RGB_image_prob_mergedarray.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "mlproof/tifffile.py:1995: UserWarning: failed to import _tifffile.decodepackbits\n",
      "  warnings.warn(\"failed to import %s\" % module_function)\n",
      "mlproof/tifffile.py:1995: UserWarning: failed to import _tifffile.decodelzw\n",
      "  warnings.warn(\"failed to import %s\" % module_function)\n",
      "mlproof/tifffile.py:1995: UserWarning: failed to import _tifffile.unpackints\n",
      "  warnings.warn(\"failed to import %s\" % module_function)\n",
      "Using gpu device 0: GeForce GTX TITAN (CNMeM is disabled, CuDNN 4007)\n",
      "/home/d/nolearn/local/lib/python2.7/site-packages/theano/tensor/signal/downsample.py:6: UserWarning: downsample module has been moved to the theano.tensor.signal.pool module.\n",
      "  \"downsample module has been moved to the theano.tensor.signal.pool module.\")\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "import mlproof as mlp\n",
    "import mlproof.nets as nets\n",
    "\n",
    "import numpy as np\n",
    "\n",
    "from nolearn.lasagne.visualize import plot_loss\n",
    "from nolearn.lasagne.visualize import plot_conv_weights\n",
    "from nolearn.lasagne.visualize import plot_conv_activity\n",
    "from nolearn.lasagne.visualize import plot_occlusion\n",
    "\n",
    "from matplotlib.pyplot import imshow\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CNN configuration: \n",
      "    Our CNN with image, prob, merged_array as RGB.\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "cnn = nets.RGBNet()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded /home/d/patches//cylinder_small1_rgba/ in 0.000605821609497 seconds.\n"
     ]
    }
   ],
   "source": [
    "X_train, y_train, X_test, y_test = mlp.Patch.load('cylinder_small1_rgba')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# Neural Network with 1547826 learnable parameters\n",
      "\n",
      "## Layer information\n",
      "\n",
      "  #  name      size\n",
      "---  --------  --------\n",
      "  0  input     3x75x75\n",
      "  1  conv1     64x63x63\n",
      "  2  pool1     64x31x31\n",
      "  3  conv2     48x19x19\n",
      "  4  pool2     48x9x9\n",
      "  5  hidden3   256\n",
      "  6  dropout3  256\n",
      "  7  output    2\n",
      "\n",
      "  epoch    train loss    valid loss    train/val    valid acc  dur\n",
      "-------  ------------  ------------  -----------  -----------  ------\n",
      "      1       \u001b[36m0.64773\u001b[0m       \u001b[32m0.65070\u001b[0m      0.99543      0.60630  15.67s\n",
      "      2       0.70791       0.67436      1.04974      0.50199  15.87s\n",
      "      3       \u001b[36m0.64677\u001b[0m       \u001b[32m0.58421\u001b[0m      1.10708      0.67623  16.47s\n",
      "      4       \u001b[36m0.49718\u001b[0m       0.61790      0.80463      0.77543  16.44s\n",
      "      5       0.51467       \u001b[32m0.47812\u001b[0m      1.07644      0.79663  16.44s\n",
      "      6       \u001b[36m0.41735\u001b[0m       \u001b[32m0.36688\u001b[0m      1.13757      0.85098  16.44s\n",
      "      7       \u001b[36m0.35540\u001b[0m       \u001b[32m0.34647\u001b[0m      1.02575      0.85301  16.45s\n",
      "      8       \u001b[36m0.29805\u001b[0m       \u001b[32m0.27239\u001b[0m      1.09421      0.88616  16.45s\n",
      "      9       \u001b[36m0.23702\u001b[0m       \u001b[32m0.24381\u001b[0m      0.97215      0.90645  16.47s\n",
      "     10       \u001b[36m0.19933\u001b[0m       \u001b[32m0.18418\u001b[0m      1.08228      0.93446  16.48s\n",
      "     11       \u001b[36m0.15782\u001b[0m       \u001b[32m0.16468\u001b[0m      0.95839      0.94634  16.45s\n",
      "     12       \u001b[36m0.12737\u001b[0m       \u001b[32m0.13567\u001b[0m      0.93884      0.95609  16.46s\n",
      "     13       \u001b[36m0.09952\u001b[0m       \u001b[32m0.12058\u001b[0m      0.82535      0.96540  16.47s\n",
      "     14       \u001b[36m0.08472\u001b[0m       \u001b[32m0.11809\u001b[0m      0.71742      0.95699  16.47s\n",
      "     15       \u001b[36m0.07161\u001b[0m       \u001b[32m0.08692\u001b[0m      0.82393      0.97145  16.44s\n",
      "     16       \u001b[36m0.05502\u001b[0m       \u001b[32m0.08046\u001b[0m      0.68381      0.97652  16.48s\n",
      "     17       \u001b[36m0.04458\u001b[0m       0.08068      0.55249      0.98152  10.61s\n",
      "     18       \u001b[36m0.03738\u001b[0m       0.10482      0.35663      0.97388  8.46s\n",
      "     19       \u001b[36m0.03263\u001b[0m       0.08573      0.38056      0.97902  7.76s\n",
      "     20       \u001b[36m0.03196\u001b[0m       \u001b[32m0.07418\u001b[0m      0.43076      0.98076  7.76s\n",
      "     21       \u001b[36m0.02277\u001b[0m       0.09111      0.24989      0.98076  7.76s\n",
      "     22       0.02763       0.08252      0.33483      0.98069  7.78s\n",
      "     23       \u001b[36m0.02098\u001b[0m       0.10571      0.19849      0.98069  7.76s\n",
      "     24       \u001b[36m0.02068\u001b[0m       \u001b[32m0.07395\u001b[0m      0.27965      0.98236  7.76s\n",
      "     25       \u001b[36m0.01654\u001b[0m       0.09244      0.17896      0.97721  7.76s\n",
      "     26       0.02640       0.07977      0.33098      0.98319  7.76s\n",
      "     27       0.01884       0.09094      0.20711      0.97471  7.77s\n",
      "     28       \u001b[36m0.01526\u001b[0m       \u001b[32m0.06694\u001b[0m      0.22791      0.98493  7.76s\n",
      "     29       \u001b[36m0.01367\u001b[0m       0.08346      0.16375      0.98493  7.76s\n",
      "     30       \u001b[36m0.01339\u001b[0m       0.08211      0.16308      0.98152  7.77s\n",
      "     31       \u001b[36m0.01030\u001b[0m       0.07063      0.14583      0.98486  7.76s\n",
      "     32       0.01111       0.07247      0.15336      0.98659  7.81s\n",
      "     33       \u001b[36m0.00876\u001b[0m       0.08002      0.10953      0.98493  7.81s\n",
      "     34       \u001b[36m0.00714\u001b[0m       0.08683      0.08225      0.98319  7.80s\n",
      "     35       \u001b[36m0.00669\u001b[0m       0.11505      0.05813      0.97895  7.81s\n",
      "     36       0.00750       0.11973      0.06262      0.97978  7.78s\n",
      "     37       0.00852       0.10673      0.07986      0.98402  7.76s\n",
      "     38       0.00851       0.08854      0.09613      0.98743  7.76s\n",
      "     39       \u001b[36m0.00518\u001b[0m       0.10385      0.04989      0.98486  7.76s\n",
      "     40       0.00708       0.08805      0.08044      0.98569  7.76s\n",
      "     41       0.00733       0.09265      0.07910      0.98236  7.81s\n",
      "     42       0.00738       0.09864      0.07484      0.98402  7.82s\n",
      "     43       0.00625       0.07928      0.07882      0.98576  7.80s\n",
      "     44       \u001b[36m0.00454\u001b[0m       0.09020      0.05031      0.98409  7.83s\n",
      "     45       0.00522       0.08340      0.06259      0.98576  7.81s\n",
      "     46       \u001b[36m0.00428\u001b[0m       0.07827      0.05467      0.98576  7.83s\n",
      "     47       0.00527       0.09142      0.05768      0.98743  7.77s\n",
      "     48       0.00464       0.08648      0.05370      0.98576  7.79s\n",
      "     49       \u001b[36m0.00360\u001b[0m       0.07405      0.04865      0.98743  7.83s\n",
      "     50       0.00624       0.10513      0.05934      0.98409  7.76s\n",
      "     51       0.00594       0.12340      0.04810      0.98326  7.76s\n",
      "     52       0.00409       0.08396      0.04875      0.98667  7.76s\n",
      "     53       0.00547       0.10369      0.05279      0.98569  7.76s\n",
      "     54       0.00403       0.10415      0.03867      0.98576  7.76s\n",
      "     55       \u001b[36m0.00347\u001b[0m       0.08631      0.04018      0.98833  7.76s\n",
      "     56       \u001b[36m0.00332\u001b[0m       0.09373      0.03537      0.98659  7.76s\n",
      "     57       \u001b[36m0.00269\u001b[0m       0.08542      0.03144      0.98833  7.76s\n",
      "     58       0.00278       0.09138      0.03044      0.98833  7.77s\n",
      "     59       0.00519       0.09574      0.05418      0.98833  7.78s\n",
      "     60       0.00771       0.10701      0.07204      0.98569  7.77s\n",
      "     61       0.01246       0.11151      0.11176      0.98159  7.78s\n",
      "     62       0.01254       0.10114      0.12401      0.98667  7.76s\n",
      "     63       0.01085       0.11013      0.09855      0.98326  7.79s\n",
      "     64       0.00869       0.08815      0.09862      0.97819  7.78s\n",
      "     65       0.00958       0.06711      0.14275      0.98659  7.76s\n",
      "     66       0.00937       0.09924      0.09445      0.97902  7.76s\n",
      "     67       0.00603       0.10937      0.05511      0.98167  7.76s\n",
      "     68       0.01014       0.09338      0.10857      0.98228  7.79s\n",
      "     69       0.00784       0.08276      0.09472      0.98833  7.79s\n",
      "     70       0.00769       0.09641      0.07976      0.98326  7.77s\n",
      "     71       0.00815       0.10228      0.07967      0.98319  7.80s\n",
      "     72       0.00680       0.09796      0.06946      0.98152  7.77s\n",
      "     73       0.00559       0.09691      0.05768      0.98152  7.76s\n",
      "     74       0.00502       0.07481      0.06705      0.98319  7.79s\n",
      "     75       0.00274       0.07636      0.03585      0.98493  7.80s\n",
      "     76       0.00434       0.08102      0.05353      0.98659  7.81s\n",
      "     77       0.00455       0.10793      0.04212      0.98326  7.80s\n",
      "     78       0.00329       0.09035      0.03642      0.98319  7.76s\n",
      "Early stopping.\n",
      "Best valid loss was 0.066942 at epoch 28.\n",
      "Loaded parameters to layer 'conv1' (shape 64x3x13x13).\n",
      "Loaded parameters to layer 'conv1' (shape 64).\n",
      "Loaded parameters to layer 'conv2' (shape 48x64x13x13).\n",
      "Loaded parameters to layer 'conv2' (shape 48).\n",
      "Loaded parameters to layer 'hidden3' (shape 3888x256).\n",
      "Loaded parameters to layer 'hidden3' (shape 256).\n",
      "Loaded parameters to layer 'output' (shape 256x2).\n",
      "Loaded parameters to layer 'output' (shape 2).\n"
     ]
    }
   ],
   "source": [
    "cnn = cnn.fit(X_train['rgba'][:,:-1,:,:], y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7898773006134969"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# this is the score on cylinder_small1 with regularization and patience counter 50 and data augmentation\n",
    "cnn.score(X_test['rgba'][:,:-1,:,:], y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<module 'matplotlib.pyplot' from '/home/d/nolearn/local/lib/python2.7/site-packages/matplotlib/pyplot.pyc'>"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYYAAAEPCAYAAABGP2P1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f68aae7c850>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# this is the plot on cylinder_small1 with regularization and patience counter 50 and data augmentation\n",
    "plot_loss(cnn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}