vggish.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "1967c1d8-9536-4f95-8213-4cd9a0e14175",
   "metadata": {},
   "source": [
    "# Use VGGish pre-trained audio feature extractor\n",
    "\n",
    "This notebook demonstrates \n",
    "- Using VGGish in pytorch to generate audio features\n",
    "- Using VGGish with OpenSoundscape\n",
    "- Adding a classification layer to the feature extractor with OpenSoundscape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5ca9ad40-93da-41dd-9fb4-a740d9cf5343",
   "metadata": {},
   "source": [
    "## quick example\n",
    "this very quick example demonstrates the use of the pytorch model zoo vggish implementation:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "id": "a90a1dcb-57bf-4c88-9480-2a22a94d4f3e",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using cache found in /Users/SML161/.cache/torch/hub/harritaylor_torchvggish_master\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "tensor([[158.,  24., 142.,  ..., 218., 117., 255.],\n",
       "        [162.,  37., 149.,  ..., 165.,   0., 255.],\n",
       "        [160.,  30., 143.,  ..., 192., 191., 255.],\n",
       "        ...,\n",
       "        [157.,  37., 148.,  ..., 138., 189., 255.],\n",
       "        [162.,  37., 155.,  ..., 156.,  14., 255.],\n",
       "        [166.,  18., 156.,  ..., 172., 136., 255.]],\n",
       "       grad_fn=<SqueezeBackward0>)"
      ]
     },
     "execution_count": 183,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "import torch\n",
    "\n",
    "model = torch.hub.load('harritaylor/torchvggish', 'vggish')\n",
    "model.eval()\n",
    "\n",
    "# Download an example audio file\n",
    "import urllib\n",
    "url, filename = (\"http://soundbible.com/grab.php?id=1698&type=wav\", \"bus_chatter.wav\")\n",
    "try: urllib.URLopener().retrieve(url, filename)\n",
    "except: urllib.request.urlretrieve(url, filename)\n",
    "\n",
    "model.forward(filename)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8bbc55f5-0964-4418-82c4-60061fc54aa6",
   "metadata": {},
   "source": [
    "Let's look at the features it's creating"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 184,
   "id": "b114504b-738e-4211-88bb-3a862fe29c38",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "id": "852ae307-a5c4-42f7-b07d-98c8a07c4da0",
   "metadata": {},
   "outputs": [],
   "source": [
    "output = model.forward(filename)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c10274cf-f3c6-44f6-8dac-d0c3b856269e",
   "metadata": {},
   "source": [
    "the output of the feature extractor is a vector of 128 values for each audio sample.\n",
    "\n",
    "In this case, the model has internally preprocesses the audio file into 19 1-second Log Mel Spectrograms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 188,
   "id": "4683c644-0369-4e17-982d-5a8a473eded6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([19, 128])"
      ]
     },
     "execution_count": 188,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "output.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 189,
   "id": "2d8cfc93-db38-4bf8-a889-043704f245d8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x2d6c01f40>"
      ]
     },
     "execution_count": 189,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAh8AAAB4CAYAAABSM5AaAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/NK7nSAAAACXBIWXMAAA9hAAAPYQGoP6dpAAA5e0lEQVR4nO2dd3hVVfb31+3puSQh5ZJK6B0SE4OoKKGLIEVFEERn+KGgMMyIYIGxAeroYBtwHAUdVIRRsdNCZwIhQOiBBAKE9EJ6cut+/0Dvud8dEsqrNxldn+fJ85x197nn7LP2PufunPXda6uEEIIYhmEYhmHchLqlK8AwDMMwzO8LHnwwDMMwDONWePDBMAzDMIxb4cEHwzAMwzBuhQcfDMMwDMO4FR58MAzDMAzjVnjwwTAMwzCMW+HBB8MwDMMwboUHHwzDMAzDuBUefDAMwzAM41Z+tcHHu+++S9HR0eTh4UGJiYmUlpb2a52KYRiGYZj/IVS/xtoun3/+OU2ZMoVWrFhBiYmJtGzZMlq3bh2dOnWKgoODm/2uw+Gg/Px88vX1JZVK9UtXjWEYhmGYXwEhBFVXV5PJZCK1+irvNsSvQEJCgpg5c6bTttvtwmQyiSVLllz1u7m5uYKI+I//+I//+I//+O9/8C83N/eqv/Va+oWxWCx04MABWrBggfMztVpNycnJlJqa2mh/s9lMZrPZaYufXsTEvDeX1J6Gy9sLq+E7mX8JAtv7tN65/e4fl0PZy1MmgZ013QB28HYd2MW32sDuEFWE9bWjy8p2hjm3bb1qoMxSowdbXYPfDd4nwC4aACYZj+PIceET/wb7o6L+zu2aoWVQ9tXpo2A/X9wd7N1vJIJdY8Jz+V60g10wFP2i97I4t9svwvY5d68J7IZwK9j+R9HnZiOYJDRoRyzdB3bu/MQmyy7Ow+tqcwqvw+qD1+mQ7gDj+Dywz54NAbvtXuUL2gnFUNYv6CLYWUPwQrJe6gm2tgbrMjp5L9h3+p4A+9XZk53b5+7GY0f8KPWlBLywiC11YJ8f4QV27OpSsC2hvtL+Sl8O71YIZRdy8X7U52O/972AddOPRb9F+V0CO3dZB7BLeyjXkv7wv6Dsnk7oU79NAWDnvRcLtnpSCdj+Hg1gvxH9FdgnLIHO7QWH7oEyax1eZ2Aw3gfvdf0E7BmZD4Bdfskb7D23fgD2fafGOrdDvfDYBX9rD7bNE/uS9/p0sBP/awE79S/xYBcmeoLtf6vy3Ovojz7bldURbFJh+4aHYHtaVuM95HMRfV7cF/viBzPfdm5HaPHt90MjJ4Cdcz++SffLwaoZP8Vwv/xcfPNSDNiz2ygHSH7mYTwYXiZtWfwh2BOGjADbdi4X7Ow34sA+etcqsMcn3OLcVvn74ak98HfLfuoM2M8ePQj24v63ge2oxt8mtRe294VZvZzbC+5fC2X1Avv5G1+PBvvAZKXfVtU4KKrfOfL1xefHlfjFBx+lpaVkt9spJAQ7XEhICGVmZjbaf8mSJfT88883+lztaSCNl8flSqotUpkH2BqD4hxvX7wJtRpstEbf1eukcvyR1XpLjS4NPjQG5XgOL/yu2i4NPmz4Xa0Oe7Ma+wNp9HgtXr74g6NzGdxoVXgdfpIfDHVYrtXJPpT8psMfbdkvGi9lf7l9XH1y+btYb9nnUhORPCSWr03j4dF0mXRu+Tockk9V8rm8r9ZftE3uq/eRfCwdXD6W2ia1kfR9b6m9tVrl+7JP5b6k8ZD6mtaB5/aQ/CQ1gkMr1dXDpa9dzUce2O81eqyb/H2dN+7fuG8q1yL3a7n9r3YstXxuD/SLr3R8L4viZ7WXdCzpoazxwvvARzqWxkvymxmP1+jaXOqq8zJjmXRdpGveLwYfqQ20so+l/uBybr2P9ByT2lsefMjt65DqqpXuOfncrn7zkwYfcj/VeMj3Jx77as9FD6u2yXKNVG/pMhu3l1p6kKnk35artLdKqbxKOpaQrlulkp8VTR+LiMgh10Uqd20D+XdG/g9NfnbI13G5fleXTPzig4/rZcGCBTR37lynXVVVRREREaTVOkijvfzDYTf6wHfUBvxB8S5QHiAJBnRyaT8cQZIZvyuPZj1ysVHCe1SAPSLgCNgrdiv/nZztjo3WIQbfmpzJxDcCPufxv9HYtdgc+f1xNPJcJo44E0POO7frY6IIyQBrS35nsEukNxmacvRbxJcFYJvm1IKdVdbWuV0ZFwplFn/JqZLZJhsf0g4NdtTamZVgm0fcBLbB5R+r0v9LgjK/c/hjUtIXb4y28dgmFdux7vVW6SatwzYtTnLpPxfxv+ySSuynsbH4g+F7Bo8Vth2v0/+uerC91Pj9qiilb2qMuO/FB8GkDm9g+aWu+J+ItgF9bgvCuhc8jufW2ZW3VyPCjkHZFx8NBtss3XJ9ZuA9s+MsvtmYFvVfsF/v3g1shzR4cUUb3g7sSwvwP+H2r+A/PKujt4N9xor/Ef7fXX8A++yENs5ttQN9FrMLfXRudBuwu/bD/+gbtrYFO3QwvkE6YcX+UbhHubauI7HfZgVLb2AHYl008XhfXDiN16m7CZ8t7V7BNlgzS7EHpP0RygZ3wzdycb7nwd5a3gXskuJAsIU0oPAdjn54t+QO5/b+okgoC8g6DbZ/Nv6Ta/Ft/kdPbu/3jtwK9pMDlTcKby5+G8oe+O/0Zo9d2wX7niEH/RKULv0+1D4KtucflLqrpZ+ptx//B9gvt+8D9sel+Nr8zD/xzVjMxMNgO+rwtyfiJaW9730Mn0t9998PdvQzUgRjGt0Qv/jgIygoiDQaDRUV4c1SVFREoaGhjfY3GAxkMBgafc4wDMMwzG+TX3yqrV6vp7i4OEpJSXF+5nA4KCUlhZKSkpr5JsMwDMMwvwd+lbDL3LlzaerUqRQfH08JCQm0bNkyqq2tpWnTbvD9DMMwDMMwvxl+lcHHfffdRyUlJbRw4UIqLCykPn360IYNGxqJUJvD4VCR6qcYq7pB0ghYMF5ZE668wDltRW1CwHGMbdVEYBxWX4PBNSHpD8rMqEg/b0Flv6ZB0U4ISb1e5ItxdFMsqsar22OM0CcXVeANbVG/0LMNzkjIrlbqIqT4okxpNsZdSYfHdkjCu6p+YWB380LVeHp2tHJuE8Yy/bMwRt8wFNvgwmAUBcR8jfoE2wb0ceFDqPS35ivnM57A9vIsQS2LRzmWV13EPmgfhMcuOiz1UUnIOTJeiZ1WWLEfJvqj1P79YSPB9juPfe3iEH+wM2sxLBnvdRbs4E0XnNu1I1BfQBnoU6ufJFBswPYdNDID7CNJqEfy/xD9UBmr3GPZ0dhvjcfRh+Zg9MvNfqjMTynvAfb7ORivDshEP9kNSht+UIk+shfhPVX0Dvoh70fUj6yfgrNAxuAtS5mP4fdN0Ur4WK/BeuVp0WeG0ub1BmJABdj5BdiGf2iYArbKpSsPM6Ju5sfbcPba1oFvgT36zXlg97oNZ16UrcBnas7z/cHOtSszrww6vKeMOrxfl+zEfj57wGawP+qGGpDgd1BfUjkH2yi/Trkv9vfDmRdDqQ/YFZ3ApNi3s8FWtcM2Gv7pk2DrO0qziGyKJmTi+r9AmV+WFCi4A82GQHwOymKC0oTm9Ya1IYqf/feiqPOd/EHS0XB246bN/cAO3SfdQwOxXLMdZ8eUzFCiEp1XYoSi/yDUeNlTJVHXDfKrCU5nzZpFs2bN+rUOzzAMwzDM/yi8tgvDMAzDMG6FBx8MwzAMw7iVFs/z0RQq1eU/IiKHlLRGJSVM8slT7E46SXeRiLY5GOOXssZDW4e2yQvnPNtJiuu6JFPRtkXNRq9gzJWxNxVjn9FFGHclaZkdlR3PlX4Oc3k81nuHc3sjNR+H0wZjnDa5wymwXeOsRET1j0s5SuZgjoIuUS7X9hrmkFDVoR9ip+eDfWg1xh+11ahPqIjD2KlfKh4/KlWJy54Zh+1b1QHzdARlSNk1a9CukXysi8UYMB1Cvxp1in5lUza2Z6odsyV2lPJ4FAxAH9d2wva/w3gS7EGeGLed84DS/vWl2I+9++G5tC8dx3O/jZlf89N7g63ylpLr3YM6HVuZouOYFrQbyp7zQT9cGI6PlUf8MY/D6yGoy7La8X8g60OYIVOjVtpsou8FKFtrRQ2IaXwW2AE7sO/sqcbsnGO8D4FtPIJ1L89Xjm8OxOdOWAbaGmvT+UiILuvYACtet13yQ32Eklvl2/K+UBb5Jd4ja+OxPOgY9q2jDtRVeN+NdY9ahDqMXn90SWj3NeazGf406k9eGZWBdhn6uD64eb8khF1osizDbG6yjIgo9lPUPpx8EZ+RnabvBzugN2r+is+gFi5Mq+j0zty3Ao+1EzU5MuXdsX39pfKob9APFimZV+Fw5fv+5zAr9GwT6mheIHyGxt+O+WwyOmH+m7Yf4XVLacQodJuinXp948d47omYj+T8SDwWPbyTbgR+88EwDMMwjFvhwQfDMAzDMG6FBx8MwzAMw7iVVqv5cDhIyfNRj/EvtRbj+nZd0/PrddWSjsID4+g1oagnMQfg/r28cX78oRpca6A+1DWnAcYnsyswX4XdG+OsNe1wHZk2RyrAFjrMl9A3EuuysUiJ46p7YU4ReW0Xu7SI2ZbNGCO2e+B1Rw2JADspYBvY5+uVa8v2xJhw7miMo544iHkhAoOwLmobxuX92qDuQl+BfrowRLnWwCNY77pQ7AsOKbhpnVAOdsBazLVgnlABtk2S5Ry8pPjFW1rsq6IA9SGXussLyeGxgndgPy6/GdvQLnD1V9e+rPHFe8J2FCPM6l6owwhKR5+XJOF94Lcf61rZE48fEKnoMHbVYXIFu7TIXVsMs9P9N90JtrcHOtVbj/bFI6jj8M51WeyrD7av2hs1PxWjcZVbcx3qrmaEYT+WMf2AKxOffSjcuS0kDU7B7ejT8C14LLvA+10cxDbyknQ6a+Nwxd57DyjrzNTapHsAF1Clz3NQA1AxGc8dHIR9yX8+Pvpxb6R+RBXYf9yH2ofYUMy1MiwE9Uahqdjxa8eh/ij1ImqAHuik5GIJ10o3jUTmPHx2dH0N6ypl1qB+bbF9D6/CZzSNUzatAr+tPyQ9Y3HhWIr9BLVKsk/rpPV4SuNwjy6LK5zbmY9jvWYdxRWRTe1RV5dZhr9jhu34LNKYpQeZRFmiounrqkdNR50Jnw2RG1A/SNLiv9cKv/lgGIZhGMat8OCDYRiGYRi30mrDLq7I02EdNrRVLm/eaxz4OsonD1/blVlwvKUx42t7uxe+Cssx4xRTiwNdJlzeONss+Pp5XCRO4/uwFlMYCzW+xrvUywi2/N5OPnd2gVK3TmdwqWmZ9iYpNXstpk/3OYPHvtQRX/O21+Or1Zku00LHn8HXdB5l+CrU3B3bwFCJfmowol2XacS61aAjVC5OVzmw/RzSct02jFyR2IQhoeJkDJ2IfHxd6SXdISfPufhN4LkW3P4d2MuPjJFOjqZdyr98yYYhBI0K+2roGmU6XWXnzlCmwzfXZAnCY9VESPeMF7aJ+TZ8leq7F9uw2ldxZLEFfaSrwle6IgTDSafL8B6qzsQwXZkntm+bLKxrdbTiuNQGdJqjVrpwifMnMYSTFY72PwpwCur5+8LBVrt0j1t64DTe/du6gp0/ABtYbr/6dujznKRPwF5bg3Wb0UWZ0vx4G1w+of0hDH3N6LgLj/XYMLCL4jD0GZCHUzM1bbGNKh1Kf7DZ8P58su8msA9U4/TWUT6Yjnudz1CwfdbtA7t2CoYIvzynTAN/NgjrKROwB59Tdm99E3teRg6j7zH1a2JPoi6fzwS77UXsp5PPDQQ7fzD261B0A9UHY7/W1GP/EJ5K3TWB+Fx6MBaXt9h4Fu/BhjT8bVHfiSE93TsYClP3wX5/CU3A5oH1Dv/bmSb2vD74zQfDMAzDMG6FBx8MwzAMw7gVHnwwDMMwDONWWq3m43J69csxVJVVWva+DqtdH6TEpBqk6VFqK8bpDIUYE/SoxHJdBY7HdCo8XpgBY2m5lUocV0jTWT/OwillWi0ey7tQmi7pid/3zsPrbH876jbKghSthbDgsWTOnMClpXXVeK6GIIxX+5zHON8Hebjs+TMlinai7UApda+UEdlei9dRHI/HDk3FNnBIMcbSPlhXi0k5gTEL9QX+OZI+xI7XZfaTpvmWYH8I6Ylp5YurcWl515TYvTtjWuhYPX7X9GMe2DmTMOWxvgqvs86OdUkzS23aVokpexRJ/bQKr7M4HrURvuexvD4c28R+EfdHrxLZypXpdnODUF8wVYOp2oX0L023IPTLwWrU3XiUoqZAkpSQyqbUXadC3UT1fTeDXTxc0vA04HWO9TkLdpwHainGhaGWwjtaud9zqjCmL0l+yBGE7XXGWgN2UGQF2DHfTAe7Z1fsT8cPRju335B0MeEpeO6VsbgMun816nAembYV7E0fSSnQ+6FuY1lZnHN7XKcMKPvs4k1g55bgdPXsKtSPNARgh5CTAngbsK4FJcqU5OMWaVqnhMVP0gPqsS/J/10vO4bTvq1xePwvapTOp7Hgsati0F4dvR3s4e/hdTSaahuGn3gUS7U7lePcDF7fB4qOt8fnN6klv/TBKcaWU3gTyVPvrUacPmvzUeo2+OQoKNvz6j/AjntFWq1+Pj4PrhV+88EwDMMwjFvhwQfDMAzDMG6FBx8MwzAMw7iVVqv5UKsFqX9aSltoMI6na4O5PNQ2JYoYpMH8BpWx0jxwaRq42iZpHXAaOF2ox3jmX8Jwjvt/vRRdh7pKylehl3ISSPHnsBqMEerLMJ5d2hPTMX+zNw7s2+NOOLfzrc2nz5VzTAT0Kbnyfj/h/1Q22CUjMBauOqHkgSjpKy1bX4mxUc8AjE/qj2I80iGlx3dIaaytGhwj+x1W/OrQSnoRKc+HdwH6tLwbtpG9DfotvwDb21O6lqzJ7zu3U+rxWDO+wBh+eGc8t5BSvfvm4nXuKowFe1KbvWBfildi6bOnrIey1RdQX0T/wLi79xmMCZcNx5ivIRt1OzU9UTvhqo3556UEKBNqKbeKtOL2qXLMMTF5giRYkPj4q0Fge+cr27d4YF8wHsEl1W0emJZ63F8w53kbjZdk47k7zEGfZ/9bWYag7gzej4ZqvO7o97E9Y4egusG2AetmKsL9y6Oxbg5vpXxG0nYoW3NsMNhqKRdHSRzmaXlzD+7ftS1q1zzPVYA93O+wc/u+7TOgLKkT6mb+krAR7Ao7PoNfuHMk2G0P9AC7OAPr6tlJ6auFdlkhglR3Qp2Nby72a9+bMN2+z0Y8XiVKX2icj3LucVOWQ1nvtIlgn7TUgV0xphfY/usOgj1+IPat/xzDJS7O/VupTKAvpsOP9cLndZ7OCPbPv5VOJD2SORjbRLflANgxauW3ZekHX0DZkjK8rrDtuEQFzacbgt98MAzDMAzjVnjwwTAMwzCMW+HBB8MwDMMwbqXVaj6EUJGQJ9L/hPUSxvWCCpteEFqeB67D1dpJX4Fx+UsdpKWmpaQFoRqM0xq+V9YP18djfn1bCMbhAk0VWJd8jKvbzuE8fxqBxzPFYtxvZaQyv3qYDuPwMgYpl0JkP1z+ea4JtSyLCPUlpQUY726bo1yb1Qd9HLYF45WZvVDjYZf8ErYL9QgDFpwAe98zUl6Bwcr55Nwo0Z+iaOfkPJwfbwjGDrCkF67H8n7urWAXn8C1Plx57FPUeCy9D9fqeLZuMti+5/C6vb7CNS7ueRnbd9YpjDH7fabs/9b4O6DsJhP2nT034zoh/oGoZdGdxjaT15kRdiwPcFka4t4J6VC21Yg5YGQNz8F+a8F+5ALuf6gY8594oIyD6kKV422X1sOo7oK5N4L2og+fCsT1WK6G7U7s98O7KAt0HF2L+Uy8th4FO2celstISzNRwxS8Bx+O3A/295MUDcBH76Gmp91R1FFtfhp9HFsxDWxPT9RG5NyL+pPIv/4X7ASDkunF5wR2jjkD8VnRVo3PsTtTHgTbtEVaT6usEOyDk/8FdtzKPzm3p9dNgbIOhOtleeRjRhrjfsytY8vBPC6+izGfyRuxX1NTvFaOGqyas/gM7JogaXSkNciEpMPLQIlHo2s5+6qSq6UuB7UpTz/3Jdi7uj4A9oRYPNbXWtS6WPYbwZbz+Gi3KhqQeWfGQ9lbHT4Hew/1oV8CfvPBMAzDMIxb4cEHwzAMwzBuhQcfDMMwDMO4lVar+bBZ1eSw/qxTQJ2FPIfZ+yLm/XDFIQW3GkKkufVdMZ5pRzkJDQk4BvY5GyYKsQ1SYsQaqRoP98Kc96drMQ5/IkGagy5pPuqjMWYY4InzyldVueRP6CFNWKc0sIzZqItJD+4A9hpPnPdf8SDqLILCMBBfmqBoCKK+lTQ3xbivSo3z+K3heF2l/TCW+kpIBti3SXoWzyhFt2G/hN8tHIYaDXU96iwayrGBl53BnBLFJahP6bIe49MXH1fW69DVYke0Sok8pFA4NQTi/ppA1Cv08cC59zN6Yfx6qOjj3K4t94QyfThql1Tta8GucuA8f0swagC6LUStzKnXsa/+feE/ndu+amzvkl7SOjEe6PNni7Gfv2TaAPaQ/8wD+44pqH34NkPRUlQ4MM7ufQ7XT1HVNf0sICLaKRXfJt3v+r0nwf4xs5tz2xgm5RtK6Ax2n+TMZs8duhfrunEe6jRm5aGuo/JO5Z7W7JX61g7UaMiM6oZ6lDFGzDnxJyl3R3O0SS4A21taXydGh/oEtReWW72ldUQkHcZb5aiVMbrIdMo8pAaSUEvLHwld8z9p27qjxiN+0aNgpz+v5Paok4RQcp4emaA96CdbE/s1Rexa5bl2dqxvM3sSOTJQF7eoLdrfXMBcKgX34DO3A6byIFVcd+d2SjfUrrXfMhNsUydZMXJj8JsPhmEYhmHcCg8+GIZhGIZxK9c9+Ni5cyeNGjWKTCYTqVQqWr9+PZQLIWjhwoUUFhZGnp6elJycTFlZ1zfdjWEYhmGY3y7Xrfmora2l3r1708MPP0xjx45tVP7qq6/SW2+9RR999BHFxMTQc889R0OHDqUTJ06Qx1Xid64IcfmPiEgYMNimqcUxk64M49t4IMn0xHi1zROP7Z0n5ebQYpw2VIOBfLteqYvxDOpJpvkfAXthPa634VnS/Hos6mpsnioz+u/5bWOUenVpfhxp9ke9gc9Z3P9bSzzYKkznTyFajGCevec95/awf03C7xpRN9E5vAjsMaEZYL9zdAzYaWYM5Hp+jfqVre/scW4nXfgzlFnx1ORZJK0Lg4eisvGonVCVSWsBZeE6FmurFMfI+qD7fTFvwzunm9cqnVyMeQQ66L7FHcibmsI3CPv85v3YYOOS8EI3HEgC26cXanxsBahtGd0Zc7W4rqky7cJwKKuPwL4RvAfvqXJr09dBRGT6G+oXFs7eA/YGn67O7dfnYV+zdUWn+n16nJpD1njI63WE1qHmo+NbyrW55pchIqqrwr6yJmZrs+emvUeaLd6V1x4/MCl+jFx9Boqupif4Iasb2N80oK6iXb7UN5thZ8+vpE88r7jfz5wZtBI/QFkVDV3VB+wOHtj3jKeVvu13vnmhRfhi7DtXu6rerz0Gts4hmtizsY5itfl2sN++hDlDnD9YN4hIV/SFycuv/bfyStwfg/qxL18e3MSelylKUrRzm+pQ09Hp7/ibJw6hfojeuoEK0g0MPoYPH07Dhw+/YpkQgpYtW0bPPvssjR49moiIPv74YwoJCaH169fT/fff3+g7ZrOZzGbl4qqqqhrtwzAMwzDMb4dfVPORk5NDhYWFlJyc7PzM39+fEhMTKTU19YrfWbJkCfn7+zv/IiIifskqMQzDMAzTyvhFBx+FhZdfn4WEhMDnISEhzjKZBQsWUGVlpfMvNzf3ivsxDMMwDPPboMXzfBgMBjIYDI0+V2sEaTSXY2gqG+o07N5oV/QKbPL4eimKUycvA3Pl5WOc7KzuArbOF2PKGrNywNzBGJ9M+Hou2GppXYp2Phih1AfhdeiqpPU1PDDOH9RD0SNUrW56DRIioqoBmODAdw/GbQM6laN912mwPXfggDLp8Dhl30u4XkruOKxL9TmMGda9hmt5tFFjBDujAWOpZY+gXuHmFCW3iq5GWlfmv3gshw4b3O9POLi1fRgN9m1z9oK98c+4vs6mYiVHgdUHjz3w2BiwPcpRu2KXtEte5zC2+kYpXufiENQIlP1BKQ/xvQhl5goj2Cnv3wy2WrrFyotRHONzF+ZS2ZyLbbrKWznf3txoKPM7iY8Rz1K8bpOhAuyn8kaA7UjBXC1BGtSI2MzK8Xe9+x6UDTX1AVulQx3G1Yhpg/3e7I3nroxVbP8z2N7GQ7iOTHM5I4iIxC1Y15sOdgJ7Qntcn2PDakVjcOZR1AdFLUIdlaxd0WRi7o3kkRin33O4H9iYPQWRc6McbcC301P9cELB62V47M0F+Az1JtRR/fXIKLDb/FXR2TV8ic+doB3NVJSIzCMwP5Hhh/1N7PnTubKbzwvjil3SCz7eBvOVfJeD6yddjbz5+GwJ3avU5VzNlf9Z/xlthPy8zwBr+yjM81E0B/f2+xyfRWFbS53bQ57G+3exCXuHAbvpDfOLvvkIDb2cmKioCG+MoqIiZxnDMAzDML9vftHBR0xMDIWGhlJKSorzs6qqKtq3bx8lJSU1802GYRiGYX4vXHfYpaamhrKzs512Tk4OZWRkUEBAAEVGRtKcOXPopZdeoo4dOzqn2ppMJhozZswvWW+GYRiGYf5Hue7BR3p6Ot1xxx1Oe+7cy7qGqVOn0qpVq2jevHlUW1tL06dPp4qKChowYABt2LDhunJ8EBF5GqykMVx+MSPUWE2VpzzLvek4r1VKka+y4cuekPR6sMu6YT235mFc9pGemINAu1WZT+3TC2N41f0wnqgKwJihz1GMN9tKcU0UeS2BglqM06f2VhL0jyi6B8ou2jA/iVqD+pKKm1CHEeePdcGIP9HxVMxB4Fq3S9OkXAs5ON+9ul5al8KCdfEowLou3Yox4Ihi3L+ySGlv2UeeF7HmmXMw9k1nUW+iGoR++P5sd7B9S/BaNnT53rnduWQK1qse+453IGo6NGY8lpDuvlgPzK0h5xHwzVVisYmB56AschDmGNmdguuphO3BOK5aWqPoEi5TQg0NWPf3nldy+vztxVVYz1fGg31qBvrh2SBc82SVHvvaQ1F43R0/Ru2E3iUdzt1Zw7CihLFxYW0+d47M+o4bwR7WdTLYFR2V50XbDHzu2E9j7g17MuoTZMQLeH/balBfsqUQtREFA5T7KrI7rvMjczjhM7BjCqaDve071GFE3oXrSNHfmz7200/+H9gvvfo+2D5qbO8gHd6DxQfRL+3jUG82JAb7x9eH+zi3o89Li7dchdpQvKlkNeHhJ/8B9mN5N1NTfFGDz1uPwuZzjpz+AHMldXokHewLi/D3IfJ5zFFy6XtlLR/z+mg8+FNo2nJR89WoLjNMYHd5DrWKdgc+U+0nFI1f1z0PQlnHeaiTs2y6Pl1VU1z34GPgwIEkmkmmolKp6IUXXqAXXnjh/6tiDMMwDMP8NuG1XRiGYRiGcSs8+GAYhmEYxq20eJ6PprA71ESOy2MjtQVjrcKGMUO7vulkHUIaXgUcktaFycc1LjSxGL+M8MNYul3guTSBAc5teQkLbT5GHL274rEcfs3Nrm9M+YFg/MBluQZ7dg4UhWtR6+C9G+1DT2PsU2Yo9cEPJD92XqbMcT/xHM45b7MI1xXpNN0I9vFjuO5E8BZch+CJgRVgLy/HdP4Rm5W4/qVO6GN1CX633feYQ6KsO8ZtT85AP8jxTn21nBhG4ZV+X4L99vk7wa6Yit+tOoF5XDo8nwH2SxGodfE5jboL0yYlS7DmVTz2B5G7we5bgGu9eOZjjpiiaeiHRpHUbOwvfp8p51752AD8rsuaFEREZMGcIXLs/PntY8AuGZACdvh2vN89CpW6f/OHDVDWaTHqQ6J+RJ3VvWfR52vb47lkNHmlYK955F/O7fs/wLw93m1xBl/I+xjjp2fRPJOHazu12YN91zEW9SsdP1K0E8vXfwJldz09T6p5BljRHTDdgW831Dad3o4arihqWkPg/cU+sCuW4nPrzwWoi5PX8jF0xWesOIBrpjwciLl10r9X8viMfgM1OT92N4J9/nnUUfiex46s9sK6JizA/vL3Re+CvalOaZNxPpgkal4g3nN/LkAdTfS65pNGWTrWN1v+QLSSk+TDO5qfHaqScmMNeAJ1OQHST4v9K2nRq0FNL2MSswj7yolnMU2G+vkY6RvSglnXCL/5YBiGYRjGrfDgg2EYhmEYt8KDD4ZhGIZh3Eqr1Xw4hIpUP+krhE6aX23BMZNXiZz3w+U40pTkCimfgU8exoSt3s3H7XzVGPcT9UqMWWCInlQxGGevqUU9iboc14aQ1QVqC9ZF1/cSNYU2OlL6JAOs2gGYS+OflTgP3FuNcT4ZWyDOtz8/Odq5HXAI46z2gRgLNahxnRhH46V8gNE+qCH4Mn0I2FXRSqPWtcNzlw7GeGRdCPrQ5o37y364u8NRsDMk7YQri46jRmNIJOYr2LwK47ZeUjcuH9cb7OTeuJZLWjC2qWN/H+f2x3vx1t3bAa/bKqU3sXti53RkYlDYW0ojUdEV/VTxoHItGhuuaaGW1kPp3g3zAsix83diUY+w4tBtWLcJeG5NOep2XOn4D8xXYbuIF7K2fUaT370Sog3Gxt8sVFboDt2HOUR0m1DjkT8X9QdyLNyYih3/wCJc+2VRCeaY2eWiKRmyZxaUdXwP+xphcSNO7I8GW2e/8n5XQtO5A9hjvDMkG9eN+bgqCOyDhc2vO9VLj8/FGpNyo7zzA+q9YglXR49ahLkyTq9AvVHAh3Vglw9BTdCSCyPB/q7Tj03Wc9Tt2N659biWS0kfvMfaoTyJOkxuflGUd35Ucti0PSAV4mWRMOPz+s3X3gZ7wjdPgN3wLT5LQpvR+PT79CTYpX9HjZfG2nSqjeuB33wwDMMwDONWePDBMAzDMIxbabVhF5VKkEp1+fWO3QtjJxHtMVxRHdH0irmBR/H9YlECjreqI9EFtdJrfIsDy/9WMhCPN0V5dW7zwu9O64pT1P617Q6wsx/FV2FBGfh6siEKX/PeHIqvyjqvVKaNRZ/D15Ey8RH4KnyD9Ir38AFcsrsD4fS3sX3w1WrEzUqK7I/fxlejQoOhDq0a2yAkDV+FVv6Ar3UfmI8pj/2llOleecrxPcvwlX9FLMY2aqPw3B1X43S3JcFYd8MFKSY0F889/ozyGn5d339B2cjP/wJ2+zQMdalsGFjLnoixke4+GDJ4PwFT+Q+vfMC5/UgStneDQ4r5LcW+UrNBSo8vRfAqvXH6+hPJ+M7YY7ASdpthxHom3/QwHms53q+vPYd9y1+Pr76/uhXDD/k2DLP8+YNHqCnkMEv+PAx9xHyD76vb7sX+kbYYz10wEEMGF9cptsAoGVlvwXOZg5sO/xIRTX/8G7CHjJsKdtYjUhveozyrJnXD9/D7y/A6erz1GNgaKYrafh+GgLUnzoEtvJqe9r82ZTUeazOe6+zgD8Hu44F9z3IQwxOFf0K/DXgiEewdy153bo/NvK/JehERaWNwCQLSNz01noho4wAMT9y1Upqy7DJruOt7eJ1yOvSN+Thte8TXAWDb1dhGndPwt+dAaQTY4WplqnWuBn/TOq3CKcIxUvjpsYWzwRb98bnn0F37e4aXgjH0vH1iR7DL9/wyK9Tzmw+GYRiGYdwKDz4YhmEYhnErPPhgGIZhGMattFrNh8WiIY32cvXU9RhLvViEMUR/w7WnV3foUJfREIA7ODwwZhjigTH//HqMR3uWKfvbh+GUwigDpmomI2o4/A+ivqAmXJqLKS17nHkJ06t798SlyZsj6xKmdrbZ8bqDu6CORqa9J5bXucxhbrsC44+WYTeBXWZGXYa6DqftVqbhdVkSsQ188lFD4Nof/DIwJXXeHRiPDN+C7V3WHesS/g2eKy8ZfS5KUQuRcUHRp0yvnwRlY5JRJ5O2C/1QmIjtG7IP69b2LuxrjXAodf1g361Q5HkB9QLRIdlge2jxHvLeg36w+OKp3vt8BNgDRh52bj9vxXtAsx31QCWv4hTj3AaMhR8+hVqnL4xxYJ+pxb56z327nNt7G5qfIxp4FPtWQyA+4koS8PuPXMBphDH3Z4GduUmJd/91CqY4f+k9bH9jlvSwGY3mK3tRX+T3FOowpsTgVM413ylTkMf7Y9kPj84Be+R9qEdI/zP69Pxw1HSYfDGOr98opYZ3IXGfpLmpaH5J9X+XS2nnD2CbeF5ELdS671eC/WKJ8v2UbqiTkZd9yB3bDmy1rvkU5rE61FlFfS/dc9OVTXMsapNsg9CncjqDk3/G+6LTH7CvZU9FbZtVWiJhTy9luYb2haij+uzWFWA//TRqmawT8LegrxHt0o1ySnSkZoKr7iYDyvLOoQ4qaRhOxb1R+M0HwzAMwzBuhQcfDMMwDMO4FR58MAzDMAzjVlqt5kOnc5DGmQNYyhuhx/i1V3HTc7sr22Oc3eGF3217BONypT0wdp5WgPHpNX0/AHt26UznduE5jPllROJ3VeUYKw1Kx7ic/fgpsKvfwvnvGhVqBEZFKWnI00ObT2Fs3opxu+qO6AejCfUq0gLM9M7xgWD7eyux1eI38dihGH6mnAqM+XvGYtxVJaVHkHU3Nk8pbn+zosOo64o+9U/D9q7CqfSN0o5f6oF9S6jw3NpqPN5rY//t3J5/6B4o20+Yc8A7G5NpBHpjKv+aMBz7h2orqDlyxitapw7tMZdCWA9ctrx0Lepock7guUU/1B91fR1j33kvN/1/yaK2uCR60gMzwNbWoE+T/Y+D/dpI7CALi1EbU9qAepT/HlDWRKi4Wc5HgQktDD/uBzvrgwywx2QNBfuDyN1gj7h9LNjhbRVdxvxO46AsJgPPXRnTvBbCGIhaB8dWvC/Wpd8OtoeLJGTsN5jHoeNy1Fn9p39fsBNfPAv2uX24rkR1BD7nsHcQVTqU+7uuDH2eM2EFNceG813xg254rshj6Ic+OzCHxcmBSv6cTjtQ+xBDh8EOex37kmVofLN1kzk9pen8Jm/3/xTsfzyT3MSel2m3QdNsuUPKVxU0BXPUzNmk1D1gJ+oB45Lxu7Y7UX9yMB5/l7qnoh7Jxw/rhko2Ip91Sk6qVxZKeqA2qH2pGC09G3BViGuG33wwDMMwDONWePDBMAzDMIxb4cEHwzAMwzBupfVqPjR20mgu6zE0hRjPVqkwnl0Zo4yh3ijHNSzaZmBsu65RLg3UUdR2wDnpHjbc/8kcjPtWRiuxOaFFvcAXqTgXW2vGWHhVVyPYmhjcX9OAY0NvHV6L63oeIhCPNfgkLveur8TrDE7F66pph/HnM3/D9VViA1BjcPpsmHP7sUGboWyFYzDYgV9gRLlgNMYQvQ5LEUhJwlMVhd1UJCj9QZ2DCSranMY4/PkRGG/W1mEbqIKlRTAa0C8eOXhuk1bRcagP47kdt2Es+8xk1MJYwqS+dQ7juIM8UX/UZfeD+P1opa63t8V8FHLej7A+eB0BR/G6R83aBfa3/VFvEPIa5ks4OF/Jp3BbBWpd5Fw50X/LAHuu/xSwZw3DdWO+z8F1hvzWol87f69oRvasxnwFxmG4b0F/bK/J54xgzwnHvipzfjzmidEkKu2tP4LHOjse26tdTBHYwyJRf1D7AuYn0gxAnc2g6NNgl1sUPcKFNzpBmWkvXvel5bgs/b3Povalqh+WW+cWgC0vRX9LmkubSGmUPqnG+3nFgvFgP7n4C7AXlWB/cZRXgB3UBvUNzxYreoZ9A3DtncnhE8C2hxjBVkvLvZ9eidqIXmmoR/HJwfskpV6xVxZgDpjK9/F+fbcCBWWGSyheK/sD5jux+KEjj3y9FezubytryUTtxL401NQH7Px1+Ftwd9YwsKOexfKqbvhQ1XTE38ma7sqz6l9H8bvBAagHPPkirtVElEI3Ar/5YBiGYRjGrfDgg2EYhmEYt9Lqwi5CXH5tZq9TXjHbHPgayFGHr+3tZuUyGmrw1ZfNhvs66vE1m83qkMqxPnZp+W9bLb6mt1uUcke9FC+QXlc6pDCKfG4hvTJ0SCEA+dxmvfIa32Y3N7uvaz2JiAgjAGSXQkKOBmm6q3Q8R71yPNnnjgapfSzSsRu1n1QupdO3W9APrn3D0YBhFZsNL8whpeN2NDRfF4dZOpcZ/VBb7ZJO39x835D94KjHusnHrqpG296obsq1NNRI11kv1cUqXYfUBmbp+3L/kO8buB+tV/mukO5XyQ9yf3E99uXjS23qcryr7WtvwEeatRbrUluN/aHKhvec3Kbkcj57o/bEY8ntbxNyX5SPjbZFahOrVam7zSrvi9clt0GddJ2yHxrVrb7p9nbU43OrXn7GSnWTyxv1Tal/2OXnmosfqg3Sc8ghPdfk5548bV9+nhukMKvU3q79o5HPpHo28oN8z1ik+1kOu8v3u0td5Oe5XWov+dkg17WRX6Q2avR74VLuqJOu29D0s58Ir6Oq5vL2z7/jzaES17KXG7l48SJFRERcfUeGYRiGYVodubm5FB7efO6pVjf4cDgclJ+fT0IIioyMpNzcXPLzk1NeMU1RVVVFERER7LfrgH12Y7Dfrh/22Y3Bfrt+WsJnQgiqrq4mk8lEanXzqo5WF3ZRq9UUHh5OVVWXFbZ+fn7c2W4A9tv1wz67Mdhv1w/77MZgv10/7vaZv7//1XciFpwyDMMwDONmePDBMAzDMIxbabWDD4PBQIsWLSKDwXD1nRkn7Lfrh312Y7Dfrh/22Y3Bfrt+WrvPWp3glGEYhmGY3zat9s0HwzAMwzC/TXjwwTAMwzCMW+HBB8MwDMMwboUHHwzDMAzDuBUefDAMwzAM41Za7eDj3XffpejoaPLw8KDExERKS0tr6Sq1GpYsWUI33XQT+fr6UnBwMI0ZM4ZOnToF+zQ0NNDMmTMpMDCQfHx8aNy4cVRUVNRCNW59LF26lFQqFc2ZM8f5GfvsyuTl5dHkyZMpMDCQPD09qWfPnpSenu4sF0LQwoULKSwsjDw9PSk5OZmysrJasMYti91up+eee45iYmLI09OTYmNj6cUXX4TFtthnRDt37qRRo0aRyWQilUpF69evh/Jr8VF5eTlNmjSJ/Pz8yGg00iOPPEI1NTVuvAr305zfrFYrPfXUU9SzZ0/y9vYmk8lEU6ZMofz8fDhGq/CbaIWsWbNG6PV68eGHH4rjx4+LP/7xj8JoNIqioqKWrlqrYOjQoWLlypXi2LFjIiMjQ4wYMUJERkaKmpoa5z4zZswQERERIiUlRaSnp4ubb75Z9O/fvwVr3XpIS0sT0dHRolevXmL27NnOz9lnjSkvLxdRUVHioYceEvv27RNnz54VGzduFNnZ2c59li5dKvz9/cX69evF4cOHxd133y1iYmJEfX19C9a85Xj55ZdFYGCg+O6770ROTo5Yt26d8PHxEW+++aZzH/aZED/88IN45plnxJdffimISHz11VdQfi0+GjZsmOjdu7fYu3ev2LVrl+jQoYOYOHGim6/EvTTnt4qKCpGcnCw+//xzkZmZKVJTU0VCQoKIi4uDY7QGv7XKwUdCQoKYOXOm07bb7cJkMoklS5a0YK1aL8XFxYKIxI4dO4QQlzugTqcT69atc+5z8uRJQUQiNTW1parZKqiurhYdO3YUmzdvFrfffrtz8ME+uzJPPfWUGDBgQJPlDodDhIaGitdee835WUVFhTAYDOKzzz5zRxVbHSNHjhQPP/wwfDZ27FgxadIkIQT77ErIP6LX4qMTJ04IIhL79+937vPjjz8KlUol8vLy3Fb3luRKgzaZtLQ0QUTi/PnzQojW47dWF3axWCx04MABSk5Odn6mVqspOTmZUlNTW7BmrZfKykoiIgoICCAiogMHDpDVagUfdunShSIjI3/3Ppw5cyaNHDkSfEPEPmuKb775huLj42nChAkUHBxMffv2pffff99ZnpOTQ4WFheA3f39/SkxM/N36rX///pSSkkKnT58mIqLDhw/T7t27afjw4UTEPrsWrsVHqampZDQaKT4+3rlPcnIyqdVq2rdvn9vr3FqprKwklUpFRqORiFqP31rdqralpaVkt9spJCQEPg8JCaHMzMwWqlXrxeFw0Jw5c+iWW26hHj16EBFRYWEh6fV6Z2f7mZCQECosLGyBWrYO1qxZQwcPHqT9+/c3KmOfXZmzZ8/S8uXLae7cufT000/T/v376YknniC9Xk9Tp051+uZK9+vv1W/z58+nqqoq6tKlC2k0GrLb7fTyyy/TpEmTiIjYZ9fAtfiosLCQgoODoVyr1VJAQAD78ScaGhroqaeeookTJzpXtm0tfmt1gw/m+pg5cyYdO3aMdu/e3dJVadXk5ubS7NmzafPmzeTh4dHS1fmfweFwUHx8PC1evJiIiPr27UvHjh2jFStW0NSpU1u4dq2TtWvX0ieffEKffvopde/enTIyMmjOnDlkMpnYZ4zbsFqtdO+995IQgpYvX97S1WlEqwu7BAUFkUajaTTLoKioiEJDQ1uoVq2TWbNm0XfffUfbtm2j8PBw5+ehoaFksViooqIC9v89+/DAgQNUXFxM/fr1I61WS1qtlnbs2EFvvfUWabVaCgkJYZ9dgbCwMOrWrRt81rVrV7pw4QIRkdM3fL8qPPnkkzR//ny6//77qWfPnvTggw/Sn/70J1qyZAkRsc+uhWvxUWhoKBUXF0O5zWaj8vLy370ffx54nD9/njZv3ux860HUevzW6gYfer2e4uLiKCUlxfmZw+GglJQUSkpKasGatR6EEDRr1iz66quvaOvWrRQTEwPlcXFxpNPpwIenTp2iCxcu/G59OGjQIDp69ChlZGQ4/+Lj42nSpEnObfZZY2655ZZG07hPnz5NUVFRREQUExNDoaGh4Leqqirat2/f79ZvdXV1pFbjo1Wj0ZDD4SAi9tm1cC0+SkpKooqKCjpw4IBzn61bt5LD4aDExES317m18PPAIysri7Zs2UKBgYFQ3mr85jZp63WwZs0aYTAYxKpVq8SJEyfE9OnThdFoFIWFhS1dtVbBo48+Kvz9/cX27dtFQUGB86+urs65z4wZM0RkZKTYunWrSE9PF0lJSSIpKakFa936cJ3tIgT77EqkpaUJrVYrXn75ZZGVlSU++eQT4eXlJVavXu3cZ+nSpcJoNIqvv/5aHDlyRIwePfp3N23UlalTp4p27do5p9p++eWXIigoSMybN8+5D/vs8syzQ4cOiUOHDgkiEm+88YY4dOiQc1bGtfho2LBhom/fvmLfvn1i9+7domPHjr/5qbbN+c1isYi7775bhIeHi4yMDPh9MJvNzmO0Br+1ysGHEEK8/fbbIjIyUuj1epGQkCD27t3b0lVqNRDRFf9Wrlzp3Ke+vl489thjok2bNsLLy0vcc889oqCgoOUq3QqRBx/ssyvz7bffih49egiDwSC6dOki/vnPf0K5w+EQzz33nAgJCREGg0EMGjRInDp1qoVq2/JUVVWJ2bNni8jISOHh4SHat28vnnnmGXj4s8+E2LZt2xWfY1OnThVCXJuPysrKxMSJE4WPj4/w8/MT06ZNE9XV1S1wNe6jOb/l5OQ0+fuwbds25zFag99UQrik3WMYhmEYhvmVaXWaD4ZhGIZhftvw4INhGIZhGLfCgw+GYRiGYdwKDz4YhmEYhnErPPhgGIZhGMat8OCDYRiGYRi3woMPhmEYhmHcCg8+GIZhGIZxKzz4YBiGYRjGrfDgg2EYhmEYt8KDD4ZhGIZh3Mr/A/Dm2VyAXF9xAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "plt.imshow(output.detach().numpy())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1ff72e06-dff9-43d7-a6e5-18265be5b104",
   "metadata": {},
   "source": [
    "we can see that some features are constant across the samples while others change"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4c8967fb-f1c5-45a8-a7d0-abfaf4435e85",
   "metadata": {},
   "source": [
    "# use with OpenSoundscape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 191,
   "id": "4f8e7e65-1b1f-44cf-9773-813f1dcede6c",
   "metadata": {},
   "outputs": [],
   "source": [
    "f=filename"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8d17c216-812a-47b5-9a3b-e49c5f9d9b02",
   "metadata": {},
   "source": [
    "we'll turn of the internal preprocessing and post-processing (which applies PCA and whitening)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "id": "978f5ee8-8f61-4d06-ad50-ce4ddbbdfb22",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using cache found in /Users/SML161/.cache/torch/hub/harritaylor_torchvggish_master\n"
     ]
    }
   ],
   "source": [
    "vgg = torch.hub.load('harritaylor/torchvggish', 'vggish')\n",
    "vgg.preprocess = False\n",
    "vgg.postprocess = False "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "868ed716-7ec2-4e19-977f-bd9c64b89b49",
   "metadata": {},
   "source": [
    "the internal preprocessor is creating MelSpectrogram samples (log valued) with shape [1,96,64]:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "id": "edca0e9b-07ee-42aa-a718-d0bc6f067f39",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([19, 1, 96, 64])"
      ]
     },
     "execution_count": 197,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vgg._preprocess(f,fs=None).shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8d89c3b9-4d5a-4849-98b3-b6ccbde7b049",
   "metadata": {},
   "source": [
    "now, let's create an OpenSoundscape CNN object using the vgg as the architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "id": "beb6ffdd-3943-451a-a3fd-2cb0f30d4dca",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from opensoundscape.torch.models.cnn import CNN"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ac2a51a5-7ec0-4989-aeb9-3a2b668ded13",
   "metadata": {},
   "source": [
    "note that we make classes a list of length 128, so that the output of the network matches the expected output size of our model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "id": "7bab6cb6-c79e-467b-b7c2-2c00d23fdc46",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/SML161/opensoundscape/opensoundscape/torch/models/cnn.py:105: UserWarning: Make sure your architecture expects the number of channels in your input samples (1). Pytorch architectures expect 3 channels by default.\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "m = CNN(architecture=vgg,classes=np.arange(128),sample_duration=1,sample_shape=[96, 64,1])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c10cc544-c970-4a2f-82f8-08c82750b96e",
   "metadata": {},
   "source": [
    "change the preprocessing behavior to use MelSpectroram (quasi-logarithmic frequencies) instead of Spectrogram (linear frequency)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "id": "0a2a0669-20ba-4d4e-aae0-3a0d8e7bdc8b",
   "metadata": {},
   "outputs": [],
   "source": [
    "from opensoundscape.spectrogram import MelSpectrogram\n",
    "from opensoundscape.preprocess.actions import Action\n",
    "m.preprocessor.pipeline.to_spec = Action(MelSpectrogram.from_audio)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b635ad7c-3fbc-43de-a1c7-12a946601ef7",
   "metadata": {},
   "source": [
    "check that we preprocess samples of the correct shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "id": "eb998ccc-0726-4846-b49d-c396b99bc496",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "torch.Size([1, 96, 64])"
      ]
     },
     "execution_count": 201,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "s,info = m.preprocessor.forward(f)\n",
    "s.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1c07673c-2681-493d-8087-28d1eb72bb22",
   "metadata": {},
   "source": [
    "and check that we can pass those samples to the vgg architecture and get the feature vector out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "id": "b6fa9520-998b-4017-adcb-decdb4dd657e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([1, 128])"
      ]
     },
     "execution_count": 202,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vgg_feat = m.network.forward(s.unsqueeze(0))\n",
    "vgg_feat.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5503eb88-a27d-4b4d-8dfc-45a65ae9e901",
   "metadata": {},
   "source": [
    "Now, use OpenSoundscape's mechanics to generate features on a list of files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "id": "11403c59-989f-4a6e-a9ff-abd60b2bcc01",
   "metadata": {
    "collapsed": true,
    "jupyter": {
     "outputs_hidden": true
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n"
     ]
    }
   ],
   "source": [
    "preds, _, _ = m.predict([f])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "id": "e47ff426-3b96-4cb9-a2f2-90cd98cf9681",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>...</th>\n",
       "      <th>118</th>\n",
       "      <th>119</th>\n",
       "      <th>120</th>\n",
       "      <th>121</th>\n",
       "      <th>122</th>\n",
       "      <th>123</th>\n",
       "      <th>124</th>\n",
       "      <th>125</th>\n",
       "      <th>126</th>\n",
       "      <th>127</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>file</th>\n",
       "      <th>start_time</th>\n",
       "      <th>end_time</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"18\" valign=\"top\">bus_chatter.wav</th>\n",
       "      <th>0.0</th>\n",
       "      <th>1.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.278471</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.151628</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.124921</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.079170</td>\n",
       "      <td>0.088153</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.283686</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1.0</th>\n",
       "      <th>2.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.257660</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.167520</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.062247</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.038897</td>\n",
       "      <td>0.008585</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.385484</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2.0</th>\n",
       "      <th>3.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.277238</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.173320</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.088737</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.381939</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3.0</th>\n",
       "      <th>4.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.235431</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.187473</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.055323</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.084498</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.390752</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4.0</th>\n",
       "      <th>5.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.252497</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.190735</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.030830</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.094437</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.338956</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5.0</th>\n",
       "      <th>6.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.231514</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.189389</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.032201</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.089037</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.383347</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6.0</th>\n",
       "      <th>7.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.293826</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.150943</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.031819</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.409308</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7.0</th>\n",
       "      <th>8.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.196446</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.212592</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.126165</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.124310</td>\n",
       "      <td>0.087042</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.333099</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8.0</th>\n",
       "      <th>9.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.236335</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.181137</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.080504</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.366061</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9.0</th>\n",
       "      <th>10.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.222815</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.176373</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.018496</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.091031</td>\n",
       "      <td>0.040314</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.344080</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10.0</th>\n",
       "      <th>11.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.242644</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.178091</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.084123</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.067649</td>\n",
       "      <td>0.048231</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.378086</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11.0</th>\n",
       "      <th>12.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.248261</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.159815</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.016548</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.062290</td>\n",
       "      <td>0.001986</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.356162</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12.0</th>\n",
       "      <th>13.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.235566</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.224073</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.067926</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.110380</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.372896</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13.0</th>\n",
       "      <th>14.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.217955</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.202602</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.050149</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.151452</td>\n",
       "      <td>0.018572</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.370058</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14.0</th>\n",
       "      <th>15.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.248950</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.156188</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.030757</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.100040</td>\n",
       "      <td>0.018022</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.395959</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15.0</th>\n",
       "      <th>16.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.306571</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.154496</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.047088</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.368833</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16.0</th>\n",
       "      <th>17.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.279230</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.175175</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000917</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.035308</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.320643</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17.0</th>\n",
       "      <th>18.0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.163487</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.244820</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.046318</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.096014</td>\n",
       "      <td>0.109004</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.314059</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>18 rows × 128 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                     0    1         2    3         4    5    \\\n",
       "file            start_time end_time                                           \n",
       "bus_chatter.wav 0.0        1.0       0.0  0.0  0.278471  0.0  0.151628  0.0   \n",
       "                1.0        2.0       0.0  0.0  0.257660  0.0  0.167520  0.0   \n",
       "                2.0        3.0       0.0  0.0  0.277238  0.0  0.173320  0.0   \n",
       "                3.0        4.0       0.0  0.0  0.235431  0.0  0.187473  0.0   \n",
       "                4.0        5.0       0.0  0.0  0.252497  0.0  0.190735  0.0   \n",
       "                5.0        6.0       0.0  0.0  0.231514  0.0  0.189389  0.0   \n",
       "                6.0        7.0       0.0  0.0  0.293826  0.0  0.150943  0.0   \n",
       "                7.0        8.0       0.0  0.0  0.196446  0.0  0.212592  0.0   \n",
       "                8.0        9.0       0.0  0.0  0.236335  0.0  0.181137  0.0   \n",
       "                9.0        10.0      0.0  0.0  0.222815  0.0  0.176373  0.0   \n",
       "                10.0       11.0      0.0  0.0  0.242644  0.0  0.178091  0.0   \n",
       "                11.0       12.0      0.0  0.0  0.248261  0.0  0.159815  0.0   \n",
       "                12.0       13.0      0.0  0.0  0.235566  0.0  0.224073  0.0   \n",
       "                13.0       14.0      0.0  0.0  0.217955  0.0  0.202602  0.0   \n",
       "                14.0       15.0      0.0  0.0  0.248950  0.0  0.156188  0.0   \n",
       "                15.0       16.0      0.0  0.0  0.306571  0.0  0.154496  0.0   \n",
       "                16.0       17.0      0.0  0.0  0.279230  0.0  0.175175  0.0   \n",
       "                17.0       18.0      0.0  0.0  0.163487  0.0  0.244820  0.0   \n",
       "\n",
       "                                     6    7    8    9    ...  118       119  \\\n",
       "file            start_time end_time                      ...                  \n",
       "bus_chatter.wav 0.0        1.0       0.0  0.0  0.0  0.0  ...  0.0  0.124921   \n",
       "                1.0        2.0       0.0  0.0  0.0  0.0  ...  0.0  0.062247   \n",
       "                2.0        3.0       0.0  0.0  0.0  0.0  ...  0.0  0.000000   \n",
       "                3.0        4.0       0.0  0.0  0.0  0.0  ...  0.0  0.055323   \n",
       "                4.0        5.0       0.0  0.0  0.0  0.0  ...  0.0  0.030830   \n",
       "                5.0        6.0       0.0  0.0  0.0  0.0  ...  0.0  0.032201   \n",
       "                6.0        7.0       0.0  0.0  0.0  0.0  ...  0.0  0.000000   \n",
       "                7.0        8.0       0.0  0.0  0.0  0.0  ...  0.0  0.126165   \n",
       "                8.0        9.0       0.0  0.0  0.0  0.0  ...  0.0  0.000000   \n",
       "                9.0        10.0      0.0  0.0  0.0  0.0  ...  0.0  0.018496   \n",
       "                10.0       11.0      0.0  0.0  0.0  0.0  ...  0.0  0.084123   \n",
       "                11.0       12.0      0.0  0.0  0.0  0.0  ...  0.0  0.016548   \n",
       "                12.0       13.0      0.0  0.0  0.0  0.0  ...  0.0  0.067926   \n",
       "                13.0       14.0      0.0  0.0  0.0  0.0  ...  0.0  0.050149   \n",
       "                14.0       15.0      0.0  0.0  0.0  0.0  ...  0.0  0.030757   \n",
       "                15.0       16.0      0.0  0.0  0.0  0.0  ...  0.0  0.047088   \n",
       "                16.0       17.0      0.0  0.0  0.0  0.0  ...  0.0  0.000917   \n",
       "                17.0       18.0      0.0  0.0  0.0  0.0  ...  0.0  0.046318   \n",
       "\n",
       "                                     120  121  122       123       124  125  \\\n",
       "file            start_time end_time                                           \n",
       "bus_chatter.wav 0.0        1.0       0.0  0.0  0.0  0.079170  0.088153  0.0   \n",
       "                1.0        2.0       0.0  0.0  0.0  0.038897  0.008585  0.0   \n",
       "                2.0        3.0       0.0  0.0  0.0  0.088737  0.000000  0.0   \n",
       "                3.0        4.0       0.0  0.0  0.0  0.084498  0.000000  0.0   \n",
       "                4.0        5.0       0.0  0.0  0.0  0.094437  0.000000  0.0   \n",
       "                5.0        6.0       0.0  0.0  0.0  0.089037  0.000000  0.0   \n",
       "                6.0        7.0       0.0  0.0  0.0  0.031819  0.000000  0.0   \n",
       "                7.0        8.0       0.0  0.0  0.0  0.124310  0.087042  0.0   \n",
       "                8.0        9.0       0.0  0.0  0.0  0.080504  0.000000  0.0   \n",
       "                9.0        10.0      0.0  0.0  0.0  0.091031  0.040314  0.0   \n",
       "                10.0       11.0      0.0  0.0  0.0  0.067649  0.048231  0.0   \n",
       "                11.0       12.0      0.0  0.0  0.0  0.062290  0.001986  0.0   \n",
       "                12.0       13.0      0.0  0.0  0.0  0.110380  0.000000  0.0   \n",
       "                13.0       14.0      0.0  0.0  0.0  0.151452  0.018572  0.0   \n",
       "                14.0       15.0      0.0  0.0  0.0  0.100040  0.018022  0.0   \n",
       "                15.0       16.0      0.0  0.0  0.0  0.000000  0.000000  0.0   \n",
       "                16.0       17.0      0.0  0.0  0.0  0.035308  0.000000  0.0   \n",
       "                17.0       18.0      0.0  0.0  0.0  0.096014  0.109004  0.0   \n",
       "\n",
       "                                          126  127  \n",
       "file            start_time end_time                 \n",
       "bus_chatter.wav 0.0        1.0       0.283686  0.0  \n",
       "                1.0        2.0       0.385484  0.0  \n",
       "                2.0        3.0       0.381939  0.0  \n",
       "                3.0        4.0       0.390752  0.0  \n",
       "                4.0        5.0       0.338956  0.0  \n",
       "                5.0        6.0       0.383347  0.0  \n",
       "                6.0        7.0       0.409308  0.0  \n",
       "                7.0        8.0       0.333099  0.0  \n",
       "                8.0        9.0       0.366061  0.0  \n",
       "                9.0        10.0      0.344080  0.0  \n",
       "                10.0       11.0      0.378086  0.0  \n",
       "                11.0       12.0      0.356162  0.0  \n",
       "                12.0       13.0      0.372896  0.0  \n",
       "                13.0       14.0      0.370058  0.0  \n",
       "                14.0       15.0      0.395959  0.0  \n",
       "                15.0       16.0      0.368833  0.0  \n",
       "                16.0       17.0      0.320643  0.0  \n",
       "                17.0       18.0      0.314059  0.0  \n",
       "\n",
       "[18 rows x 128 columns]"
      ]
     },
     "execution_count": 204,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preds"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9ded6bc0-56d1-404b-80b3-7774662edb57",
   "metadata": {},
   "source": [
    "# add classification layer after vgg feature extractor \n",
    "we can add other layers after the VGGish feature extractor and train them on a desired number of classes. The simplest approach is to add a fully connected layer mapping 128 inputs to [the number of classes] outputs. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "id": "514fa494-b69e-479a-a691-7b3f4ec65310",
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch import nn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "id": "f4339074-0503-4ed8-ba8e-63133d268350",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/SML161/opensoundscape/opensoundscape/torch/models/cnn.py:105: UserWarning: Make sure your architecture expects the number of channels in your input samples (1). Pytorch architectures expect 3 channels by default.\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "m = CNN(architecture=vgg,classes=np.arange(128),sample_duration=1,sample_shape=[96, 64,1])\n",
    "m.preprocessor.pipeline.to_spec = Action(MelSpectrogram.from_audio)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "384a24c1-60dc-40bd-a948-0ae932a8d53f",
   "metadata": {},
   "source": [
    "### freeze VGGish weights\n",
    "If we were to train this model, we would probably want to freeze vggish and only train the layers we are adding on. Let's freeze all parameters currently in the architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 205,
   "id": "820a7a03-5a76-4098-88b7-002432111733",
   "metadata": {},
   "outputs": [],
   "source": [
    "for param in m.network.parameters():\n",
    "    param.requires_grad=False"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8aa9816f-c46b-4cfd-a7b0-e421e718c0c4",
   "metadata": {},
   "source": [
    "### add fully connected layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 206,
   "id": "90ef4700-31d0-4706-b189-849773b5e256",
   "metadata": {},
   "outputs": [],
   "source": [
    "m.classes = ['frog','notfrog'] #set of classes we want to recognize\n",
    "fc = nn.Linear(in_features=128,out_features=len(m.classes))\n",
    "m.network.embeddings.add_module('fc',fc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 208,
   "id": "3f92197f-a860-4b64-afb4-9e561d55c531",
   "metadata": {
    "collapsed": true,
    "jupyter": {
     "outputs_hidden": true
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n",
      "/Users/SML161/opensoundscape/opensoundscape/audio.py:188: UserWarning: Failed to load metadata: 'artist'. Metadata will be None\n",
      "  warnings.warn(f\"Failed to load metadata: {e}. Metadata will be None\")\n",
      "/Users/SML161/opensoundscape/opensoundscape/spectrogram.py:702: FutureWarning: Pass sr=44100, n_fft=512 as keyword args. From version 0.10 passing these as positional arguments will result in an error\n",
      "  fb = librosa.filters.mel(\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>frog</th>\n",
       "      <th>notfrog</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>file</th>\n",
       "      <th>start_time</th>\n",
       "      <th>end_time</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"18\" valign=\"top\">bus_chatter.wav</th>\n",
       "      <th>0.0</th>\n",
       "      <th>1.0</th>\n",
       "      <td>0.037825</td>\n",
       "      <td>-0.044471</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1.0</th>\n",
       "      <th>2.0</th>\n",
       "      <td>0.048731</td>\n",
       "      <td>-0.025217</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2.0</th>\n",
       "      <th>3.0</th>\n",
       "      <td>0.029733</td>\n",
       "      <td>-0.019102</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3.0</th>\n",
       "      <th>4.0</th>\n",
       "      <td>0.058458</td>\n",
       "      <td>-0.038375</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4.0</th>\n",
       "      <th>5.0</th>\n",
       "      <td>0.038907</td>\n",
       "      <td>-0.020259</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5.0</th>\n",
       "      <th>6.0</th>\n",
       "      <td>0.035583</td>\n",
       "      <td>-0.032007</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6.0</th>\n",
       "      <th>7.0</th>\n",
       "      <td>0.036369</td>\n",
       "      <td>-0.012072</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7.0</th>\n",
       "      <th>8.0</th>\n",
       "      <td>0.061777</td>\n",
       "      <td>-0.037487</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8.0</th>\n",
       "      <th>9.0</th>\n",
       "      <td>0.038710</td>\n",
       "      <td>-0.015797</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9.0</th>\n",
       "      <th>10.0</th>\n",
       "      <td>0.039489</td>\n",
       "      <td>-0.018754</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10.0</th>\n",
       "      <th>11.0</th>\n",
       "      <td>0.047219</td>\n",
       "      <td>-0.022538</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11.0</th>\n",
       "      <th>12.0</th>\n",
       "      <td>0.041551</td>\n",
       "      <td>-0.025012</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12.0</th>\n",
       "      <th>13.0</th>\n",
       "      <td>0.050307</td>\n",
       "      <td>-0.034695</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13.0</th>\n",
       "      <th>14.0</th>\n",
       "      <td>0.053348</td>\n",
       "      <td>-0.026717</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14.0</th>\n",
       "      <th>15.0</th>\n",
       "      <td>0.051504</td>\n",
       "      <td>-0.029674</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15.0</th>\n",
       "      <th>16.0</th>\n",
       "      <td>0.041777</td>\n",
       "      <td>-0.022258</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16.0</th>\n",
       "      <th>17.0</th>\n",
       "      <td>0.043781</td>\n",
       "      <td>-0.043807</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17.0</th>\n",
       "      <th>18.0</th>\n",
       "      <td>0.060804</td>\n",
       "      <td>-0.041682</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                         frog   notfrog\n",
       "file            start_time end_time                    \n",
       "bus_chatter.wav 0.0        1.0       0.037825 -0.044471\n",
       "                1.0        2.0       0.048731 -0.025217\n",
       "                2.0        3.0       0.029733 -0.019102\n",
       "                3.0        4.0       0.058458 -0.038375\n",
       "                4.0        5.0       0.038907 -0.020259\n",
       "                5.0        6.0       0.035583 -0.032007\n",
       "                6.0        7.0       0.036369 -0.012072\n",
       "                7.0        8.0       0.061777 -0.037487\n",
       "                8.0        9.0       0.038710 -0.015797\n",
       "                9.0        10.0      0.039489 -0.018754\n",
       "                10.0       11.0      0.047219 -0.022538\n",
       "                11.0       12.0      0.041551 -0.025012\n",
       "                12.0       13.0      0.050307 -0.034695\n",
       "                13.0       14.0      0.053348 -0.026717\n",
       "                14.0       15.0      0.051504 -0.029674\n",
       "                15.0       16.0      0.041777 -0.022258\n",
       "                16.0       17.0      0.043781 -0.043807\n",
       "                17.0       18.0      0.060804 -0.041682"
      ]
     },
     "execution_count": 208,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preds, _, _ = m.predict([f])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 209,
   "id": "b91a0804-ca96-45f2-80c9-88f48a8a41fd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>frog</th>\n",
       "      <th>notfrog</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>file</th>\n",
       "      <th>start_time</th>\n",
       "      <th>end_time</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"18\" valign=\"top\">bus_chatter.wav</th>\n",
       "      <th>0.0</th>\n",
       "      <th>1.0</th>\n",
       "      <td>0.037825</td>\n",
       "      <td>-0.044471</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1.0</th>\n",
       "      <th>2.0</th>\n",
       "      <td>0.048731</td>\n",
       "      <td>-0.025217</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2.0</th>\n",
       "      <th>3.0</th>\n",
       "      <td>0.029733</td>\n",
       "      <td>-0.019102</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3.0</th>\n",
       "      <th>4.0</th>\n",
       "      <td>0.058458</td>\n",
       "      <td>-0.038375</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4.0</th>\n",
       "      <th>5.0</th>\n",
       "      <td>0.038907</td>\n",
       "      <td>-0.020259</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5.0</th>\n",
       "      <th>6.0</th>\n",
       "      <td>0.035583</td>\n",
       "      <td>-0.032007</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6.0</th>\n",
       "      <th>7.0</th>\n",
       "      <td>0.036369</td>\n",
       "      <td>-0.012072</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7.0</th>\n",
       "      <th>8.0</th>\n",
       "      <td>0.061777</td>\n",
       "      <td>-0.037487</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8.0</th>\n",
       "      <th>9.0</th>\n",
       "      <td>0.038710</td>\n",
       "      <td>-0.015797</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9.0</th>\n",
       "      <th>10.0</th>\n",
       "      <td>0.039489</td>\n",
       "      <td>-0.018754</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10.0</th>\n",
       "      <th>11.0</th>\n",
       "      <td>0.047219</td>\n",
       "      <td>-0.022538</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11.0</th>\n",
       "      <th>12.0</th>\n",
       "      <td>0.041551</td>\n",
       "      <td>-0.025012</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12.0</th>\n",
       "      <th>13.0</th>\n",
       "      <td>0.050307</td>\n",
       "      <td>-0.034695</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13.0</th>\n",
       "      <th>14.0</th>\n",
       "      <td>0.053348</td>\n",
       "      <td>-0.026717</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14.0</th>\n",
       "      <th>15.0</th>\n",
       "      <td>0.051504</td>\n",
       "      <td>-0.029674</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15.0</th>\n",
       "      <th>16.0</th>\n",
       "      <td>0.041777</td>\n",
       "      <td>-0.022258</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16.0</th>\n",
       "      <th>17.0</th>\n",
       "      <td>0.043781</td>\n",
       "      <td>-0.043807</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17.0</th>\n",
       "      <th>18.0</th>\n",
       "      <td>0.060804</td>\n",
       "      <td>-0.041682</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                         frog   notfrog\n",
       "file            start_time end_time                    \n",
       "bus_chatter.wav 0.0        1.0       0.037825 -0.044471\n",
       "                1.0        2.0       0.048731 -0.025217\n",
       "                2.0        3.0       0.029733 -0.019102\n",
       "                3.0        4.0       0.058458 -0.038375\n",
       "                4.0        5.0       0.038907 -0.020259\n",
       "                5.0        6.0       0.035583 -0.032007\n",
       "                6.0        7.0       0.036369 -0.012072\n",
       "                7.0        8.0       0.061777 -0.037487\n",
       "                8.0        9.0       0.038710 -0.015797\n",
       "                9.0        10.0      0.039489 -0.018754\n",
       "                10.0       11.0      0.047219 -0.022538\n",
       "                11.0       12.0      0.041551 -0.025012\n",
       "                12.0       13.0      0.050307 -0.034695\n",
       "                13.0       14.0      0.053348 -0.026717\n",
       "                14.0       15.0      0.051504 -0.029674\n",
       "                15.0       16.0      0.041777 -0.022258\n",
       "                16.0       17.0      0.043781 -0.043807\n",
       "                17.0       18.0      0.060804 -0.041682"
      ]
     },
     "execution_count": 209,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preds"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "opso_dev",
   "language": "python",
   "name": "opso_dev"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}