diff --git a/index.html b/index.html
index eafa664..fba5d9b 100644
--- a/index.html
+++ b/index.html
@@ -48,6 +48,8 @@
   <script src="/test/convolutional/Convolution1D.js"></script>
   <script src="/test/convolutional/data_Convolution2D.js"></script>
   <script src="/test/convolutional/Convolution2D.js"></script>
+  <script src="/test/convolutional/data_AtrousConvolution2D.js"></script>
+  <script src="/test/convolutional/AtrousConvolution2D.js"></script>
   <script src="/test/convolutional/data_Convolution3D.js"></script>
   <script src="/test/convolutional/Convolution3D.js"></script>
   <script src="/test/convolutional/data_UpSampling1D.js"></script>
diff --git a/notebooks/convolutional/AtrousConvolution2D.ipynb b/notebooks/convolutional/AtrousConvolution2D.ipynb
new file mode 100644
index 0000000..c8a1e4e
--- /dev/null
+++ b/notebooks/convolutional/AtrousConvolution2D.ipynb
@@ -0,0 +1,367 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Using TensorFlow backend.\n"
+     ]
+    }
+   ],
+   "source": [
+    "import numpy as np\n",
+    "from keras.models import Model\n",
+    "from keras.layers import Input\n",
+    "from keras.layers.convolutional import AtrousConvolution2D\n",
+    "from keras import backend as K"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {
+    "collapsed": true
+   },
+   "outputs": [],
+   "source": [
+    "def format_decimal(arr, places=6):\n",
+    "    return [round(x * 10**places) / 10**places for x in arr]"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### AtrousConvolution2D"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "**[convolutional.AtrousConvolution2D.0] 4 3x3 filters on 5x5x2 input, activation='linear', border_mode='valid', subsample=(1,1), atrous_rate=(2, 2), dim_ordering='tf', bias=True**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "W shape: (3, 3, 2, 4)\n",
+      "W: [0.08681, -0.443261, -0.150965, 0.689552, -0.990562, -0.756862, 0.341498, 0.651706, -0.726587, 0.150187, 0.782644, -0.581596, -0.629344, -0.783246, -0.560605, 0.957248, 0.623366, -0.656118, 0.632449, -0.451853, -0.136592, 0.88006, 0.635299, -0.327776, -0.649179, -0.254336, -0.988623, -0.495147, 0.591325, -0.96949, 0.197687, 0.207609, -0.789705, -0.236113, -0.927048, 0.780823, 0.961842, -0.880116, 0.781092, 0.153803, 0.484959, 0.260368, 0.163684, -0.959122, -0.579947, 0.08937, 0.53823, -0.49861, -0.428209, 0.70479, 0.950013, 0.769707, -0.280984, 0.197718, -0.290409, -0.31962, -0.643838, -0.524612, -0.910275, 0.010863, -0.247495, 0.185611, 0.259884, -0.714799, 0.867683, 0.89276, 0.204593, -0.224467, -0.273624, -0.591309, -0.44647, -0.506928]\n",
+      "b shape: (4,)\n",
+      "b: [0.08681, -0.443261, -0.150965, 0.689552]\n",
+      "\n",
+      "in shape: (5, 5, 2)\n",
+      "in: [-0.990562, -0.756862, 0.341498, 0.651706, -0.726587, 0.150187, 0.782644, -0.581596, -0.629344, -0.783246, -0.560605, 0.957248, 0.623366, -0.656118, 0.632449, -0.451853, -0.136592, 0.88006, 0.635299, -0.327776, -0.649179, -0.254336, -0.988623, -0.495147, 0.591325, -0.96949, 0.197687, 0.207609, -0.789705, -0.236113, -0.927048, 0.780823, 0.961842, -0.880116, 0.781092, 0.153803, 0.484959, 0.260368, 0.163684, -0.959122, -0.579947, 0.08937, 0.53823, -0.49861, -0.428209, 0.70479, 0.950013, 0.769707, -0.280984, 0.197718]\n",
+      "out shape: (1, 1, 4)\n",
+      "out: [-0.449265, 0.56076, -2.92837, 1.056555]\n"
+     ]
+    }
+   ],
+   "source": [
+    "data_in_shape = (5, 5, 2)\n",
+    "conv = AtrousConvolution2D(4, 3, 3, activation='linear', border_mode='valid', \n",
+    "                           subsample=(1, 1), atrous_rate=(2, 2), dim_ordering='tf', bias=True)\n",
+    "\n",
+    "layer_0 = Input(shape=data_in_shape)\n",
+    "layer_1 = conv(layer_0)\n",
+    "model = Model(input=layer_0, output=layer_1)\n",
+    "\n",
+    "# set weights to random (use seed for reproducibility)\n",
+    "weights = []\n",
+    "for w in model.get_weights():\n",
+    "    np.random.seed(100)\n",
+    "    weights.append(2 * np.random.random(w.shape) - 1)\n",
+    "model.set_weights(weights)\n",
+    "print('W shape:', weights[0].shape)\n",
+    "print('W:', format_decimal(weights[0].ravel().tolist()))\n",
+    "print('b shape:', weights[1].shape)\n",
+    "print('b:', format_decimal(weights[1].ravel().tolist()))\n",
+    "\n",
+    "data_in = 2 * np.random.random(data_in_shape) - 1\n",
+    "print('')\n",
+    "print('in shape:', data_in_shape)\n",
+    "print('in:', format_decimal(data_in.ravel().tolist()))\n",
+    "result = model.predict(np.array([data_in]))\n",
+    "print('out shape:', result[0].shape)\n",
+    "print('out:', format_decimal(result[0].ravel().tolist()))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "**[convolutional.AtrousConvolution2D.1] 4 3x3 filters on 5x5x2 input, activation='linear', border_mode='valid', subsample=(1,1), atrous_rate=(2, 2), dim_ordering='tf', bias=False**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "W shape: (3, 3, 2, 4)\n",
+      "W: [0.032797, 0.141335, -0.943052, -0.656957, 0.370554, 0.667794, -0.386068, 0.787226, 0.443088, -0.620122, 0.108455, -0.295736, -0.636215, 0.571204, 0.930966, -0.535293, -0.832877, 0.207097, 0.457986, -0.447522, 0.370613, 0.035735, -0.903031, -0.724262, -0.626065, 0.988636, 0.041331, 0.157579, 0.469638, 0.083924, 0.826307, 0.61584, -0.194004, -0.285551, 0.905753, -0.312737, 0.7302, 0.660555, 0.076323, 0.844939, -0.805707, -0.794305, 0.403015, 0.78096, -0.680879, -0.448855, 0.344983, -0.671394, 0.402742, -0.02473, 0.361356, 0.043096, -0.913207, -0.552127, 0.15041, -0.759133, 0.000233, -0.723981, -0.894383, -0.643446, -0.115264, 0.755175, 0.898528, -0.043665, -0.077761, 0.274578, -0.350784, -0.764844, -0.897798, 0.275317, 0.624532, 0.340521]\n",
+      "\n",
+      "in shape: (5, 5, 2)\n",
+      "in: [0.303535, -0.150862, 0.313191, -0.581677, 0.319849, 0.059247, 0.497041, -0.812486, 0.569044, 0.374484, 0.390157, -0.006267, 0.950722, -0.592945, -0.401959, -0.544688, -0.903662, 0.807943, -0.839793, 0.214433, 0.261693, -0.244116, -0.973518, 0.684439, -0.230125, 0.103332, 0.421076, 0.350558, 0.389129, -0.315086, -0.175219, -0.520418, 0.937672, -0.422886, -0.705377, -0.741319, 0.888112, -0.297131, 0.467135, 0.827779, 0.401975, -0.222937, 0.884519, 0.472983, -0.523071, 0.647547, 0.521227, -0.210582, -0.599624, 0.425193]\n",
+      "out shape: (1, 1, 4)\n",
+      "out: [-0.578839, 1.046696, 1.078234, 0.69352]\n"
+     ]
+    }
+   ],
+   "source": [
+    "data_in_shape = (5, 5, 2)\n",
+    "conv = AtrousConvolution2D(4, 3, 3, activation='linear', border_mode='valid', \n",
+    "                           subsample=(1, 1), atrous_rate=(2, 2), dim_ordering='tf', bias=False)\n",
+    "\n",
+    "layer_0 = Input(shape=data_in_shape)\n",
+    "layer_1 = conv(layer_0)\n",
+    "model = Model(input=layer_0, output=layer_1)\n",
+    "\n",
+    "# set weights to random (use seed for reproducibility)\n",
+    "weights = []\n",
+    "for w in model.get_weights():\n",
+    "    np.random.seed(101)\n",
+    "    weights.append(2 * np.random.random(w.shape) - 1)\n",
+    "model.set_weights(weights)\n",
+    "print('W shape:', weights[0].shape)\n",
+    "print('W:', format_decimal(weights[0].ravel().tolist()))\n",
+    "\n",
+    "data_in = 2 * np.random.random(data_in_shape) - 1\n",
+    "print('')\n",
+    "print('in shape:', data_in_shape)\n",
+    "print('in:', format_decimal(data_in.ravel().tolist()))\n",
+    "result = model.predict(np.array([data_in]))\n",
+    "print('out shape:', result[0].shape)\n",
+    "print('out:', format_decimal(result[0].ravel().tolist()))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "**[convolutional.AtrousConvolution2D.2] 4 3x3 filters on 7x7x2 input, activation='relu', border_mode='valid', subsample=(1,1), atrous_rate=(3,3), dim_ordering='tf', bias=True**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "W shape: (3, 3, 2, 4)\n",
+      "W: [0.195363, 0.351974, -0.401437, 0.461481, 0.157479, 0.618035, -0.665503, -0.37571, -0.284137, -0.016505, -0.019604, 0.78882, -0.635179, -0.277676, 0.621183, -0.333106, 0.037901, -0.47676, -0.738144, -0.755166, 0.103315, -0.941233, 0.20555, 0.501765, -0.501859, -0.604645, 0.147912, 0.803597, 0.032433, -0.197603, 0.034979, 0.874782, 0.700816, -0.648994, 0.039638, 0.011529, -0.302879, -0.858314, 0.151964, -0.196575, 0.82283, 0.617771, -0.735886, 0.429729, -0.306674, 0.228823, -0.425021, 0.971139, -0.455574, 0.638783, 0.199032, 0.44091, 0.91975, -0.814286, -0.088979, 0.765096, 0.395408, -0.357503, -0.275416, 0.467275, 0.245085, -0.563436, 0.96541, -0.199388, 0.43292, -0.389609, 0.298369, -0.25199, -0.266434, -0.949491, -0.303867, -0.024641]\n",
+      "b shape: (4,)\n",
+      "b: [0.195363, 0.351974, -0.401437, 0.461481]\n",
+      "\n",
+      "in shape: (7, 7, 2)\n",
+      "in: [0.157479, 0.618035, -0.665503, -0.37571, -0.284137, -0.016505, -0.019604, 0.78882, -0.635179, -0.277676, 0.621183, -0.333106, 0.037901, -0.47676, -0.738144, -0.755166, 0.103315, -0.941233, 0.20555, 0.501765, -0.501859, -0.604645, 0.147912, 0.803597, 0.032433, -0.197603, 0.034979, 0.874782, 0.700816, -0.648994, 0.039638, 0.011529, -0.302879, -0.858314, 0.151964, -0.196575, 0.82283, 0.617771, -0.735886, 0.429729, -0.306674, 0.228823, -0.425021, 0.971139, -0.455574, 0.638783, 0.199032, 0.44091, 0.91975, -0.814286, -0.088979, 0.765096, 0.395408, -0.357503, -0.275416, 0.467275, 0.245085, -0.563436, 0.96541, -0.199388, 0.43292, -0.389609, 0.298369, -0.25199, -0.266434, -0.949491, -0.303867, -0.024641, -0.221058, -0.27166, -0.014907, -0.92768, -0.54906, -0.731149, 0.600323, 0.352064, -0.730288, 0.065041, 0.53791, -0.398966, -0.180102, -0.233353, 0.95265, 0.076114, 0.820989, -0.637627, -0.434933, -0.836859, -0.781756, 0.2022, 0.420213, -0.359623, 0.460477, 0.643565, -0.143274, -0.63885, -0.803257, 0.287415]\n",
+      "out shape: (1, 1, 4)\n",
+      "out: [0.0, 2.241294, 0.0, 1.107507]\n"
+     ]
+    }
+   ],
+   "source": [
+    "data_in_shape = (7, 7, 2)\n",
+    "conv = AtrousConvolution2D(4, 3, 3, activation='relu', border_mode='valid', \n",
+    "                           subsample=(1, 1), atrous_rate=(3, 3), dim_ordering='tf', bias=True)\n",
+    "\n",
+    "layer_0 = Input(shape=data_in_shape)\n",
+    "layer_1 = conv(layer_0)\n",
+    "model = Model(input=layer_0, output=layer_1)\n",
+    "\n",
+    "# set weights to random (use seed for reproducibility)\n",
+    "weights = []\n",
+    "for w in model.get_weights():\n",
+    "    np.random.seed(102)\n",
+    "    weights.append(2 * np.random.random(w.shape) - 1)\n",
+    "model.set_weights(weights)\n",
+    "print('W shape:', weights[0].shape)\n",
+    "print('W:', format_decimal(weights[0].ravel().tolist()))\n",
+    "print('b shape:', weights[1].shape)\n",
+    "print('b:', format_decimal(weights[1].ravel().tolist()))\n",
+    "\n",
+    "data_in = 2 * np.random.random(data_in_shape) - 1\n",
+    "print('')\n",
+    "print('in shape:', data_in_shape)\n",
+    "print('in:', format_decimal(data_in.ravel().tolist()))\n",
+    "result = model.predict(np.array([data_in]))\n",
+    "print('out shape:', result[0].shape)\n",
+    "print('out:', format_decimal(result[0].ravel().tolist()))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "**[convolutional.AtrousConvolution2D.3] 3 4x4 filters on 4x8x3 input, activation='relu', border_mode='same', subsample=(1,1), atrous_rate=(2,2), dim_ordering='tf', bias=True**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "W shape: (4, 4, 3, 3)\n",
+      "W: [-0.135778, -0.651569, -0.658113, 0.655264, 0.174343, -0.081293, 0.64537, 0.643096, -0.385759, -0.598211, -0.193535, 0.895779, 0.355089, 0.21795, 0.344727, -0.984035, -0.32688, -0.285455, -0.02788, 0.756565, 0.509817, 0.26093, -0.247033, 0.180323, 0.474657, 0.192293, -0.815041, -0.195458, 0.179373, -0.769174, -0.223611, -0.880477, 0.167916, -0.666549, -0.303809, -0.156202, -0.501052, 0.539906, 0.10262, -0.947416, 0.541952, -0.813565, 0.405592, 0.159078, 0.198436, 0.71454, 0.102339, -0.835312, 0.466992, 0.067715, -0.315822, -0.056778, -0.556856, -0.647049, -0.050928, -0.277526, 0.64888, -0.708, 0.639138, -0.632262, 0.126027, -0.751824, 0.387838, -0.924068, -0.411933, 0.793635, -0.432359, -0.568279, -0.942986, -0.929277, 0.746907, 0.954599, -0.470971, 0.887461, -0.919761, -0.423457, 0.359481, 0.23833, 0.4367, 0.136035, 0.837921, 0.01252, 0.76166, -0.067006, -0.569453, 0.707085, -0.685992, -0.972653, 0.424172, -0.023413, -0.494544, 0.239537, -0.357146, -0.832116, 0.911805, -0.493355, 0.799492, 0.907763, 0.590698, -0.657319, -0.899599, 0.66576, -0.026552, -0.808517, 0.761201, 0.060031, 0.911167, -0.541875, -0.052366, -0.677874, -0.540037, 0.858037, 0.841428, 0.664757, 0.159247, -0.408513, 0.157763, 0.953228, -0.419758, 0.13644, 0.857617, 0.935383, -0.725736, 0.482768, 0.919937, 0.516163, 0.078243, 0.504286, 0.149417, 0.153285, -0.686413, -0.515197, 0.529735, -0.671812, 0.737835, 0.760286, -0.825563, 0.862511, -0.185154, 0.436687, -0.452091, -0.048684, 0.142544, 0.590622]\n",
+      "b shape: (3,)\n",
+      "b: [-0.135778, -0.651569, -0.658113]\n",
+      "\n",
+      "in shape: (4, 8, 3)\n",
+      "in: [0.655264, 0.174343, -0.081293, 0.64537, 0.643096, -0.385759, -0.598211, -0.193535, 0.895779, 0.355089, 0.21795, 0.344727, -0.984035, -0.32688, -0.285455, -0.02788, 0.756565, 0.509817, 0.26093, -0.247033, 0.180323, 0.474657, 0.192293, -0.815041, -0.195458, 0.179373, -0.769174, -0.223611, -0.880477, 0.167916, -0.666549, -0.303809, -0.156202, -0.501052, 0.539906, 0.10262, -0.947416, 0.541952, -0.813565, 0.405592, 0.159078, 0.198436, 0.71454, 0.102339, -0.835312, 0.466992, 0.067715, -0.315822, -0.056778, -0.556856, -0.647049, -0.050928, -0.277526, 0.64888, -0.708, 0.639138, -0.632262, 0.126027, -0.751824, 0.387838, -0.924068, -0.411933, 0.793635, -0.432359, -0.568279, -0.942986, -0.929277, 0.746907, 0.954599, -0.470971, 0.887461, -0.919761, -0.423457, 0.359481, 0.23833, 0.4367, 0.136035, 0.837921, 0.01252, 0.76166, -0.067006, -0.569453, 0.707085, -0.685992, -0.972653, 0.424172, -0.023413, -0.494544, 0.239537, -0.357146, -0.832116, 0.911805, -0.493355, 0.799492, 0.907763, 0.590698]\n",
+      "out shape: (4, 8, 3)\n",
+      "out: [1.131317, 0.0, 0.0, 0.809904, 0.0, 0.0, 0.056086, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.251846, 0.0, 0.650556, 0.628197, 0.0, 0.0, 1.029379, 0.0, 2.939306, 0.23459, 0.0, 0.385831, 0.0, 0.0, 1.017464, 0.18825, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 2.247988, 0.083916, 0.0, 0.0, 0.202801, 0.0, 0.137876, 1.309465, 0.0, 0.0, 0.969843, 0.531891, 0.0, 0.0, 0.06576, 0.0, 0.0, 0.0, 0.0, 0.096874, 0.0, 1.147747, 0.0, 0.0, 0.155811, 0.0, 0.331866, 0.0, 0.319746, 0.0, 0.0, 0.0, 0.0, 1.384199, 0.0, 1.856335, 0.0, 1.686605, 0.0, 0.0, 0.0, 0.0, 0.0, 1.802758, 0.0, 0.0, 0.0, 0.102891, 0.0, 0.0, 1.089357, 0.617667, 0.0, 0.0, 0.543751]\n"
+     ]
+    }
+   ],
+   "source": [
+    "data_in_shape = (4, 8, 3)\n",
+    "conv = AtrousConvolution2D(3, 4, 4, activation='relu', border_mode='same', \n",
+    "                           subsample=(1, 1), atrous_rate=(2, 2), dim_ordering='tf', bias=True)\n",
+    "\n",
+    "layer_0 = Input(shape=data_in_shape)\n",
+    "layer_1 = conv(layer_0)\n",
+    "model = Model(input=layer_0, output=layer_1)\n",
+    "\n",
+    "# set weights to random (use seed for reproducibility)\n",
+    "weights = []\n",
+    "for w in model.get_weights():\n",
+    "    np.random.seed(103)\n",
+    "    weights.append(2 * np.random.random(w.shape) - 1)\n",
+    "model.set_weights(weights)\n",
+    "print('W shape:', weights[0].shape)\n",
+    "print('W:', format_decimal(weights[0].ravel().tolist()))\n",
+    "print('b shape:', weights[1].shape)\n",
+    "print('b:', format_decimal(weights[1].ravel().tolist()))\n",
+    "\n",
+    "data_in = 2 * np.random.random(data_in_shape) - 1\n",
+    "print('')\n",
+    "print('in shape:', data_in_shape)\n",
+    "print('in:', format_decimal(data_in.ravel().tolist()))\n",
+    "result = model.predict(np.array([data_in]))\n",
+    "print('out shape:', result[0].shape)\n",
+    "print('out:', format_decimal(result[0].ravel().tolist()))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "**[convolutional.AtrousConvolution2D.4] 4 3x3 filters on 8x8x2 input, activation='relu', border_mode='same', subsample=(1,1), atrous_rate=(4,4), dim_ordering='tf', bias=True**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "W shape: (3, 3, 2, 4)\n",
+      "W: [-0.704159, -0.543403, 0.614987, -0.403051, -0.628587, 0.544275, -0.189121, 0.991809, -0.028349, 0.30284, 0.201092, -0.953154, -0.343083, 0.640537, -0.117391, -0.526633, -0.163892, 0.861458, -0.665795, 0.51919, -0.753596, 0.337708, 0.402785, 0.315812, 0.541968, 0.870102, -0.141283, -0.469735, -0.574406, -0.571765, 0.046045, -0.790043, -0.953079, -0.774733, -0.58544, -0.882147, 0.282331, 0.214534, 0.862629, -0.446078, 0.230215, 0.308982, -0.361978, -0.850926, 0.409799, 0.710712, -0.391058, -0.542441, -0.41638, 0.558765, -0.149949, -0.864061, -0.759997, -0.265417, -0.331629, -0.845702, 0.426935, -0.731659, 0.062865, -0.255582, 0.396332, 0.639196, -0.078254, -0.253246, 0.792036, 0.044105, 0.879927, -0.987868, -0.35422, 0.924714, 0.304954, -0.870234]\n",
+      "b shape: (4,)\n",
+      "b: [-0.704159, -0.543403, 0.614987, -0.403051]\n",
+      "\n",
+      "in shape: (8, 8, 2)\n",
+      "in: [-0.628587, 0.544275, -0.189121, 0.991809, -0.028349, 0.30284, 0.201092, -0.953154, -0.343083, 0.640537, -0.117391, -0.526633, -0.163892, 0.861458, -0.665795, 0.51919, -0.753596, 0.337708, 0.402785, 0.315812, 0.541968, 0.870102, -0.141283, -0.469735, -0.574406, -0.571765, 0.046045, -0.790043, -0.953079, -0.774733, -0.58544, -0.882147, 0.282331, 0.214534, 0.862629, -0.446078, 0.230215, 0.308982, -0.361978, -0.850926, 0.409799, 0.710712, -0.391058, -0.542441, -0.41638, 0.558765, -0.149949, -0.864061, -0.759997, -0.265417, -0.331629, -0.845702, 0.426935, -0.731659, 0.062865, -0.255582, 0.396332, 0.639196, -0.078254, -0.253246, 0.792036, 0.044105, 0.879927, -0.987868, -0.35422, 0.924714, 0.304954, -0.870234, -0.611808, 0.138739, 0.526129, -0.170329, 0.504868, -0.454139, 0.245234, 0.479283, 0.651594, -0.447885, -0.78716, -0.537367, -0.578885, -0.3999, -0.981424, 0.584097, -0.461375, 0.948907, 0.763872, -0.720626, -0.572361, 0.780094, -0.385637, 0.90998, -0.697513, 0.366373, -0.427147, 0.423096, -0.55829, 0.769474, -0.484912, -0.731131, -0.794623, 0.19042, 0.537997, -0.390004, -0.697993, -0.710914, -0.222541, -0.375822, 0.236956, -0.978421, -0.834872, 0.29127, 0.125375, 0.82053, 0.953645, 0.931806, 0.170904, -0.474968, -0.266042, -0.021253, 0.449315, 0.266589, 0.405989, 0.663497, 0.762836, -0.421562, 0.664906, 0.130767]\n",
+      "out shape: (8, 8, 4)\n",
+      "out: [1.008111, 0.862141, 1.53732, 0.0, 0.0, 0.0, 2.278924, 0.0, 0.192175, 0.256028, 1.00272, 0.0, 0.0, 0.0, 0.0, 1.28356, 0.0, 0.0, 1.295987, 0.0, 0.0, 0.0, 0.522589, 0.0, 0.0, 0.0, 1.593745, 0.0, 0.094799, 1.375385, 1.350499, 0.637961, 0.0, 0.393093, 1.508163, 1.045792, 0.0, 0.580916, 0.774659, 0.0, 0.0, 0.0, 1.090937, 0.2098, 0.0, 0.0, 0.706958, 0.948835, 0.0, 0.0, 0.70246, 1.233015, 0.0, 0.0, 0.0, 0.0, 0.0, 0.237716, 0.150121, 0.0, 0.012937, 1.067614, 0.259571, 0.895802, 0.0, 0.127769, 0.0, 0.0, 0.0, 0.0, 0.0, 0.517957, 0.0, 0.0, 0.524233, 0.079491, 0.0, 0.0, 0.0, 1.424679, 0.0, 0.0, 0.798433, 0.0, 0.752754, 0.49452, 0.564184, 1.320468, 0.0, 0.0, 1.469927, 0.075539, 0.0, 0.760117, 0.0, 0.961547, 0.938599, 1.264279, 0.857871, 0.0, 0.114331, 0.0, 0.753579, 0.0, 0.0, 0.0, 0.020539, 0.0, 0.0, 0.0, 1.035394, 0.0, 0.0, 0.0, 0.745973, 0.0, 0.0, 0.070491, 0.0, 0.0, 0.0, 0.0, 0.308166, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.911015, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.178493, 1.44698, 0.0, 0.0, 0.0, 1.45985, 0.298695, 0.0, 0.0, 0.0, 0.0, 0.0, 0.402046, 0.536043, 1.160239, 0.0, 0.0, 0.0, 0.0, 0.575623, 0.009206, 0.639194, 0.0, 0.353097, 0.0, 0.471996, 0.411527, 1.148312, 0.966958, 1.172197, 0.0, 0.419191, 0.0, 2.142206, 0.0, 0.0, 0.0, 0.0, 0.0, 0.508378, 0.0, 1.110668, 1.82902, 0.0, 0.0, 2.081267, 0.05382, 0.0, 0.0, 1.5163, 1.476838, 1.364327, 0.033732, 1.076706, 0.609169, 0.0, 0.149234, 2.181282, 0.780896, 0.0, 0.0, 0.763411, 0.0, 0.0, 0.0, 2.053607, 0.12022, 0.0, 0.0, 0.0, 0.29116, 0.0, 0.0, 0.656705, 0.0, 0.0, 0.0, 1.055828, 0.634612, 0.0, 0.0, 0.0, 0.363467, 1.779839, 0.0, 1.270641, 0.15881, 0.0, 0.020913, 0.854418, 0.0, 0.0, 0.0, 0.436751, 0.0, 0.0, 0.0, 0.0, 0.0, 0.436203, 0.236268, 0.0, 0.0, 0.0, 0.0, 0.189446, 0.0, 0.0, 0.0, 0.827813, 0.0, 0.0, 0.0, 0.31375, 0.0, 0.0, 0.0, 0.755046, 0.0]\n"
+     ]
+    }
+   ],
+   "source": [
+    "data_in_shape = (8, 8, 2)\n",
+    "conv = AtrousConvolution2D(4, 3, 3, activation='relu', border_mode='same', \n",
+    "                           subsample=(1, 1), atrous_rate=(4, 4), dim_ordering='tf', bias=True)\n",
+    "\n",
+    "layer_0 = Input(shape=data_in_shape)\n",
+    "layer_1 = conv(layer_0)\n",
+    "model = Model(input=layer_0, output=layer_1)\n",
+    "\n",
+    "# set weights to random (use seed for reproducibility)\n",
+    "weights = []\n",
+    "for w in model.get_weights():\n",
+    "    np.random.seed(104)\n",
+    "    weights.append(2 * np.random.random(w.shape) - 1)\n",
+    "model.set_weights(weights)\n",
+    "print('W shape:', weights[0].shape)\n",
+    "print('W:', format_decimal(weights[0].ravel().tolist()))\n",
+    "print('b shape:', weights[1].shape)\n",
+    "print('b:', format_decimal(weights[1].ravel().tolist()))\n",
+    "\n",
+    "data_in = 2 * np.random.random(data_in_shape) - 1\n",
+    "print('')\n",
+    "print('in shape:', data_in_shape)\n",
+    "print('in:', format_decimal(data_in.ravel().tolist()))\n",
+    "result = model.predict(np.array([data_in]))\n",
+    "print('out shape:', result[0].shape)\n",
+    "print('out:', format_decimal(result[0].ravel().tolist()))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": true
+   },
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "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": "3.5.2"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 0
+}
diff --git a/src/layers/convolutional/AtrousConvolution2D.js b/src/layers/convolutional/AtrousConvolution2D.js
new file mode 100644
index 0000000..84174a2
--- /dev/null
+++ b/src/layers/convolutional/AtrousConvolution2D.js
@@ -0,0 +1,260 @@
+import * as activations from '../../activations'
+import Tensor from '../../Tensor'
+import Layer from '../../engine/Layer'
+import ops from 'ndarray-ops'
+import gemm from 'ndarray-gemm'
+import unpack from 'ndarray-unpack'
+import flattenDeep from 'lodash/flattenDeep'
+
+/**
+ * AtrousConvolution2D layer class
+ */
+export default class AtrousConvolution2D extends Layer {
+  /**
+   * Creates a AtrousConvolution2D layer
+   * @param {number} nbFilter - Number of convolution filters to use.
+   * @param {number} nbRow - Number of rows in the convolution kernel.
+   * @param {number} nbCol - Number of columns in the convolution kernel.
+   * @param {Object} [attrs] - layer attributes
+   */
+  constructor (nbFilter, nbRow, nbCol, attrs = {}) {
+    super(attrs)
+    const {
+      activation = 'linear',
+      borderMode = 'valid',
+      subsample = [1, 1],
+      atrousRate = [1, 1],
+      dimOrdering = 'tf',
+      bias = true
+    } = attrs
+
+    this.kernelShape = [nbFilter, nbRow, nbCol]
+
+    this.activation = activations[activation]
+
+    if (borderMode === 'valid' || borderMode === 'same') {
+      this.borderMode = borderMode
+    } else {
+      throw new Error(`${this.name} [AtrousConvolution2D layer] Invalid borderMode.`)
+    }
+
+    this.subsample = subsample
+
+    this.atrousRate = atrousRate
+
+    if (dimOrdering === 'tf' || dimOrdering === 'th') {
+      this.dimOrdering = dimOrdering
+    } else {
+      throw new Error(`${this.name} [AtrousConvolution2D layer] Only tf and th dim ordering are allowed.`)
+    }
+
+    this.bias = bias
+
+    // Layer weights specification
+    this.params = this.bias ? ['W', 'b'] : ['W']
+  }
+
+  /**
+   * Method for setting layer weights. Extends `super` method.
+   * W weight tensor is converted to `tf` mode if in `th` mode.
+   * In `tf` mode, W weight tensor has shape [nbRow, nbCol, inputChannels, nbFilter]
+   * In `th` mode, W weight tensor has shape [nbFilter, inputChannels, nbRow, nbCol]
+   * @param {Tensor[]} weightsArr - array of weights which are instances of Tensor
+   */
+  setWeights (weightsArr) {
+    if (this.dimOrdering === 'th') {
+      // W
+      weightsArr[0].tensor = weightsArr[0].tensor.transpose(2, 3, 1, 0)
+    }
+    super.setWeights(weightsArr)
+  }
+
+  /**
+   * Method for computing output dimensions and padding, based on input
+   * dimensions, kernel size, and padding mode.
+   * For tensorflow implementation of padding, see:
+   * https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/framework/common_shape_fns.cc
+   * @param {Tensor} x
+   * @returns {number[]} [outputRows, outputCols, outputChannels]
+   */
+  _calcOutputShape = x => {
+    const inputRows = x.tensor.shape[0]
+    const inputCols = x.tensor.shape[1]
+    const [nbFilter, nbRow, nbCol] = this.kernelShape
+
+    // effective shape after filter dilation
+    const nbRowDilated = nbRow + (nbRow - 1) * (this.atrousRate[0] - 1)
+    const nbColDilated = nbCol + (nbCol - 1) * (this.atrousRate[1] - 1)
+
+    const outputRows = this.borderMode === 'same'
+      ? Math.floor((inputRows + this.subsample[0] - 1) / this.subsample[0])
+      : Math.floor((inputRows - nbRowDilated + this.subsample[0]) / this.subsample[0])
+    const outputCols = this.borderMode === 'same'
+      ? Math.floor((inputCols + this.subsample[1] - 1) / this.subsample[1])
+      : Math.floor((inputCols - nbColDilated + this.subsample[1]) / this.subsample[1])
+    const outputChannels = nbFilter
+
+    const paddingRow = this.borderMode === 'same'
+      ? Math.max(0, Math.floor((outputRows - 1) * this.subsample[0] + nbRowDilated - inputRows))
+      : 0
+    const paddingCol = this.borderMode === 'same'
+      ? Math.max(0, Math.floor((outputCols - 1) * this.subsample[1] + nbColDilated - inputCols))
+      : 0
+    const paddingRowBefore = Math.floor(paddingRow / 2)
+    const paddingRowAfter = paddingRow - paddingRowBefore
+    const paddingColBefore = Math.floor(paddingCol / 2)
+    const paddingColAfter = paddingCol - paddingColBefore
+
+    this.outputShape = [outputRows, outputCols, outputChannels]
+    this.inputPadding = [paddingRowBefore, paddingRowAfter, paddingColBefore, paddingColAfter]
+  }
+
+  /**
+   * Pad input tensor if necessary, for borderMode='same'.
+   * See above for notes on calculating padding.
+   * @param {Tensor} x
+   * @returns {Tensor} x
+   */
+  _padInput = x => {
+    if (this.borderMode === 'same') {
+      const [inputRows, inputCols, inputChannels] = x.tensor.shape
+      const [paddingRowBefore, paddingRowAfter, paddingColBefore, paddingColAfter] = this.inputPadding
+      const newRows = inputRows + paddingRowBefore + paddingRowAfter
+      const newCols = inputCols + paddingColBefore + paddingColAfter
+      let _x = new Tensor([], [newRows, newCols, inputChannels])
+      ops.assign(
+        _x.tensor
+          .hi(inputRows + paddingRowBefore, inputCols + paddingColBefore, inputChannels)
+          .lo(paddingRowBefore, paddingColBefore, 0),
+        x.tensor
+      )
+      x.tensor = _x.tensor
+    }
+    return x
+  }
+
+  /**
+   * Convert input image to column matrix
+   * @param {Tensor} x
+   * @returns {Tensor} x
+   */
+  _im2col = x => {
+    const [inputRows, inputCols, inputChannels] = x.tensor.shape
+    const nbRow = this.kernelShape[1]
+    const nbCol = this.kernelShape[2]
+    const outputRows = this.outputShape[0]
+    const outputCols = this.outputShape[1]
+    const nbPatches = outputRows * outputCols
+    const patchLen = nbRow * nbCol * inputChannels
+
+    // effective shape after filter dilation
+    const nbRowDilated = nbRow + (nbRow - 1) * (this.atrousRate[0] - 1)
+    const nbColDilated = nbCol + (nbCol - 1) * (this.atrousRate[1] - 1)
+
+    const imColsMat = new Tensor([], [nbPatches, patchLen])
+
+    let patch = new Tensor([], [patchLen])
+    let n = 0
+    for (let i = 0; i <= inputRows - nbRowDilated; i += this.subsample[0]) {
+      for (let j = 0; j <= inputCols - nbColDilated; j += this.subsample[1]) {
+        const patchData = flattenDeep(unpack(
+          x.tensor
+            .hi(i + nbRowDilated, j + nbColDilated, inputChannels)
+            .lo(i, j, 0)
+            .step(this.atrousRate[0], this.atrousRate[1], 1)
+        ))
+        patch.replaceTensorData(patchData)
+        ops.assign(imColsMat.tensor.pick(n, null), patch.tensor)
+        n += 1
+      }
+    }
+
+    return imColsMat
+  }
+
+  /**
+   * Convert filter weights to row matrix
+   * @param {Tensor} x
+   * @returns {Tensor} x
+   */
+  _w2row = x => {
+    const inputChannels = x.tensor.shape[2]
+    const [nbFilter, nbRow, nbCol] = this.kernelShape
+    const patchLen = nbRow * nbCol * inputChannels
+
+    const wRowsMat = new Tensor([], [patchLen, nbFilter])
+
+    let patch = new Tensor([], [patchLen])
+    for (let n = 0; n < nbFilter; n++) {
+      const patchData = flattenDeep(unpack(
+        this.weights.W.tensor.pick(null, null, null, n)
+      ))
+      patch.replaceTensorData(patchData)
+      ops.assign(wRowsMat.tensor.pick(null, n), patch.tensor)
+    }
+
+    return wRowsMat
+  }
+
+  /**
+   * Method for layer computational logic
+   * @param {Tensor} x
+   * @returns {Tensor} x
+   */
+  call = x => {
+    // convert to tf ordering
+    if (this.dimOrdering === 'th') {
+      x.tensor = x.tensor.transpose(1, 2, 0)
+    }
+
+    this._calcOutputShape(x)
+    this._padInput(x)
+
+    const imColsMat = this._im2col(x)
+    const wRowsMat = this._w2row(x)
+
+    const nbFilter = this.kernelShape[0]
+    const outputRows = this.outputShape[0]
+    const outputCols = this.outputShape[1]
+    const nbPatches = outputRows * outputCols
+    const matMul = new Tensor([], [nbPatches, nbFilter])
+    if (this.bias) {
+      for (let n = 0; n < nbFilter; n++) {
+        ops.assigns(matMul.tensor.pick(null, n), this.weights.b.tensor.get(n))
+      }
+    }
+
+    if (x._useWeblas) {
+      const bias = this.bias
+        ? this.weights.b.tensor.data
+        : new Float32Array(wRowsMat.tensor.shape[1])
+      matMul.tensor.data = weblas.sgemm(
+        imColsMat.tensor.shape[0], wRowsMat.tensor.shape[1], imColsMat.tensor.shape[1], // M, N, K
+        1, imColsMat.tensor.data, wRowsMat.tensor.data, // alpha, A, B
+        1, bias // beta, C
+      )
+    } else {
+      gemm(matMul.tensor, imColsMat.tensor, wRowsMat.tensor, 1, 1)
+    }
+
+    let output = new Tensor([], this.outputShape)
+    let outputChannel = new Tensor([], [outputRows, outputCols])
+    for (let n = 0; n < nbFilter; n++) {
+      const outputChannelData = flattenDeep(unpack(
+        matMul.tensor.pick(null, n)
+      ))
+      outputChannel.replaceTensorData(outputChannelData)
+      ops.assign(output.tensor.pick(null, null, n), outputChannel.tensor)
+    }
+    x.tensor = output.tensor
+
+    this.activation(x)
+
+    // convert back to th ordering if necessary
+    if (this.dimOrdering === 'th') {
+      x.tensor = x.tensor.transpose(2, 0, 1)
+    }
+
+    return x
+  }
+}
diff --git a/src/layers/convolutional/index.js b/src/layers/convolutional/index.js
index 93fe4bf..e2890ef 100644
--- a/src/layers/convolutional/index.js
+++ b/src/layers/convolutional/index.js
@@ -1,5 +1,6 @@
 import Convolution1D from './Convolution1D'
 import Convolution2D from './Convolution2D'
+import AtrousConvolution2D from './AtrousConvolution2D'
 import Convolution3D from './Convolution3D'
 import UpSampling1D from './UpSampling1D'
 import UpSampling2D from './UpSampling2D'
@@ -11,6 +12,7 @@ import ZeroPadding3D from './ZeroPadding3D'
 export {
   Convolution1D,
   Convolution2D,
+  AtrousConvolution2D,
   Convolution3D,
   UpSampling1D,
   UpSampling2D,
diff --git a/test/convolutional/AtrousConvolution2D.js b/test/convolutional/AtrousConvolution2D.js
new file mode 100644
index 0000000..22295ad
--- /dev/null
+++ b/test/convolutional/AtrousConvolution2D.js
@@ -0,0 +1,110 @@
+/* eslint-env browser, mocha */
+
+describe('convolutional layer: AtrousConvolution2D', function () {
+  const assert = chai.assert
+  const styles = testGlobals.styles
+  const logTime = testGlobals.logTime
+  const stringifyCondensed = testGlobals.stringifyCondensed
+  const approxEquals = KerasJS.testUtils.approxEquals
+  const layers = KerasJS.layers
+
+  const testParams = [
+    {
+      inputShape: [5, 5, 2],
+      kernelShape: [4, 3, 3],
+      attrs: { activation: 'linear', borderMode: 'valid', subsample: [1, 1], atrousRate: [2, 2], dimOrdering: 'tf', bias: true }
+    },
+    {
+      inputShape: [5, 5, 2],
+      kernelShape: [4, 3, 3],
+      attrs: { activation: 'linear', borderMode: 'valid', subsample: [1, 1], atrousRate: [2, 2], dimOrdering: 'tf', bias: false }
+    },
+    {
+      inputShape: [7, 7, 2],
+      kernelShape: [4, 3, 3],
+      attrs: { activation: 'relu', borderMode: 'valid', subsample: [1, 1], atrousRate: [3, 3], dimOrdering: 'tf', bias: true }
+    },
+    {
+      inputShape: [4, 8, 3],
+      kernelShape: [3, 4, 4],
+      attrs: { activation: 'relu', borderMode: 'same', subsample: [1, 1], atrousRate: [2, 2], dimOrdering: 'tf', bias: true }
+    },
+    {
+      inputShape: [8, 8, 2],
+      kernelShape: [4, 3, 3],
+      attrs: { activation: 'relu', borderMode: 'same', subsample: [1, 1], atrousRate: [4, 4], dimOrdering: 'tf', bias: true }
+    }
+  ]
+
+  before(function () {
+    console.log('\n%cconvolutional layer: AtrousConvolution2D', styles.h1)
+  })
+
+  /*********************************************************
+  * CPU
+  *********************************************************/
+
+  describe('CPU', function () {
+    before(function () {
+      console.log('\n%cCPU', styles.h2)
+    })
+
+    testParams.forEach(({ inputShape, kernelShape, attrs }, i) => {
+      const key = `convolutional.AtrousConvolution2D.${i}`
+      const [inputRows, inputCols, inputChannels] = inputShape
+      const [nbFilter, nbRow, nbCol] = kernelShape
+      const title = `[${key}] [CPU] test: ${nbFilter} ${nbRow}x${nbCol} filters on ${inputRows}x${inputCols}x${inputChannels} input, activation='${attrs.activation}', border_mode='${attrs.borderMode}', subsample=${attrs.subsample}, atrousRate=${attrs.atrousRate}, dim_ordering='${attrs.dimOrdering}', bias=${attrs.bias}`
+
+      it(title, function () {
+        console.log(`\n%c${title}`, styles.h3)
+        let testLayer = new layers.AtrousConvolution2D(nbFilter, nbRow, nbCol, attrs)
+        testLayer.setWeights(TEST_DATA[key].weights.map(w => new KerasJS.Tensor(w.data, w.shape)))
+        let t = new KerasJS.Tensor(TEST_DATA[key].input.data, TEST_DATA[key].input.shape)
+        console.log('%cin', styles.h4, stringifyCondensed(t.tensor))
+        const startTime = performance.now()
+        t = testLayer.call(t)
+        const endTime = performance.now()
+        console.log('%cout', styles.h4, stringifyCondensed(t.tensor))
+        logTime(startTime, endTime)
+        const dataExpected = new Float32Array(TEST_DATA[key].expected.data)
+        const shapeExpected = TEST_DATA[key].expected.shape
+        assert.deepEqual(t.tensor.shape, shapeExpected)
+        assert.isTrue(approxEquals(t.tensor, dataExpected))
+      })
+    })
+  })
+
+  /*********************************************************
+  * GPU
+  *********************************************************/
+
+  describe('GPU', function () {
+    before(function () {
+      console.log('\n%cGPU', styles.h2)
+    })
+
+    testParams.forEach(({ inputShape, kernelShape, attrs }, i) => {
+      const key = `convolutional.AtrousConvolution2D.${i}`
+      const [inputRows, inputCols, inputChannels] = inputShape
+      const [nbFilter, nbRow, nbCol] = kernelShape
+      const title = `[${key}] [GPU] test: ${nbFilter} ${nbRow}x${nbCol} filters on ${inputRows}x${inputCols}x${inputChannels} input, activation='${attrs.activation}', border_mode='${attrs.borderMode}', subsample=${attrs.subsample}, atrousRate=${attrs.atrousRate}, dim_ordering='${attrs.dimOrdering}', bias=${attrs.bias}`
+
+      it(title, function () {
+        console.log(`\n%c${title}`, styles.h3)
+        let testLayer = new layers.AtrousConvolution2D(nbFilter, nbRow, nbCol, attrs)
+        testLayer.setWeights(TEST_DATA[key].weights.map(w => new KerasJS.Tensor(w.data, w.shape)))
+        let t = new KerasJS.Tensor(TEST_DATA[key].input.data, TEST_DATA[key].input.shape, { useWeblas: true })
+        console.log('%cin', styles.h4, stringifyCondensed(t.tensor))
+        const startTime = performance.now()
+        t = testLayer.call(t)
+        const endTime = performance.now()
+        console.log('%cout', styles.h4, stringifyCondensed(t.tensor))
+        logTime(startTime, endTime)
+        const dataExpected = new Float32Array(TEST_DATA[key].expected.data)
+        const shapeExpected = TEST_DATA[key].expected.shape
+        assert.deepEqual(t.tensor.shape, shapeExpected)
+        assert.isTrue(approxEquals(t.tensor, dataExpected))
+      })
+    })
+  })
+})
diff --git a/test/convolutional/data_AtrousConvolution2D.js b/test/convolutional/data_AtrousConvolution2D.js
new file mode 100644
index 0000000..467b55d
--- /dev/null
+++ b/test/convolutional/data_AtrousConvolution2D.js
@@ -0,0 +1,106 @@
+// TEST DATA
+// Keyed by mocha test ID
+// Python code for generating test data can be found in the matching jupyter notebook in folder `notebooks/`.
+
+(function () {
+  var DATA = {
+    'convolutional.AtrousConvolution2D.0': {
+      input: {
+        data: [-0.990562, -0.756862, 0.341498, 0.651706, -0.726587, 0.150187, 0.782644, -0.581596, -0.629344, -0.783246, -0.560605, 0.957248, 0.623366, -0.656118, 0.632449, -0.451853, -0.136592, 0.88006, 0.635299, -0.327776, -0.649179, -0.254336, -0.988623, -0.495147, 0.591325, -0.96949, 0.197687, 0.207609, -0.789705, -0.236113, -0.927048, 0.780823, 0.961842, -0.880116, 0.781092, 0.153803, 0.484959, 0.260368, 0.163684, -0.959122, -0.579947, 0.08937, 0.53823, -0.49861, -0.428209, 0.70479, 0.950013, 0.769707, -0.280984, 0.197718],
+        shape: [5, 5, 2]
+      },
+      weights: [
+        {
+          data: [0.08681, -0.443261, -0.150965, 0.689552, -0.990562, -0.756862, 0.341498, 0.651706, -0.726587, 0.150187, 0.782644, -0.581596, -0.629344, -0.783246, -0.560605, 0.957248, 0.623366, -0.656118, 0.632449, -0.451853, -0.136592, 0.88006, 0.635299, -0.327776, -0.649179, -0.254336, -0.988623, -0.495147, 0.591325, -0.96949, 0.197687, 0.207609, -0.789705, -0.236113, -0.927048, 0.780823, 0.961842, -0.880116, 0.781092, 0.153803, 0.484959, 0.260368, 0.163684, -0.959122, -0.579947, 0.08937, 0.53823, -0.49861, -0.428209, 0.70479, 0.950013, 0.769707, -0.280984, 0.197718, -0.290409, -0.31962, -0.643838, -0.524612, -0.910275, 0.010863, -0.247495, 0.185611, 0.259884, -0.714799, 0.867683, 0.89276, 0.204593, -0.224467, -0.273624, -0.591309, -0.44647, -0.506928],
+          shape: [3, 3, 2, 4]
+        },
+        {
+          data: [0.08681, -0.443261, -0.150965, 0.689552],
+          shape: [4]
+        }
+      ],
+      expected: {
+        data: [-0.449265, 0.56076, -2.92837, 1.056555],
+        shape: [1, 1, 4]
+      }
+    },
+    'convolutional.AtrousConvolution2D.1': {
+      input: {
+        data: [0.303535, -0.150862, 0.313191, -0.581677, 0.319849, 0.059247, 0.497041, -0.812486, 0.569044, 0.374484, 0.390157, -0.006267, 0.950722, -0.592945, -0.401959, -0.544688, -0.903662, 0.807943, -0.839793, 0.214433, 0.261693, -0.244116, -0.973518, 0.684439, -0.230125, 0.103332, 0.421076, 0.350558, 0.389129, -0.315086, -0.175219, -0.520418, 0.937672, -0.422886, -0.705377, -0.741319, 0.888112, -0.297131, 0.467135, 0.827779, 0.401975, -0.222937, 0.884519, 0.472983, -0.523071, 0.647547, 0.521227, -0.210582, -0.599624, 0.425193],
+        shape: [5, 5, 2]
+      },
+      weights: [
+        {
+          data: [0.032797, 0.141335, -0.943052, -0.656957, 0.370554, 0.667794, -0.386068, 0.787226, 0.443088, -0.620122, 0.108455, -0.295736, -0.636215, 0.571204, 0.930966, -0.535293, -0.832877, 0.207097, 0.457986, -0.447522, 0.370613, 0.035735, -0.903031, -0.724262, -0.626065, 0.988636, 0.041331, 0.157579, 0.469638, 0.083924, 0.826307, 0.61584, -0.194004, -0.285551, 0.905753, -0.312737, 0.7302, 0.660555, 0.076323, 0.844939, -0.805707, -0.794305, 0.403015, 0.78096, -0.680879, -0.448855, 0.344983, -0.671394, 0.402742, -0.02473, 0.361356, 0.043096, -0.913207, -0.552127, 0.15041, -0.759133, 0.000233, -0.723981, -0.894383, -0.643446, -0.115264, 0.755175, 0.898528, -0.043665, -0.077761, 0.274578, -0.350784, -0.764844, -0.897798, 0.275317, 0.624532, 0.340521],
+          shape: [3, 3, 2, 4]
+        }
+      ],
+      expected: {
+        data: [-0.578839, 1.046696, 1.078234, 0.69352],
+        shape: [1, 1, 4]
+      }
+    },
+    'convolutional.AtrousConvolution2D.2': {
+      input: {
+        data: [0.157479, 0.618035, -0.665503, -0.37571, -0.284137, -0.016505, -0.019604, 0.78882, -0.635179, -0.277676, 0.621183, -0.333106, 0.037901, -0.47676, -0.738144, -0.755166, 0.103315, -0.941233, 0.20555, 0.501765, -0.501859, -0.604645, 0.147912, 0.803597, 0.032433, -0.197603, 0.034979, 0.874782, 0.700816, -0.648994, 0.039638, 0.011529, -0.302879, -0.858314, 0.151964, -0.196575, 0.82283, 0.617771, -0.735886, 0.429729, -0.306674, 0.228823, -0.425021, 0.971139, -0.455574, 0.638783, 0.199032, 0.44091, 0.91975, -0.814286, -0.088979, 0.765096, 0.395408, -0.357503, -0.275416, 0.467275, 0.245085, -0.563436, 0.96541, -0.199388, 0.43292, -0.389609, 0.298369, -0.25199, -0.266434, -0.949491, -0.303867, -0.024641, -0.221058, -0.27166, -0.014907, -0.92768, -0.54906, -0.731149, 0.600323, 0.352064, -0.730288, 0.065041, 0.53791, -0.398966, -0.180102, -0.233353, 0.95265, 0.076114, 0.820989, -0.637627, -0.434933, -0.836859, -0.781756, 0.2022, 0.420213, -0.359623, 0.460477, 0.643565, -0.143274, -0.63885, -0.803257, 0.287415],
+        shape: [7, 7, 2]
+      },
+      weights: [
+        {
+          data: [0.195363, 0.351974, -0.401437, 0.461481, 0.157479, 0.618035, -0.665503, -0.37571, -0.284137, -0.016505, -0.019604, 0.78882, -0.635179, -0.277676, 0.621183, -0.333106, 0.037901, -0.47676, -0.738144, -0.755166, 0.103315, -0.941233, 0.20555, 0.501765, -0.501859, -0.604645, 0.147912, 0.803597, 0.032433, -0.197603, 0.034979, 0.874782, 0.700816, -0.648994, 0.039638, 0.011529, -0.302879, -0.858314, 0.151964, -0.196575, 0.82283, 0.617771, -0.735886, 0.429729, -0.306674, 0.228823, -0.425021, 0.971139, -0.455574, 0.638783, 0.199032, 0.44091, 0.91975, -0.814286, -0.088979, 0.765096, 0.395408, -0.357503, -0.275416, 0.467275, 0.245085, -0.563436, 0.96541, -0.199388, 0.43292, -0.389609, 0.298369, -0.25199, -0.266434, -0.949491, -0.303867, -0.024641],
+          shape: [3, 3, 2, 4]
+        },
+        {
+          data: [0.195363, 0.351974, -0.401437, 0.461481],
+          shape: [4]
+        }
+      ],
+      expected: {
+        data: [0.0, 2.241294, 0.0, 1.107507],
+        shape: [1, 1, 4]
+      }
+    },
+    'convolutional.AtrousConvolution2D.3': {
+      input: {
+        data: [0.655264, 0.174343, -0.081293, 0.64537, 0.643096, -0.385759, -0.598211, -0.193535, 0.895779, 0.355089, 0.21795, 0.344727, -0.984035, -0.32688, -0.285455, -0.02788, 0.756565, 0.509817, 0.26093, -0.247033, 0.180323, 0.474657, 0.192293, -0.815041, -0.195458, 0.179373, -0.769174, -0.223611, -0.880477, 0.167916, -0.666549, -0.303809, -0.156202, -0.501052, 0.539906, 0.10262, -0.947416, 0.541952, -0.813565, 0.405592, 0.159078, 0.198436, 0.71454, 0.102339, -0.835312, 0.466992, 0.067715, -0.315822, -0.056778, -0.556856, -0.647049, -0.050928, -0.277526, 0.64888, -0.708, 0.639138, -0.632262, 0.126027, -0.751824, 0.387838, -0.924068, -0.411933, 0.793635, -0.432359, -0.568279, -0.942986, -0.929277, 0.746907, 0.954599, -0.470971, 0.887461, -0.919761, -0.423457, 0.359481, 0.23833, 0.4367, 0.136035, 0.837921, 0.01252, 0.76166, -0.067006, -0.569453, 0.707085, -0.685992, -0.972653, 0.424172, -0.023413, -0.494544, 0.239537, -0.357146, -0.832116, 0.911805, -0.493355, 0.799492, 0.907763, 0.590698],
+        shape: [4, 8, 3]
+      },
+      weights: [
+        {
+          data: [-0.135778, -0.651569, -0.658113, 0.655264, 0.174343, -0.081293, 0.64537, 0.643096, -0.385759, -0.598211, -0.193535, 0.895779, 0.355089, 0.21795, 0.344727, -0.984035, -0.32688, -0.285455, -0.02788, 0.756565, 0.509817, 0.26093, -0.247033, 0.180323, 0.474657, 0.192293, -0.815041, -0.195458, 0.179373, -0.769174, -0.223611, -0.880477, 0.167916, -0.666549, -0.303809, -0.156202, -0.501052, 0.539906, 0.10262, -0.947416, 0.541952, -0.813565, 0.405592, 0.159078, 0.198436, 0.71454, 0.102339, -0.835312, 0.466992, 0.067715, -0.315822, -0.056778, -0.556856, -0.647049, -0.050928, -0.277526, 0.64888, -0.708, 0.639138, -0.632262, 0.126027, -0.751824, 0.387838, -0.924068, -0.411933, 0.793635, -0.432359, -0.568279, -0.942986, -0.929277, 0.746907, 0.954599, -0.470971, 0.887461, -0.919761, -0.423457, 0.359481, 0.23833, 0.4367, 0.136035, 0.837921, 0.01252, 0.76166, -0.067006, -0.569453, 0.707085, -0.685992, -0.972653, 0.424172, -0.023413, -0.494544, 0.239537, -0.357146, -0.832116, 0.911805, -0.493355, 0.799492, 0.907763, 0.590698, -0.657319, -0.899599, 0.66576, -0.026552, -0.808517, 0.761201, 0.060031, 0.911167, -0.541875, -0.052366, -0.677874, -0.540037, 0.858037, 0.841428, 0.664757, 0.159247, -0.408513, 0.157763, 0.953228, -0.419758, 0.13644, 0.857617, 0.935383, -0.725736, 0.482768, 0.919937, 0.516163, 0.078243, 0.504286, 0.149417, 0.153285, -0.686413, -0.515197, 0.529735, -0.671812, 0.737835, 0.760286, -0.825563, 0.862511, -0.185154, 0.436687, -0.452091, -0.048684, 0.142544, 0.590622],
+          shape: [4, 4, 3, 3]
+        },
+        {
+          data: [-0.135778, -0.651569, -0.658113],
+          shape: [3]
+        }
+      ],
+      expected: {
+        data: [1.131317, 0.0, 0.0, 0.809904, 0.0, 0.0, 0.056086, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.251846, 0.0, 0.650556, 0.628197, 0.0, 0.0, 1.029379, 0.0, 2.939306, 0.23459, 0.0, 0.385831, 0.0, 0.0, 1.017464, 0.18825, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 2.247988, 0.083916, 0.0, 0.0, 0.202801, 0.0, 0.137876, 1.309465, 0.0, 0.0, 0.969843, 0.531891, 0.0, 0.0, 0.06576, 0.0, 0.0, 0.0, 0.0, 0.096874, 0.0, 1.147747, 0.0, 0.0, 0.155811, 0.0, 0.331866, 0.0, 0.319746, 0.0, 0.0, 0.0, 0.0, 1.384199, 0.0, 1.856335, 0.0, 1.686605, 0.0, 0.0, 0.0, 0.0, 0.0, 1.802758, 0.0, 0.0, 0.0, 0.102891, 0.0, 0.0, 1.089357, 0.617667, 0.0, 0.0, 0.543751],
+        shape: [4, 8, 3]
+      }
+    },
+    'convolutional.AtrousConvolution2D.4': {
+      input: {
+        data: [-0.628587, 0.544275, -0.189121, 0.991809, -0.028349, 0.30284, 0.201092, -0.953154, -0.343083, 0.640537, -0.117391, -0.526633, -0.163892, 0.861458, -0.665795, 0.51919, -0.753596, 0.337708, 0.402785, 0.315812, 0.541968, 0.870102, -0.141283, -0.469735, -0.574406, -0.571765, 0.046045, -0.790043, -0.953079, -0.774733, -0.58544, -0.882147, 0.282331, 0.214534, 0.862629, -0.446078, 0.230215, 0.308982, -0.361978, -0.850926, 0.409799, 0.710712, -0.391058, -0.542441, -0.41638, 0.558765, -0.149949, -0.864061, -0.759997, -0.265417, -0.331629, -0.845702, 0.426935, -0.731659, 0.062865, -0.255582, 0.396332, 0.639196, -0.078254, -0.253246, 0.792036, 0.044105, 0.879927, -0.987868, -0.35422, 0.924714, 0.304954, -0.870234, -0.611808, 0.138739, 0.526129, -0.170329, 0.504868, -0.454139, 0.245234, 0.479283, 0.651594, -0.447885, -0.78716, -0.537367, -0.578885, -0.3999, -0.981424, 0.584097, -0.461375, 0.948907, 0.763872, -0.720626, -0.572361, 0.780094, -0.385637, 0.90998, -0.697513, 0.366373, -0.427147, 0.423096, -0.55829, 0.769474, -0.484912, -0.731131, -0.794623, 0.19042, 0.537997, -0.390004, -0.697993, -0.710914, -0.222541, -0.375822, 0.236956, -0.978421, -0.834872, 0.29127, 0.125375, 0.82053, 0.953645, 0.931806, 0.170904, -0.474968, -0.266042, -0.021253, 0.449315, 0.266589, 0.405989, 0.663497, 0.762836, -0.421562, 0.664906, 0.130767],
+        shape: [8, 8, 2]
+      },
+      weights: [
+        {
+          data: [-0.704159, -0.543403, 0.614987, -0.403051, -0.628587, 0.544275, -0.189121, 0.991809, -0.028349, 0.30284, 0.201092, -0.953154, -0.343083, 0.640537, -0.117391, -0.526633, -0.163892, 0.861458, -0.665795, 0.51919, -0.753596, 0.337708, 0.402785, 0.315812, 0.541968, 0.870102, -0.141283, -0.469735, -0.574406, -0.571765, 0.046045, -0.790043, -0.953079, -0.774733, -0.58544, -0.882147, 0.282331, 0.214534, 0.862629, -0.446078, 0.230215, 0.308982, -0.361978, -0.850926, 0.409799, 0.710712, -0.391058, -0.542441, -0.41638, 0.558765, -0.149949, -0.864061, -0.759997, -0.265417, -0.331629, -0.845702, 0.426935, -0.731659, 0.062865, -0.255582, 0.396332, 0.639196, -0.078254, -0.253246, 0.792036, 0.044105, 0.879927, -0.987868, -0.35422, 0.924714, 0.304954, -0.870234],
+          shape: [3, 3, 2, 4]
+        },
+        {
+          data: [-0.704159, -0.543403, 0.614987, -0.403051],
+          shape: [4]
+        }
+      ],
+      expected: {
+        data: [1.008111, 0.862141, 1.53732, 0.0, 0.0, 0.0, 2.278924, 0.0, 0.192175, 0.256028, 1.00272, 0.0, 0.0, 0.0, 0.0, 1.28356, 0.0, 0.0, 1.295987, 0.0, 0.0, 0.0, 0.522589, 0.0, 0.0, 0.0, 1.593745, 0.0, 0.094799, 1.375385, 1.350499, 0.637961, 0.0, 0.393093, 1.508163, 1.045792, 0.0, 0.580916, 0.774659, 0.0, 0.0, 0.0, 1.090937, 0.2098, 0.0, 0.0, 0.706958, 0.948835, 0.0, 0.0, 0.70246, 1.233015, 0.0, 0.0, 0.0, 0.0, 0.0, 0.237716, 0.150121, 0.0, 0.012937, 1.067614, 0.259571, 0.895802, 0.0, 0.127769, 0.0, 0.0, 0.0, 0.0, 0.0, 0.517957, 0.0, 0.0, 0.524233, 0.079491, 0.0, 0.0, 0.0, 1.424679, 0.0, 0.0, 0.798433, 0.0, 0.752754, 0.49452, 0.564184, 1.320468, 0.0, 0.0, 1.469927, 0.075539, 0.0, 0.760117, 0.0, 0.961547, 0.938599, 1.264279, 0.857871, 0.0, 0.114331, 0.0, 0.753579, 0.0, 0.0, 0.0, 0.020539, 0.0, 0.0, 0.0, 1.035394, 0.0, 0.0, 0.0, 0.745973, 0.0, 0.0, 0.070491, 0.0, 0.0, 0.0, 0.0, 0.308166, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.911015, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.178493, 1.44698, 0.0, 0.0, 0.0, 1.45985, 0.298695, 0.0, 0.0, 0.0, 0.0, 0.0, 0.402046, 0.536043, 1.160239, 0.0, 0.0, 0.0, 0.0, 0.575623, 0.009206, 0.639194, 0.0, 0.353097, 0.0, 0.471996, 0.411527, 1.148312, 0.966958, 1.172197, 0.0, 0.419191, 0.0, 2.142206, 0.0, 0.0, 0.0, 0.0, 0.0, 0.508378, 0.0, 1.110668, 1.82902, 0.0, 0.0, 2.081267, 0.05382, 0.0, 0.0, 1.5163, 1.476838, 1.364327, 0.033732, 1.076706, 0.609169, 0.0, 0.149234, 2.181282, 0.780896, 0.0, 0.0, 0.763411, 0.0, 0.0, 0.0, 2.053607, 0.12022, 0.0, 0.0, 0.0, 0.29116, 0.0, 0.0, 0.656705, 0.0, 0.0, 0.0, 1.055828, 0.634612, 0.0, 0.0, 0.0, 0.363467, 1.779839, 0.0, 1.270641, 0.15881, 0.0, 0.020913, 0.854418, 0.0, 0.0, 0.0, 0.436751, 0.0, 0.0, 0.0, 0.0, 0.0, 0.436203, 0.236268, 0.0, 0.0, 0.0, 0.0, 0.189446, 0.0, 0.0, 0.0, 0.827813, 0.0, 0.0, 0.0, 0.31375, 0.0, 0.0, 0.0, 0.755046, 0.0],
+        shape: [8, 8, 4]
+      }
+    }
+  }
+
+  window.TEST_DATA = Object.assign({}, window.TEST_DATA, DATA)
+})()