 7))\n",
    "    ax[0][0].imshow(reconstruction.squeeze(0).permute((1,2,0)).detach().cpu().clip(0.,1.))\n",
    "    ax[0][0].set_title('Reconstruction')\n",
    "    ax[1][0].imshow(image.squeeze(0).cpu().permute((1,2,0)))\n",
    "    ax[1][0].set_title('Input Image')\n",
    "    ax[2][0].imshow(reconstruction_.squeeze(0).permute((1,2,0)).detach().cpu().clip(0.,1.))\n",
    "    ax[2][0].set_title('Permuted Recon.')\n",
    "    for i in range(num_slots):\n",
    "        texture = textures.squeeze(0)[i].detach().cpu().clip(0.,1.)\n",
    "        ax[0][i + 1].imshow(texture, vmin=0., vmax=1.)\n",
    "        ax[0][i + 1].set_title('Texture %s' % str(i + 1))\n",
    "        ax[1][i + 1].imshow(masks.squeeze(0)[i].detach().cpu(), vmin=0., vmax=1.)\n",
    "        ax[1][i + 1].set_title('Mask %s' % str(i + 1))\n",
    "        texture = textures_.squeeze(0)[i].detach().cpu().clip(0.,1.)\n",
    "        ax[2][i + 1].imshow(texture, vmin=0., vmax=1.)\n",
    "        ax[2][i + 1].set_title('Perm. Texture %s' % str(i + 1))\n",
    "    for i in range(len(ax)):\n",
    "        for j in range(len(ax[i])):\n",
    "            ax[i][j].grid(False)\n",
    "            ax[i][j].axis('off')\n",
    "    plt.rcParams[\"svg.fonttype\"] = \"none\"\n",
    "    # fig.savefig(\"comp.svg\")\n",
    "\n",
    "if model_ == \"disa\":\n",
    "    reconstruction, textures, masks, slots, edge_mask, attn = model(image)\n",
    "else:\n",
    "    reconstruction, textures, masks, slots, attn = model(image)\n",
    "s_pos, s_scale = slots[:, :, -4:-2].unsqueeze(2).unsqueeze(2), slots[:, :, -2:].unsqueeze(2).unsqueeze(2)\n",
    "slots = slots[:, :, :-4]\n",
    "# s_pos[:,2,:,:,1] -= 0.25\n",
    "# s_scale[:,2,:,:,0] += 0.1\n",
    "\n",
    "# permute textures\n",
    "perm_ind = torch.randperm(slots.shape[1])\n",
    "slots_texture = slots[:, perm_ind, :slots_dim]\n",
    "slots_mask = slots[:, :, slots_dim:]\n",
    "# slots_texture[:, :] = slots_texture[:, :].mean(dim=1) + torch.randn(slots_texture[:, :].shape).to(device) * 0.035\n",
    "slots = torch.concat([slots_texture, slots_mask], dim=-1)\n",
    "reconstruction_, textures_, masks_ = model.decode(slots, s_pos, s_scale)\n",
    "\n",
    "visualize_perm(image, reconstruction, textures, masks_, reconstruction_, textures_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1.2402148058754392e-05, 0.040451530367136, 0.04046393185853958)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "var_mean_tex = slots[:, :, :slots_dim].view(-1, slots_dim).var(0).mean()\n",
    "var_mean_shape = slots[:, :, slots_dim:2*slots_dim].view(-1, slots_dim).var(0).mean()\n",
    "var_mean_tex.item(), var_mean_shape.item(), (var_mean_tex + var_mean_shape).item()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[0.0000, 0.0655, 0.1167, 0.0826],\n",
       "        [0.0655, 0.0000, 0.0780, 0.0853],\n",
       "        [0.1167, 0.0780, 0.0000, 0.1360],\n",
       "        [0.0826, 0.0853, 0.1360, 0.0000]], device='cuda:0',\n",
       "       grad_fn=<SumBackward1>)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "s = slots[0, :, :slots_dim]\n",
    "(s - s.unsqueeze(1)).abs().sum(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[0.0000, 8.0286, 3.6597, 4.8766],\n",
       "        [8.0286, 0.0000, 7.6119, 7.4978],\n",
       "        [3.6597, 7.6119, 0.0000, 4.3791],\n",
       "        [4.8766, 7.4978, 4.3791, 0.0000]], device='cuda:0',\n",
       "       grad_fn=<SumBackward1>)"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "s = slots[0, :, slots_dim:2*slots_dim]\n",
    "(s - s.unsqueeze(1)).abs().sum(-1)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "ISA",
   "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": "3.10.6"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "bab60645eafc042904fa343452c0e41c36880faaad87b24731ba3c6ffa337918"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}