Convolutional Autoencoder.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch import nn\n",
    "import numpy as np\n",
    "from torchvision import datasets\n",
    "import torchvision.transforms as transforms\n",
    "import matplotlib.pyplot as plt\n",
    "import torch.nn.functional as F"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# convert data to torch.FloatTensor\n",
    "transform = transforms.Compose([transforms.ToTensor()])\n",
    "\n",
    "# load the training and test datasets\n",
    "train_data = datasets.MNIST(root='data', train=True,\n",
    "                                   download=True, transform=transform)\n",
    "test_data = datasets.MNIST(root='data', train=False,\n",
    "                                  download=True, transform=transform)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# how many samples per batch to load\n",
    "batch_size = 32\n",
    "\n",
    "# prepare data loaders\n",
    "train_loader = torch.utils.data.DataLoader(train_data, batch_size=batch_size)\n",
    "test_loader = torch.utils.data.DataLoader(test_data, batch_size=batch_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# obtain one batch of test images\n",
    "dataiter = iter(test_loader)\n",
    "images, labels = dataiter.next()\n",
    "\n",
    "images_flatten = images.view(images.size(0), -1)\n",
    "# get sample outputs\n",
    "output = model(images_flatten)\n",
    "\n",
    "images = images.numpy()\n",
    "output = output.view(batch_size, 1, 28, 28)\n",
    "output = output.detach().numpy()\n",
    "\n",
    "# plot the first ten input images and then reconstructed images\n",
    "fig, axes = plt.subplots(nrows=2, ncols=10, sharex=True, sharey=True, figsize=(25,4))\n",
    "\n",
    "# input images on top row, reconstructions on bottom\n",
    "for images, row in zip([images, output], axes):\n",
    "    for img, ax in zip(images, row):\n",
    "        ax.imshow(np.squeeze(img), cmap='gray')\n",
    "        ax.get_xaxis().set_visible(False)\n",
    "        ax.get_yaxis().set_visible(False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training on GPU\n"
     ]
    }
   ],
   "source": [
    "train_on_gpu = torch.cuda.is_available()\n",
    "\n",
    "if train_on_gpu:\n",
    "    print(\"Training on GPU\")\n",
    "else:\n",
    "    print(\"Training on CPU\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train(model, train_loader, lr=0.001, n_epochs=3, noise=False, train_on_gpu=False):\n",
    "    \"\"\"\n",
    "    Train given model based on parameters given\n",
    "    :param model: autoencoder model\n",
    "    :param criterion: loss function\n",
    "    :param optimizer: optimizer  function\n",
    "    :param train_loader: dataloader\n",
    "    :param lr: learing rate\n",
    "    :return: max-pooled feature vector\n",
    "    \"\"\"\n",
    "#     optimizer.param_groups[0]['lr']\n",
    "    criterion = nn.MSELoss()\n",
    "    optimizer = torch.optim.Adam(model.parameters(), lr=lr )\n",
    "    \n",
    "    for epoch in range(1, n_epochs+1):\n",
    "\n",
    "        train_loss = 0.0\n",
    "        for data in train_loader:\n",
    "            images, labels = data\n",
    "            if noise:\n",
    "                images = images + noise_factor * torch.randn(*images.shape)\n",
    "                images = np.clip(images, 0., 1.)\n",
    "            \n",
    "            if train_on_gpu:\n",
    "                images = images.cuda()\n",
    "            optimizer.zero_grad()\n",
    "            outputs = model(images)\n",
    "            loss = criterion(outputs, images)\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            # update running training loss\n",
    "            train_loss += loss.item()*images.size(0)\n",
    "\n",
    "        # print avg training statistics \n",
    "        train_loss = train_loss/len(train_loader)\n",
    "        print('Epoch: {} \\tTraining Loss: {:.6f}'.format(\n",
    "            epoch, \n",
    "            train_loss\n",
    "            ))\n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Helper function\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def save(model, model_name):\n",
    "    torch.save(model.state_dict(),f\"brain\\{model_name}.pt\")\n",
    "    \n",
    "def load_model(model, path):\n",
    "    return model.load_state_dict(torch.load(path))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Convolution Auroencoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ConvAutoencoder(\n",
      "  (conv1): Conv2d(1, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
      "  (conv2): Conv2d(32, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "  (pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
      "  (t_conv1): ConvTranspose2d(8, 16, kernel_size=(2, 2), stride=(2, 2))\n",
      "  (t_conv2): ConvTranspose2d(16, 1, kernel_size=(2, 2), stride=(2, 2))\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "class ConvAutoencoder(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(ConvAutoencoder, self).__init__()\n",
    "        ## encoder layers ##\n",
    "        # conv layer (depth from 1 --> 32), 5x5 kernels\n",
    "        self.conv1 = nn.Conv2d(1, 32, 5, padding=2)  \n",
    "        # conv layer (depth from 32 --> 8), 3x3 kernels\n",
    "        self.conv2 = nn.Conv2d(32, 8, 3, padding=1)\n",
    "        # pooling layer to reduce x-y dims by two; kernel and stride of 2\n",
    "        self.pool = nn.MaxPool2d(2, 2)\n",
    "        \n",
    "        ## decoder layers ##\n",
    "        ## a kernel of 2 and a stride of 2 will increase the spatial dims by 2\n",
    "        self.t_conv1 = nn.ConvTranspose2d(8, 16, 2, stride=2)\n",
    "        self.t_conv2 = nn.ConvTranspose2d(16, 1, 2, stride=2)\n",
    "\n",
    "\n",
    "    def forward(self, x):\n",
    "        ## encode ##\n",
    "        x = F.relu(self.conv1(x))\n",
    "        x = self.pool(x)\n",
    "        x = F.relu(self.conv2(x))\n",
    "        x = self.pool(x)  # compressed representation\n",
    "        \n",
    "        ## decode ##\n",
    "        x = F.relu(self.t_conv1(x))\n",
    "        # output layer (with sigmoid for scaling from 0 to 1)\n",
    "        x = F.sigmoid(self.t_conv2(x))\n",
    "                \n",
    "        return x\n",
    "\n",
    "# initialize the NN\n",
    "model = ConvAutoencoder()\n",
    "print(model)\n",
    "\n",
    "if train_on_gpu:\n",
    "    model = model.cuda()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1 \tTraining Loss: 4.953319\n",
      "Epoch: 2 \tTraining Loss: 4.953319\n",
      "Epoch: 3 \tTraining Loss: 4.953319\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "ConvAutoencoder(\n",
       "  (conv1): Conv2d(1, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "  (conv2): Conv2d(32, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
       "  (pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "  (t_conv1): ConvTranspose2d(8, 16, kernel_size=(2, 2), stride=(2, 2))\n",
       "  (t_conv2): ConvTranspose2d(16, 1, kernel_size=(2, 2), stride=(2, 2))\n",
       ")"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train(model, criterion, optimizer, train_loader, train_on_gpu=train_on_gpu)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "scrolled": true,
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1800x288 with 20 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot(model, test_loader):\n",
    "    dataiter = iter(test_loader)\n",
    "    images, labels = dataiter.next()\n",
    "\n",
    "#     images_flatten = images.view(images.size(0), -1)\n",
    "    # get sample outputs\n",
    "    model.cpu()\n",
    "    output = model(images)\n",
    "\n",
    "    images = images.numpy()\n",
    "    output = output.view(batch_size, 1, 28, 28)\n",
    "    output = output.detach().numpy()\n",
    "\n",
    "    # plot the first ten input images and then reconstructed images\n",
    "    fig, axes = plt.subplots(nrows=2, ncols=10, sharex=True, sharey=True, figsize=(25,4))\n",
    "\n",
    "    # input images on top row, reconstructions on bottom\n",
    "    for images, row in zip([images, output], axes):\n",
    "        for img, ax in zip(images, row):\n",
    "            ax.imshow(np.squeeze(img), cmap='gray')\n",
    "            ax.get_xaxis().set_visible(False)\n",
    "            ax.get_yaxis().set_visible(False)\n",
    "plot(model, test_loader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "save(model,\"conv-autoencoder\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.6"
  },
  "pycharm": {
   "stem_cell": {
    "cell_type": "raw",
    "source": [],
    "metadata": {
     "collapsed": false
    }
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}