Skip to content

Instantly share code, notes, and snippets.

@jcheong0428

jcheong0428/CNN_git_gist.ipynb

Created August 13, 2018 06:06
Show Gist options
  • Save jcheong0428/1ee086e9bbfa1b37ac5bb0cff3e7f23c to your computer and use it in GitHub Desktop.
Save jcheong0428/1ee086e9bbfa1b37ac5bb0cff3e7f23c to your computer and use it in GitHub Desktop.
Download ZIP
Display the source blob
Display the rendered blob
Raw
CNN_git_gist.ipynb
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Load the fashion mnist dataset."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T05:52:45.120183",
"start_time": "2018-08-13T05:52:44.983209Z"
}
},
"outputs": [],
"source": [
"import numpy as np\n",
"fashion = np.load('fashion.npz')\n",
"# split to traning and testing set\n",
"train_data = fashion['train_data']\n",
"test_data = fashion['test_data']\n",
"train_labels = fashion['train_labels']\n",
"test_labels = fashion['test_labels']"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T05:54:39.267488",
"start_time": "2018-08-13T05:54:39.263517Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"array([[0., 1., 0.],\n",
" [0., 1., 0.],\n",
" [0., 1., 0.],\n",
" ...,\n",
" [1., 0., 0.],\n",
" [1., 0., 0.],\n",
" [0., 1., 0.]])"
]
},
"execution_count": 11,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# the labels are a samples x targets \n",
"train_labels"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T05:52:45.158668",
"start_time": "2018-08-13T05:52:45.149314Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"(600, 28, 28, 1)"
]
},
"execution_count": 2,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# check data shape. (samples, 28 px by 28px, 1 color)\n",
"train_data.shape"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T05:52:45.701006",
"start_time": "2018-08-13T05:52:45.311364Z"
}
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"\n",
"Bad key \"ckend\" on line 1 in\n",
"/Users/jinhyuncheong/.matplotlib/matplotlibrc.\n",
"You probably need to get an updated matplotlibrc file from\n",
"http://github.com/matplotlib/matplotlib/blob/master/matplotlibrc.template\n",
"or from the matplotlib source distribution\n"
]
},
{
"data": {
"text/plain": [
"<matplotlib.image.AxesImage at 0x1063c3198>"
]
},
"execution_count": 3,
"metadata": {},
"output_type": "execute_result"
},
{
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
"text/plain": [
"<matplotlib.figure.Figure at 0x104f2c1d0>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# check out what an image looks like\n",
"import matplotlib.pyplot as plt\n",
"%matplotlib inline\n",
"plt.imshow(train_data[50,:,:,0])"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T05:52:47.775863",
"start_time": "2018-08-13T05:52:45.702851Z"
}
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"/Users/jinhyuncheong/anaconda2/envs/py36/lib/python3.6/site-packages/h5py/__init__.py:34: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
" from ._conv import register_converters as _register_converters\n",
"Using TensorFlow backend.\n"
]
}
],
"source": [
"from keras.models import Sequential\n",
"from keras.layers import Dense\n",
"# reshape to 1D data (simple NN)\n",
"train_data_1d = train_data.reshape((train_data.shape[0],-1))\n",
"test_data_1d = test_data.reshape((test_data.shape[0],-1))"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T05:53:28.546679",
"start_time": "2018-08-13T05:53:27.527802Z"
},
"scrolled": false
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Train on 540 samples, validate on 60 samples\n",
"Epoch 1/10\n",
"540/540 [==============================] - 0s 384us/step - loss: 0.4365 - acc: 0.8389 - val_loss: 0.2445 - val_acc: 0.9333\n",
"Epoch 2/10\n",
"540/540 [==============================] - 0s 118us/step - loss: 0.1523 - acc: 0.9426 - val_loss: 0.2819 - val_acc: 0.9000\n",
"Epoch 3/10\n",
"540/540 [==============================] - 0s 116us/step - loss: 0.1253 - acc: 0.9556 - val_loss: 0.2449 - val_acc: 0.9333\n",
"Epoch 4/10\n",
"540/540 [==============================] - 0s 118us/step - loss: 0.1070 - acc: 0.9630 - val_loss: 0.2555 - val_acc: 0.9333\n",
"Epoch 5/10\n",
"540/540 [==============================] - 0s 118us/step - loss: 0.0986 - acc: 0.9667 - val_loss: 0.3319 - val_acc: 0.9167\n",
"Epoch 6/10\n",
"540/540 [==============================] - 0s 125us/step - loss: 0.0892 - acc: 0.9704 - val_loss: 0.2906 - val_acc: 0.9000\n",
"Epoch 7/10\n",
"540/540 [==============================] - 0s 122us/step - loss: 0.0797 - acc: 0.9759 - val_loss: 0.3043 - val_acc: 0.9333\n",
"Epoch 8/10\n",
"540/540 [==============================] - 0s 127us/step - loss: 0.0816 - acc: 0.9722 - val_loss: 0.2834 - val_acc: 0.9000\n",
"Epoch 9/10\n",
"540/540 [==============================] - 0s 118us/step - loss: 0.0665 - acc: 0.9833 - val_loss: 0.3975 - val_acc: 0.9000\n",
"Epoch 10/10\n",
"540/540 [==============================] - 0s 116us/step - loss: 0.0613 - acc: 0.9796 - val_loss: 0.3244 - val_acc: 0.9000\n"
]
},
{
"data": {
"text/plain": [
"<keras.callbacks.History at 0x115be0940>"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# init model\n",
"dense = Sequential()\n",
"# input layer\n",
"dense.add(Dense(10,input_shape = (train_data_1d.shape[-1], ),\n",
" activation='relu'\n",
" ))\n",
"# output layer, three units b/c three classes\n",
"dense.add(Dense(3,activation='softmax'))\n",
"dense.compile(optimizer='adam', # sgd, adam: how to step through gradient descent\n",
" loss='categorical_crossentropy', # categorical_crossentropy, mean_squared_error\n",
" metrics=['accuracy']\n",
" )\n",
"dense.fit(train_data_1d,train_labels,batch_size=10,epochs=10,\n",
" validation_split=.1)"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T05:53:33.626209",
"start_time": "2018-08-13T05:53:33.503077Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"0.91\n"
]
},
{
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAAA5wAAAA0CAYAAADmFTfPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4xLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvAOZPmwAAB6BJREFUeJzt3V2oZXUZx/HvrzOOpkK+DaIzU1pa\nMVRqHMQoQixoLMkuIpUCiUKCIpMirJvoooug0i6kELW8CDVMSEKSMKFuMseMfMuaLJ3x9eRo5QSj\nMz1d7FUdxdxrPP+1z9nL7+dmzlrnP2s9e+/n/+z9nL1eUlVIkiRJktTaq1Y7AEmSJEnSONlwSpIk\nSZIGYcMpSZIkSRqEDackSZIkaRA2nJIkSZKkQdhwSpIkSZIGMXjDmWRrkvuTbE9y8dD7k1pIsjnJ\nrUnuTXJPkgu79Uck+VmSP3b/Hr7asUrTJFlIcmeSn3TLxye5ravL1yVZv9oxStMkOSzJ9Ul+n+S+\nJO+wJmseJbmo+2xxd5JrkhxkXdaYDdpwJlkALgPOBLYA5yXZMuQ+pUb2Ap+vqi3AacCnu9y9GLil\nqk4EbumWpbXuQuC+ZctfBy6pqhOAp4BPrEpU0v75NvDTqnozcBKTnLYma64k2Qh8FlisqrcAC8C5\nWJc1YkN/w3kqsL2qHqiqZ4FrgbMH3qe0YlX1aFX9pvv5H0w+2Gxkkr9Xd8OuBj60OhFK/STZBHwA\nuKJbDnAGcH03xDzWmpfkNcC7gSsBqurZqnoaa7Lm0zrg1UnWAQcDj2Jd1ogN3XBuBHYsW97ZrZPm\nRpLjgFOA24Cjq+rR7lePAUevUlhSX5cCXwT+1S0fCTxdVXu7Zeuy5sHxwBLwve7w8CuSHII1WXOm\nqh4GvgE8xKTR/BtwB9ZljZgXDZJeQpJDgR8Bn6uqvy//XVUVUKsSmNRDkrOAJ6rqjtWORVqhdcDb\nge9U1SnAbl5w+Kw1WfOgO8/4bCZ/RDkWOATYuqpBSQMbuuF8GNi8bHlTt05a85IcwKTZ/EFV3dCt\nfjzJMd3vjwGeWK34pB7eCXwwyV+YnNJwBpPz4A7rDuUC67Lmw05gZ1Xd1i1fz6QBtSZr3rwX+HNV\nLVXVc8ANTGq1dVmjNXTDeTtwYnflrfVMToq+ceB9SivWned2JXBfVX1r2a9uBM7vfj4f+PGsY5P6\nqqovVdWmqjqOSf39eVV9FLgV+HA3zDzWmldVjwE7krypW/Ue4F6syZo/DwGnJTm4+6zxn1y2Lmu0\nMjkCZcAdJO9ncg7RAnBVVX1t0B1KDSR5F/BL4C7+d+7bl5mcx/lD4LXAg8BHqmrXqgQp7YckpwNf\nqKqzkryeyTeeRwB3Ah+rqj2rGZ80TZKTmVz8aj3wAPBxJn84tyZrriT5KnAOkyvi3wl8ksk5m9Zl\njdLgDackSZIk6ZXJiwZJkiRJkgZhwylJkiRJGoQNpyRJkiRpEDackiRJkqRBzKThTHLBLPYjDc1c\n1hiYxxoLc1ljYB5r7Gb1DacTSWNhLmsMzGONhbmsMTCPNWoeUitJkiRJGsQg9+E86oiFOm7zAf9d\nXnpyHxuOXNjv7dz11IapY956+NLUMX/43cFTx+zZfMjUMQfu2D11zDRvfNs/p47pE2+r7czqcUO7\nmFvpE88Lc3DfM7tZOPT5z9m85UUrfeZnK7Oc5632Nc3Lyb9WXiyP+zzuPvH0mQ+t6k6r7bSYN7Oc\nD330eT37aDWv+ng5c+859nAAB+53PK3yq8V+WtWCVnO4lVnO4T5aPc+t4lnuxfK4j7WWO7N8rVpp\nMSfm8X2vRX3bu2sX+57ZnT5jB2k4F086qH598+YVb+cN131q6pg/nfPdqWPed+zJU8dsv+S0qWNO\nuOhXU8dMc/Mjv506pk+8rbYzq8cN7WJupU88fXJw3vKilT7PTSuznOet9jVNq/xrpc/jbjUfWtWd\nVttpMW9m+Vr10ef17KPVvOpjrb2ft3hcrfKv1eehWebpLOdwH2vpPb+VtZY7s3ytWmkxJ+bxfa9F\nfXvkm5ey56EdvRrOXofUJtma5P4k25NcvLLwJEmSJEmvBFMbziQLwGXAmcAW4LwkW4YOTJIkSZI0\n3/p8w3kqsL2qHqiqZ4FrgbOHDUuSJEmSNO/6NJwbgR3Llnd2654nyQVJtiXZtvTkvlbxSZIkSZLm\nVLPbolTV5VW1WFWLL+eKtJIkSZKkcenTcD4MLL/k7KZunSRJkiRJ/1efhvN24MQkxydZD5wL3Dhs\nWJIkSZKkebdu2oCq2pvkM8DNwAJwVVXdM3hkkta0md5j7ZzpQ1rdC7CPafcIm+X9ZNeaWb4OffSL\nZ+X3fJvVfc/Wolk+9rE+h5I0ZlMbToCqugm4aeBYJEmSJEkj0uc+nFcleSLJ3bMISJIkSZI0Dn3O\n4fw+sHXgOCRJkiRJIzO14ayqXwC7ZhCLJEmSJGlEmt2HM8kFSbYl2bb05L5Wm5UkSZIkzalmDWdV\nXV5Vi1W1uOHIhVablSRJkiTNqWYNpyRJkiRJy9lwSpIkSZIGkap66QHJNcDpwFHA48BXqurKKf9n\nCXhw2aqjgL+uKFJpbTCXNQbmscbCXNYYmMeaR6+rqg19Bk5tOFtIsq2qFgffkTQwc1ljYB5rLMxl\njYF5rLHzkFpJkiRJ0iBsOCVJkiRJg5hVw3n5jPYjDc1c1hiYxxoLc1ljYB5r1GZyDqckSZIk6ZXH\nQ2olSZIkSYOw4ZQkSZIkDcKGU5IkSZI0CBtOSZIkSdIgbDglSZIkSYP4N3tppF+aPqOkAAAAAElF\nTkSuQmCC\n",
"text/plain": [
"<matplotlib.figure.Figure at 0x115e3ffd0>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# Check accuracy\n",
"pred_test_lables = dense.predict(test_data_1d).argmax(axis=1)\n",
"plt.matshow([test_labels.argmax(axis=1), pred_test_lables])\n",
"print(np.mean(pred_test_lables == test_labels.argmax(axis=1)))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Deep network: add more layers"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T05:55:48.165963",
"start_time": "2018-08-13T05:55:46.543463Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Train on 540 samples, validate on 60 samples\n",
"Epoch 1/10\n",
"540/540 [==============================] - 0s 650us/step - loss: 0.4715 - acc: 0.8296 - val_loss: 0.3065 - val_acc: 0.8833\n",
"Epoch 2/10\n",
"540/540 [==============================] - 0s 155us/step - loss: 0.1524 - acc: 0.9352 - val_loss: 0.2829 - val_acc: 0.8833\n",
"Epoch 3/10\n",
"540/540 [==============================] - 0s 145us/step - loss: 0.1241 - acc: 0.9556 - val_loss: 0.3179 - val_acc: 0.9000\n",
"Epoch 4/10\n",
"540/540 [==============================] - 0s 137us/step - loss: 0.0916 - acc: 0.9648 - val_loss: 0.3148 - val_acc: 0.9000\n",
"Epoch 5/10\n",
"540/540 [==============================] - 0s 142us/step - loss: 0.1085 - acc: 0.9611 - val_loss: 0.2935 - val_acc: 0.9000\n",
"Epoch 6/10\n",
"540/540 [==============================] - 0s 145us/step - loss: 0.0893 - acc: 0.9648 - val_loss: 0.3901 - val_acc: 0.9000\n",
"Epoch 7/10\n",
"540/540 [==============================] - 0s 144us/step - loss: 0.0816 - acc: 0.9667 - val_loss: 0.3236 - val_acc: 0.9000\n",
"Epoch 8/10\n",
"540/540 [==============================] - 0s 135us/step - loss: 0.0527 - acc: 0.9833 - val_loss: 0.5578 - val_acc: 0.9167\n",
"Epoch 9/10\n",
"540/540 [==============================] - 0s 139us/step - loss: 0.0889 - acc: 0.9685 - val_loss: 0.3668 - val_acc: 0.9167\n",
"Epoch 10/10\n",
"540/540 [==============================] - 0s 139us/step - loss: 0.0558 - acc: 0.9833 - val_loss: 0.4266 - val_acc: 0.9167\n",
"0.92\n"
]
},
{
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAAA5wAAAA0CAYAAADmFTfPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
"text/plain": [
"<matplotlib.figure.Figure at 0x1162029e8>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"dense = Sequential()\n",
"# add input layer\n",
"dense.add(Dense(10,input_shape = (train_data_1d.shape[-1], ),\n",
" activation='relu'\n",
" ))\n",
"# add more layers\n",
"dense.add(Dense(100,activation='relu'))\n",
"dense.add(Dense(100,activation='relu'))\n",
"dense.add(Dense(100,activation='relu'))\n",
"# output layer, three units b/c three classes\n",
"dense.add(Dense(3,activation='softmax'))\n",
"# compile model\n",
"dense.compile(optimizer='adam', # sgd, adam: how to step through gradient descent\n",
" loss='categorical_crossentropy', # categorical_crossentropy, mean_squared_error\n",
" metrics=['accuracy']\n",
" )\n",
"# train model\n",
"dense.fit(train_data_1d,train_labels,batch_size=10,epochs=10,\n",
" validation_split=.1)\n",
"# evaluate model\n",
"pred_test_lables = dense.predict(test_data_1d).argmax(axis=1)\n",
"plt.matshow([test_labels.argmax(axis=1), pred_test_lables])\n",
"print(np.mean(pred_test_lables == test_labels.argmax(axis=1)))"
]
},
{
"cell_type": "code",
"execution_count": 16,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T05:56:54.253425",
"start_time": "2018-08-13T05:56:54.244342Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"100/100 [==============================] - 0s 45us/step\n"
]
},
{
"data": {
"text/plain": [
"[0.38440396517515185, 0.92]"
]
},
"execution_count": 16,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# loss, accuracy are shown\n",
"dense.evaluate(test_data_1d,test_labels)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Train a Convolutional Neural Network (CNN)"
]
},
{
"cell_type": "code",
"execution_count": 18,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T05:59:38.035286",
"start_time": "2018-08-13T05:59:38.023618Z"
}
},
"outputs": [],
"source": [
"from keras.layers import Conv2D, Flatten\n",
"from keras.optimizers import Adam\n",
"opt = Adam(lr=10e-3)"
]
},
{
"cell_type": "code",
"execution_count": 19,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T05:59:41.856558",
"start_time": "2018-08-13T05:59:38.301564Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Train on 540 samples, validate on 60 samples\n",
"Epoch 1/10\n",
"540/540 [==============================] - 1s 1ms/step - loss: 0.3294 - acc: 0.9000 - val_loss: 0.2575 - val_acc: 0.9667\n",
"Epoch 2/10\n",
"540/540 [==============================] - 0s 552us/step - loss: 0.2112 - acc: 0.9426 - val_loss: 0.3937 - val_acc: 0.9167\n",
"Epoch 3/10\n",
"540/540 [==============================] - 0s 556us/step - loss: 0.1249 - acc: 0.9630 - val_loss: 0.2403 - val_acc: 0.9333\n",
"Epoch 4/10\n",
"540/540 [==============================] - 0s 565us/step - loss: 0.0776 - acc: 0.9741 - val_loss: 0.3273 - val_acc: 0.9333\n",
"Epoch 5/10\n",
"540/540 [==============================] - 0s 565us/step - loss: 0.1128 - acc: 0.9648 - val_loss: 0.3107 - val_acc: 0.9500\n",
"Epoch 6/10\n",
"540/540 [==============================] - 0s 553us/step - loss: 0.0746 - acc: 0.9741 - val_loss: 0.3468 - val_acc: 0.9500\n",
"Epoch 7/10\n",
"540/540 [==============================] - 0s 555us/step - loss: 0.0616 - acc: 0.9833 - val_loss: 0.4066 - val_acc: 0.9333\n",
"Epoch 8/10\n",
"540/540 [==============================] - 0s 574us/step - loss: 0.0415 - acc: 0.9852 - val_loss: 0.5781 - val_acc: 0.9500\n",
"Epoch 9/10\n",
"540/540 [==============================] - 0s 550us/step - loss: 0.0326 - acc: 0.9907 - val_loss: 0.4597 - val_acc: 0.9500\n",
"Epoch 10/10\n",
"540/540 [==============================] - 0s 539us/step - loss: 0.0272 - acc: 0.9944 - val_loss: 0.5759 - val_acc: 0.9333\n"
]
},
{
"data": {
"text/plain": [
"<keras.callbacks.History at 0x1166f1fd0>"
]
},
"execution_count": 19,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"cnn = Sequential()\n",
"# Add convolutional layer\n",
"cnn.add(Conv2D(10, kernel_size=3, # 3x3 pixel \n",
" input_shape=(train_data.shape[1],train_data.shape[2],train_data.shape[3]),\n",
" activation='relu'))\n",
"# Add another convolutional layer\n",
"cnn.add(Conv2D(10, kernel_size = 3, activation='relu'))\n",
"# flatten the outputs\n",
"cnn.add(Flatten())\n",
"# predict 3 classes using softmax function\n",
"cnn.add(Dense(3,activation='softmax'))\n",
"# compile the model\n",
"cnn.compile(loss='categorical_crossentropy', \n",
" optimizer=opt,\n",
" metrics=['accuracy'])\n",
"# train model\n",
"cnn.fit(train_data,train_labels,epochs=10,batch_size=10,validation_split=.1)"
]
},
{
"cell_type": "code",
"execution_count": 20,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T06:00:00.402144",
"start_time": "2018-08-13T06:00:00.396650Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"_________________________________________________________________\n",
"Layer (type) Output Shape Param # \n",
"=================================================================\n",
"conv2d_1 (Conv2D) (None, 26, 26, 10) 100 \n",
"_________________________________________________________________\n",
"conv2d_2 (Conv2D) (None, 24, 24, 10) 910 \n",
"_________________________________________________________________\n",
"flatten_1 (Flatten) (None, 5760) 0 \n",
"_________________________________________________________________\n",
"dense_10 (Dense) (None, 3) 17283 \n",
"=================================================================\n",
"Total params: 18,293\n",
"Trainable params: 18,293\n",
"Non-trainable params: 0\n",
"_________________________________________________________________\n"
]
}
],
"source": [
"# Display structure of model\n",
"# parameter = 10 units, 9 pix / unit, + 10 bias term (where threshold is), = 910\n",
"# less params up front, more params later to read out. \n",
"cnn.summary()"
]
},
{
"cell_type": "code",
"execution_count": 21,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T06:00:01.316430",
"start_time": "2018-08-13T06:00:01.310133Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"_________________________________________________________________\n",
"Layer (type) Output Shape Param # \n",
"=================================================================\n",
"dense_5 (Dense) (None, 10) 7850 \n",
"_________________________________________________________________\n",
"dense_6 (Dense) (None, 100) 1100 \n",
"_________________________________________________________________\n",
"dense_7 (Dense) (None, 100) 10100 \n",
"_________________________________________________________________\n",
"dense_8 (Dense) (None, 100) 10100 \n",
"_________________________________________________________________\n",
"dense_9 (Dense) (None, 3) 303 \n",
"=================================================================\n",
"Total params: 29,453\n",
"Trainable params: 29,453\n",
"Non-trainable params: 0\n",
"_________________________________________________________________\n"
]
}
],
"source": [
"# more parameters up front,\n",
"dense.summary()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# CNN using max pooling"
]
},
{
"cell_type": "code",
"execution_count": 22,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T06:00:12.256784",
"start_time": "2018-08-13T06:00:12.254111Z"
}
},
"outputs": [],
"source": [
"from keras.layers import MaxPool2D"
]
},
{
"cell_type": "code",
"execution_count": 23,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T06:01:25.159554",
"start_time": "2018-08-13T06:01:21.791130Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Train on 540 samples, validate on 60 samples\n",
"Epoch 1/10\n",
"540/540 [==============================] - 1s 1ms/step - loss: 0.3876 - acc: 0.8722 - val_loss: 0.2111 - val_acc: 0.9167\n",
"Epoch 2/10\n",
"540/540 [==============================] - 0s 504us/step - loss: 0.1593 - acc: 0.9278 - val_loss: 0.1940 - val_acc: 0.9000\n",
"Epoch 3/10\n",
"540/540 [==============================] - 0s 509us/step - loss: 0.1337 - acc: 0.9426 - val_loss: 0.1937 - val_acc: 0.9500\n",
"Epoch 4/10\n",
"540/540 [==============================] - 0s 508us/step - loss: 0.0919 - acc: 0.9593 - val_loss: 0.1876 - val_acc: 0.9167\n",
"Epoch 5/10\n",
"540/540 [==============================] - 0s 508us/step - loss: 0.0794 - acc: 0.9741 - val_loss: 0.2761 - val_acc: 0.9167\n",
"Epoch 6/10\n",
"540/540 [==============================] - 0s 510us/step - loss: 0.0717 - acc: 0.9759 - val_loss: 0.2321 - val_acc: 0.9167\n",
"Epoch 7/10\n",
"540/540 [==============================] - 0s 501us/step - loss: 0.0617 - acc: 0.9741 - val_loss: 0.2351 - val_acc: 0.9000\n",
"Epoch 8/10\n",
"540/540 [==============================] - 0s 510us/step - loss: 0.0521 - acc: 0.9815 - val_loss: 0.3101 - val_acc: 0.9333\n",
"Epoch 9/10\n",
"540/540 [==============================] - 0s 507us/step - loss: 0.0512 - acc: 0.9815 - val_loss: 0.2474 - val_acc: 0.9333\n",
"Epoch 10/10\n",
"540/540 [==============================] - 0s 507us/step - loss: 0.0419 - acc: 0.9796 - val_loss: 0.2255 - val_acc: 0.9333\n"
]
},
{
"data": {
"text/plain": [
"<keras.callbacks.History at 0x116d90ac8>"
]
},
"execution_count": 23,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"cnn_pool = Sequential()\n",
"# input layer with convolution\n",
"cnn_pool.add(Conv2D(10, kernel_size=3, # 3x3 pixel \n",
" input_shape=(train_data.shape[1],train_data.shape[2],train_data.shape[3]),\n",
" activation='relu'))\n",
"# pool the results, get a smaller input shape\n",
"cnn_pool.add(MaxPool2D(2)) #max pool over 2x2 pixels\n",
"# add another convolution\n",
"cnn_pool.add(Conv2D(10, kernel_size = 3, activation='relu'))\n",
"# pool results again\n",
"cnn_pool.add(MaxPool2D(2)) #max pool over 2x2 pixels\n",
"# flatten layer\n",
"cnn_pool.add(Flatten())\n",
"# output layer\n",
"cnn_pool.add(Dense(3,activation='softmax'))\n",
"# compile and train model\n",
"cnn_pool.compile(loss='categorical_crossentropy', \n",
" optimizer=opt,\n",
" metrics=['accuracy'])\n",
"cnn_pool.fit(train_data,train_labels,epochs=10,batch_size=10,validation_split=.1)"
]
},
{
"cell_type": "code",
"execution_count": 24,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T06:01:25.804509",
"start_time": "2018-08-13T06:01:25.797654Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"_________________________________________________________________\n",
"Layer (type) Output Shape Param # \n",
"=================================================================\n",
"conv2d_3 (Conv2D) (None, 26, 26, 10) 100 \n",
"_________________________________________________________________\n",
"max_pooling2d_1 (MaxPooling2 (None, 13, 13, 10) 0 \n",
"_________________________________________________________________\n",
"conv2d_4 (Conv2D) (None, 11, 11, 10) 910 \n",
"_________________________________________________________________\n",
"max_pooling2d_2 (MaxPooling2 (None, 5, 5, 10) 0 \n",
"_________________________________________________________________\n",
"flatten_2 (Flatten) (None, 250) 0 \n",
"_________________________________________________________________\n",
"dense_11 (Dense) (None, 3) 753 \n",
"=================================================================\n",
"Total params: 1,763\n",
"Trainable params: 1,763\n",
"Non-trainable params: 0\n",
"_________________________________________________________________\n"
]
}
],
"source": [
"# pooling reduces the field of input\n",
"cnn_pool.summary()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Ariel has: Segmentation network from CNNs\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Use dropout to silence few units every time for less overfitting on training"
]
},
{
"cell_type": "code",
"execution_count": 26,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T06:02:08.326842",
"start_time": "2018-08-13T06:02:08.324222Z"
}
},
"outputs": [],
"source": [
"from keras.layers import MaxPooling2D, Dropout"
]
},
{
"cell_type": "code",
"execution_count": 27,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T06:02:37.466724",
"start_time": "2018-08-13T06:02:33.822929Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Train on 540 samples, validate on 60 samples\n",
"Epoch 1/10\n",
"540/540 [==============================] - 1s 1ms/step - loss: 0.3716 - acc: 0.8537 - val_loss: 0.2375 - val_acc: 0.9167\n",
"Epoch 2/10\n",
"540/540 [==============================] - 0s 547us/step - loss: 0.1398 - acc: 0.9463 - val_loss: 0.1949 - val_acc: 0.9167\n",
"Epoch 3/10\n",
"540/540 [==============================] - 0s 546us/step - loss: 0.1448 - acc: 0.9463 - val_loss: 0.1728 - val_acc: 0.9500\n",
"Epoch 4/10\n",
"540/540 [==============================] - 0s 549us/step - loss: 0.1227 - acc: 0.9481 - val_loss: 0.1925 - val_acc: 0.9000\n",
"Epoch 5/10\n",
"540/540 [==============================] - 0s 564us/step - loss: 0.1013 - acc: 0.9556 - val_loss: 0.2489 - val_acc: 0.9167\n",
"Epoch 6/10\n",
"540/540 [==============================] - 0s 549us/step - loss: 0.0931 - acc: 0.9704 - val_loss: 0.1529 - val_acc: 0.9333\n",
"Epoch 7/10\n",
"540/540 [==============================] - 0s 551us/step - loss: 0.0786 - acc: 0.9574 - val_loss: 0.2000 - val_acc: 0.9333\n",
"Epoch 8/10\n",
"540/540 [==============================] - 0s 562us/step - loss: 0.0747 - acc: 0.9685 - val_loss: 0.2328 - val_acc: 0.9167\n",
"Epoch 9/10\n",
"540/540 [==============================] - 0s 539us/step - loss: 0.0616 - acc: 0.9741 - val_loss: 0.2282 - val_acc: 0.9167\n",
"Epoch 10/10\n",
"540/540 [==============================] - 0s 553us/step - loss: 0.0791 - acc: 0.9741 - val_loss: 0.1887 - val_acc: 0.9167\n"
]
}
],
"source": [
"cnn_pool = Sequential()\n",
"cnn_pool.add(Conv2D(10, kernel_size=3, # 3x3 pixel \n",
" input_shape=(train_data.shape[1],train_data.shape[2],train_data.shape[3]),\n",
" activation='relu'))\n",
"# add a dropout layer\n",
"cnn_pool.add(Dropout(0.2)) # 20 percent weights ignored. \n",
"cnn_pool.add(MaxPool2D(2, #max pool over 2x2 pixels\n",
" ))\n",
"cnn_pool.add(Conv2D(10, kernel_size = 3, activation='relu'))\n",
"cnn_pool.add(MaxPool2D(2, #max pool over 2x2 pixels\n",
" ))\n",
"cnn_pool.add(Flatten())\n",
"cnn_pool.add(Dense(3,activation='softmax'))\n",
"cnn_pool.compile(loss='categorical_crossentropy', \n",
" optimizer=opt,\n",
" metrics=['accuracy'])\n",
"training = cnn_pool.fit(train_data,train_labels,epochs=10,batch_size=10,validation_split=.1)"
]
},
{
"cell_type": "code",
"execution_count": 29,
"metadata": {
"ExecuteTime": {
"end_time": "2018-08-13T06:03:09.108531",
"start_time": "2018-08-13T06:03:08.976040Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"[<matplotlib.lines.Line2D at 0x1166c0978>]"
]
},
"execution_count": 29,
"metadata": {},
"output_type": "execute_result"
},
{
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD8CAYAAACb4nSYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRv4f8OfOLZXhGLoAAAAASUVORK5CYII=\n",
"text/plain": [
"<matplotlib.figure.Figure at 0x1166c0128>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# Plot loss function on training and validation \n",
"plt.plot(training.history['loss'],'o--')\n",
"plt.plot(training.history['val_loss'],'o-')"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# visualize model"
]
},
{
"cell_type": "code",
"execution_count": 167,
"metadata": {},
"outputs": [],
"source": [
"from keras.utils import plot_model\n",
"plot_model(dense, to_file='model.png')"
]
}
],
"metadata": {
"anaconda-cloud": {},
"kernelspec": {
"display_name": "Python [conda env:py36]",
"language": "python",
"name": "conda-env-py36-py"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.6.1"
},
"toc": {
"nav_menu": {
"height": "120px",
"width": "252px"
},
"navigate_menu": true,
"number_sections": true,
"sideBar": true,
"threshold": 4,
"toc_cell": false,
"toc_section_display": "block",
"toc_window_display": false
}
},
"nbformat": 4,
"nbformat_minor": 2
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment