03_Dilated-Convolution.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import mxnet as mx\n",
    "from mxnet import gluon\n",
    "ctx = mx.gpu(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "attachments": {
    "image.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Dilated convolution for Conditional PixelCNN\n",
    "Autoregressive models based on convolutional layers are restricted to refer just nearby pixels when predicting a value of new pixel. To enlarge the area where context is taken into account, WaveNet \\[[arxiv:1609.03499](https://arxiv.org/pdf/1609.03499.pdf)\\] uses dilated convolution. As such, the network can model relatively long-scale context of the sequential data.\n",
    "\n",
    "![image.png](attachment:image.png)\n",
    "$$ \\mathrm{Dilated\\ convolution\\ of\\ WaveNet} $$\n",
    "\n",
    "With the dilation technique, the conditional PixelCNN is expected to be aware wider area and generate more consistent output."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "class VStack(gluon.nn.Conv2D):\n",
    "    '''Vertical stack of a PixelCNN layer\n",
    "    \n",
    "    Gluon's 2D-convolution layer modified as to capture the image context\n",
    "    descending from top line to target line.\n",
    "    When kernel_type is set to 'A'(default), the convolution does not rely on\n",
    "    the value of same position of previous layer, which preserves *causality*.\n",
    "    '''\n",
    "    def __init__(self, channels, kernel_size, strides=(1, 1),\n",
    "                 dilation=(1, 1), groups=1, layout='NCHW',\n",
    "                 activation=None, use_bias=True, weight_initializer=None,\n",
    "                 bias_initializer='zeros', in_channels=0, kernel_type='A', **kwargs):\n",
    "        pad_h = (kernel_size[0]-1)*dilation[0] + 1 * ('A' is kernel_type)\n",
    "        pad_w = (kernel_size[1]//2)*dilation[1]\n",
    "        super(VStack, self).__init__(channels, kernel_size, strides,\n",
    "                 (pad_h, pad_w), dilation, groups, layout,\n",
    "                 activation, use_bias, weight_initializer,\n",
    "                 bias_initializer, in_channels, **kwargs)\n",
    "\n",
    "    def hybrid_forward(self, F, x, weight, bias=None):\n",
    "        act = super(VStack, self).hybrid_forward(F, x, weight, bias)\n",
    "        return F.slice_like(act, x, axes=[2,3])\n",
    "\n",
    "class HStack(gluon.nn.Conv2D):\n",
    "    '''Horizontal stack of a PixelCNN layer\n",
    "    \n",
    "    Gluon's 2D-convolution layer modified as to capture the image context\n",
    "    propagating from left pixel to right pixel.\n",
    "    When kernel_type is set to 'A'(default), the convolution does not rely on\n",
    "    the value of same position of previous layer, which preserves *causality*.\n",
    "    '''\n",
    "    def __init__(self, channels, kernel_len, strides=(1, 1),\n",
    "                 dilation=(1,1), groups=1, layout='NCHW',\n",
    "                 activation=None, use_bias=True, weight_initializer=None,\n",
    "                 bias_initializer='zeros', in_channels=0, kernel_type='A', **kwargs):\n",
    "        pad_w = (kernel_len-1)*dilation[1] + 1 * ('A' is kernel_type)\n",
    "        super(HStack, self).__init__(channels, (1,kernel_len), strides,\n",
    "                 (1, pad_w), dilation, groups, layout,\n",
    "                 activation, use_bias, weight_initializer,\n",
    "                 bias_initializer, in_channels, **kwargs)\n",
    "\n",
    "    def hybrid_forward(self, F, x, weight, bias=None):\n",
    "        act = super(HStack, self).hybrid_forward(F, x, weight, bias)\n",
    "        return F.slice_like(act, x, axes=[2,3])\n",
    "    \n",
    "class GatedLayer(gluon.HybridBlock):\n",
    "    '''Implementation of a single layer in Gated PixelCNN (A. van den Oord, et.al., NIPS2016 6527)\n",
    "    '''\n",
    "    def __init__(self, channels, kernel_size, strides=(1, 1),\n",
    "                 dilation=(1, 1), groups=1, use_bias=True, weight_initializer=None,\n",
    "                 bias_initializer='zeros', in_channels=0, kernel_type='A', residual=True, **kwargs):\n",
    "        super(GatedLayer, self).__init__(**kwargs)\n",
    "        self.channels = channels\n",
    "        self.residual = residual\n",
    "        with self.name_scope():\n",
    "            self.vstack = VStack(channels*2, kernel_size, strides,\n",
    "                 (dilation[0], 1), groups, 'NCHW', None, use_bias, weight_initializer,\n",
    "                 bias_initializer, in_channels, kernel_type, **kwargs)\n",
    "            self.hstack = HStack(channels*2, kernel_size[1], strides,\n",
    "                 (1, dilation[1]), groups, 'NCHW', None, use_bias, weight_initializer,\n",
    "                 bias_initializer, in_channels, kernel_type, **kwargs)\n",
    "            self.vtoh = gluon.nn.Conv2D(channels*2, (1,1), strides,\n",
    "                 (0,0), (1,1), groups, 'NCHW', None, use_bias, weight_initializer,\n",
    "                 bias_initializer, channels*2, **kwargs)\n",
    "            self.htoh = gluon.nn.Conv2D(channels, (1,1), strides,\n",
    "                 (0,0), (1,1), groups, 'NCHW', None, use_bias, weight_initializer,\n",
    "                 bias_initializer, channels, **kwargs)\n",
    "            \n",
    "    def hybrid_forward(self, F, x):\n",
    "        vx, hx = F.split(x, axis=1, num_outputs=2)\n",
    "        vstack = self.vstack(vx)\n",
    "        hstack = self.hstack(hx) + self.vtoh(vstack)\n",
    "        \n",
    "        vstack_presig, vstack_pretanh = F.split(vstack, axis=1, num_outputs=2)\n",
    "        hstack_presig, hstack_pretanh = F.split(hstack, axis=1, num_outputs=2)\n",
    "        vstack_out = F.sigmoid(vstack_presig) * F.tanh(vstack_pretanh)\n",
    "        hstack_preres = self.htoh(F.sigmoid(hstack_presig) * F.tanh(hstack_pretanh))\n",
    "        # residual connection for hstack\n",
    "        if not self.residual:\n",
    "            hstack_out = hstack_preres\n",
    "        elif hx.shape[1] < self.channels:\n",
    "            zpad_shape = list(hx.shape)\n",
    "            zpad_shape[1] = self.channels - zpad_shape[1]\n",
    "            hstack_out = hstack_preres + F.concat(hx, F.zeros(zpad_shape), dim=1)\n",
    "        else:\n",
    "            hstack_out = hstack_preres + hx\n",
    "            \n",
    "        return mx.nd.concat(vstack_out, hstack_out, dim=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "class ConditionalLayer(GatedLayer):\n",
    "    def __init__(self, channels, kernel_size, strides=(1, 1),\n",
    "                 dilation=(1, 1), groups=1, use_bias=True, weight_initializer=None,\n",
    "                 bias_initializer='zeros', in_channels=0, kernel_type='A', residual=True, **kwargs):\n",
    "        super(ConditionalLayer, self).__init__(channels, kernel_size, strides,\n",
    "                 dilation, groups, use_bias, weight_initializer,\n",
    "                 bias_initializer, in_channels, kernel_type, residual, **kwargs)\n",
    "        with self.name_scope():\n",
    "            self.embedder = gluon.nn.Conv2D(channels*4, (1,1))\n",
    "            \n",
    "    def hybrid_forward(self, F, x, h):\n",
    "        for _ in range(2):\n",
    "            h = F.expand_dims(h, axis=-1)\n",
    "        vstack_presig_cond, vstack_pretanh_cond, hstack_presig_cond, hstack_pretanh_cond = \\\n",
    "            F.split(self.embedder(h), axis=1, num_outputs=4)\n",
    "        \n",
    "        vx, hx = F.split(x, axis=1, num_outputs=2)\n",
    "        vstack = self.vstack(vx)\n",
    "        hstack = self.hstack(hx) + self.vtoh(vstack)\n",
    "        \n",
    "        vstack_presig, vstack_pretanh = F.split(vstack, axis=1, num_outputs=2)\n",
    "        hstack_presig, hstack_pretanh = F.split(hstack, axis=1, num_outputs=2)\n",
    "        vstack_out = F.sigmoid(vstack_presig + vstack_presig_cond) * F.tanh(vstack_pretanh + vstack_pretanh_cond)\n",
    "        hstack_preres = self.htoh(F.sigmoid(hstack_presig + hstack_presig_cond) * F.tanh(hstack_pretanh + hstack_pretanh_cond))\n",
    "        # residual connection for hstack\n",
    "        if not self.residual:\n",
    "            hstack_out = hstack_preres\n",
    "        elif hx.shape[1] < self.channels:\n",
    "            zpad_shape = list(hx.shape)\n",
    "            zpad_shape[1] = self.channels - zpad_shape[1]\n",
    "            hstack_out = hstack_preres + F.concat(hx, F.zeros(zpad_shape), dim=1)\n",
    "        else:\n",
    "            hstack_out = hstack_preres + hx\n",
    "            \n",
    "        return mx.nd.concat(vstack_out, hstack_out, dim=1)\n",
    "    \n",
    "class ConditionalSequential(gluon.nn.HybridSequential):\n",
    "    def __init__(self, **kwargs):\n",
    "        super(ConditionalSequential, self).__init__(**kwargs)\n",
    "        \n",
    "    def hybrid_forward(self, F, x, h):\n",
    "        for block in self._children.values():\n",
    "            x = block(x, h)\n",
    "        return x, h\n",
    "    \n",
    "class ConditionalPixelCNN(gluon.nn.HybridBlock):\n",
    "    def __init__(self, **kwargs):\n",
    "        super(ConditionalPixelCNN, self).__init__(**kwargs)\n",
    "        net1 = ConditionalSequential()\n",
    "        net2 = gluon.nn.HybridSequential()\n",
    "        with self.name_scope():\n",
    "            net1.add(ConditionalLayer(16, (5,5), kernel_type='A', residual=False))\n",
    "            net1.add(ConditionalLayer(16, (5,5), dilation=(1,1), kernel_type='B', residual=False))\n",
    "            net1.add(ConditionalLayer(16, (5,5), dilation=(2,1), kernel_type='B', residual=False))\n",
    "            net1.add(ConditionalLayer(16, (5,5), dilation=(4,1), kernel_type='B', residual=False))\n",
    "            net2.add(GatedLayer(16, (5,5), dilation=(1,1), kernel_type='B', residual=False))\n",
    "            net2.add(GatedLayer(16, (5,5), dilation=(1,1), kernel_type='B', residual=False))\n",
    "            net2.add(gluon.nn.Conv2D(1, (1,1), activation='relu'))\n",
    "        self.net1 = net1\n",
    "        self.net2 = net2\n",
    "            \n",
    "    def hybrid_forward(self, F, x, h):\n",
    "        x = F.tile(x, [1,2,1,1])\n",
    "        x, _ = self.net1(x, h)\n",
    "        return self.net2(x)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Construction of the entine network\n",
    "\n",
    "Residual connection is omitted because it seemed degrading the quality of output images."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "net = ConditionalPixelCNN()\n",
    "net.initialize(ctx=ctx)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_loader = gluon.data.DataLoader(\n",
    "    gluon.data.vision.MNIST('.', transform=lambda x,y: (mx.nd.transpose(x, [2,0,1]).astype(np.float32), y)), batch_size=24)\n",
    "test_loader = gluon.data.DataLoader(\n",
    "    gluon.data.vision.MNIST('.', train=False, transform=lambda x,y: (mx.nd.transpose(x, [2,0,1]).astype(np.float32), y)), batch_size=24)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "training loss: 1294091.96, test loss: 726075.14\n",
      "training loss: 532066.42, test loss: 748445.37\n",
      "training loss: 512952.24, test loss: 716197.27\n",
      "training loss: 501033.49, test loss: 583300.34\n",
      "training loss: 491468.25, test loss: 678552.08\n",
      "training loss: 485389.03, test loss: 574276.62\n",
      "training loss: 478458.28, test loss: 665976.96\n",
      "training loss: 478665.85, test loss: 681455.02\n",
      "training loss: 475491.67, test loss: 593564.66\n",
      "training loss: 469478.30, test loss: 645025.04\n",
      "training loss: 469785.80, test loss: 687960.18\n",
      "training loss: 466937.34, test loss: 621148.13\n",
      "training loss: 465580.18, test loss: 614686.83\n",
      "training loss: 463552.69, test loss: 550659.15\n",
      "training loss: 460203.52, test loss: 630666.10\n",
      "training loss: 463249.93, test loss: 583233.87\n",
      "training loss: 457788.98, test loss: 525475.25\n",
      "training loss: 456011.76, test loss: 541979.24\n",
      "training loss: 454344.21, test loss: 564983.57\n",
      "training loss: 453483.99, test loss: 514895.35\n",
      "training loss: 453024.85, test loss: 602587.30\n",
      "training loss: 456461.52, test loss: 567488.57\n",
      "training loss: 451860.87, test loss: 531824.29\n",
      "training loss: 451091.71, test loss: 559971.99\n",
      "training loss: 449813.12, test loss: 522692.88\n",
      "training loss: 448032.91, test loss: 525662.08\n",
      "training loss: 448115.24, test loss: 561038.13\n",
      "training loss: 447984.05, test loss: 546199.05\n",
      "training loss: 446495.86, test loss: 511367.27\n",
      "training loss: 445557.32, test loss: 505466.65\n"
     ]
    }
   ],
   "source": [
    "trainer = gluon.Trainer(net.collect_params(), 'adam', {'learning_rate': 0.001})\n",
    "\n",
    "for _ in range(30):\n",
    "    cumulative_loss = 0\n",
    "    for x,y in train_loader:\n",
    "        x = mx.nd.array(x, ctx=ctx)\n",
    "        y = mx.nd.array(y, ctx=ctx)\n",
    "        with mx.autograd.record():\n",
    "            cond = mx.nd.one_hot(y, depth=10)\n",
    "            h = net(x,cond)\n",
    "            loss = mx.nd.sum((h - x)**2, axis=[1,2,3])\n",
    "        loss.backward()\n",
    "        trainer.step(24)\n",
    "        cumulative_loss += mx.nd.mean(loss).asscalar()\n",
    "    test_loss = 0\n",
    "    for x, y in test_loader:\n",
    "        x = mx.nd.array(x, ctx=ctx)\n",
    "        y = mx.nd.array(y, ctx=ctx)\n",
    "        cond = mx.nd.one_hot(y, depth=10)\n",
    "        h = net(x, cond)\n",
    "        test_loss += mx.nd.mean(mx.nd.sum((h - x)**2, axis=[1,2,3])).asscalar()\n",
    "    print('training loss: %.2f, test loss: %.2f'%(cumulative_loss/len(train_loader), test_loss/len(test_loader)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Image completion"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "for n, (x, y) in enumerate(test_loader):\n",
    "    x = mx.nd.array(x, ctx=ctx)\n",
    "    y = mx.nd.array(y, ctx=ctx)\n",
    "    reconst =mx.nd.array(x, ctx=ctx)\n",
    "\n",
    "for i in range(14,28):\n",
    "    for j in range(28):\n",
    "        cond = mx.nd.one_hot(y, depth=10)\n",
    "        h = net(reconst,cond)\n",
    "        reconst[:,:,i,j] = h[:,:,i,j]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1728x144 with 24 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(2,12, figsize=(24,2))\n",
    "for i in range(12):\n",
    "    a1, a2 = ax[:, i]\n",
    "    a1.set_xticks([])\n",
    "    a2.set_xticks([])\n",
    "    a1.imshow(x.asnumpy()[i,0,:,:])\n",
    "    a1.plot([0,28], [14,14], 'r')\n",
    "    a2.imshow(reconst.asnumpy()[i,0,:,:])\n",
    "    a2.plot([0,28], [14,14], 'r')\n",
    "    a1.set_xticks([])\n",
    "    a2.set_xticks([])\n",
    "    a1.set_xlim([0,28])\n",
    "    a2.set_xlim([0,28])\n",
    "    a1.set_yticks([])\n",
    "    a2.set_yticks([])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "1st row: Original images are taken from the test dataset.\n",
    "\n",
    "2nd row: The lower half of the images (below the red line) are once removed, and estimated by the Conditional PixelCNN based on the information contained in upper half of the images."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Image generation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = mx.nd.arange(10, ctx=ctx)\n",
    "gen =mx.nd.zeros_like(x[:10], ctx=ctx)\n",
    "\n",
    "for i in range(28):\n",
    "    for j in range(28):\n",
    "        cond = mx.nd.one_hot(y, depth=10)\n",
    "        h = net(gen,cond)\n",
    "        gen[:,:,i,j] = h[:,:,i,j]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x144 with 10 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1,10, figsize=(10,2))\n",
    "for i,a in enumerate(ax):\n",
    "    a.set_xticks([])\n",
    "    a.imshow(gen.asnumpy()[i,0,:,:])\n",
    "    a.set_xlim([0,28])\n",
    "    a.set_yticks([])"
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}