Iceberg Recognition from Satellite Images using Kears Deep NN (aribis…

…369#57) * Create spam_data.csv * Add files via upload * Delete spam_data.csv * Add files via upload * Rename Spam_filter (1).ipynb to Spam_filter.ipynb * Create Spam_Filter * Delete Spam_Filter * Add files via upload * Rename MNIST.ipynb to Digit_Recognition.ipynb * Delete Digit_Recognition.ipynb * Add files via upload * Add files via upload * Rename BiomechanicalPatients.csv.csv to examples-pytorch/BiomechanicalPatients.csv.csv * Rename Pytorch_Tutorial_1.ipynb to examples-pytorch/Pytorch_Tutorial_1.ipynb * Rename BiomechanicalPatients.csv.csv to BiomechanicalPatients.csv * Add Pytorch_Tutorial_2_nn.ipynb * Added Pytorch_CNN.ipynb * Delete Pytorch_CNN.ipynb * Add Pytorch_CNN_updated.ipynb * Delete Pytorch_CNN_updated.ipynb * Add Pytorch_CNN_Tutorial.ipynb * Add Hand_Sign_Pytorch.ipynb * Add Keras_Iceberg.ipynb * Move Keras_Iceberg.ipnb to examples-keras
kamrul3000 · Dec 19, 2017 · 298a819 · 298a819
1 parent cbf4629
commit 298a819
Showing 1 changed file with 302 additions and 0 deletions.
diff --git a/examples-keras/Keras_Iceberg.ipynb b/examples-keras/Keras_Iceberg.ipynb
@@ -0,0 +1,302 @@
+{
+  "cells": [
+    {
+      "metadata": {
+        "_uuid": "a1b81d4cb2e75a6e2ef83433ca3fd86786a7d072"
+      },
+      "cell_type": "markdown",
+      "source": "# Iceberg Recognition from Salelite Images Using Keras Deep NN"
+    },
+    {
+      "metadata": {
+        "_uuid": "a728862ccf216b4c8e43178e91a22653206780c8"
+      },
+      "cell_type": "markdown",
+      "source": "In this notebook I am going to develop a Deep Neural Network to recognise Iceberg in ocean from satelite images . For this image recognition problem I am going to use Keras deep learning features . Keras is high level deep NN library which can be built over a tensorflow or theano backend . Keras provides much more abstraction and flexibilty to built deep networks in a very simple way . "
+    },
+    {
+      "metadata": {
+        "_cell_guid": "4f18337e-24ff-479b-9983-c462cf195701",
+        "_uuid": "84e8fb50043642e4df9025319e87d0d54c405418",
+        "trusted": true
+      },
+      "cell_type": "code",
+      "source": "# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load in \n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the \"../input/\" directory.\n# For example, running this (by clicking run or pressing Shift+Enter) will list the files in the input directory\n\nfrom subprocess import check_output\nprint(check_output([\"ls\", \"../input\"]).decode(\"utf8\"))\n\n# Any results you write to the current directory are saved as output.",
+      "execution_count": 1,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "text": "sample_submission.csv\ntest.json\ntrain.json\n\n",
+          "name": "stdout"
+        }
+      ]
+    },
+    {
+      "metadata": {
+        "_uuid": "9ad1ccdab3de414f4e42520493646172c7f30157"
+      },
+      "cell_type": "markdown",
+      "source": "The data for this problem is a image data in json format . We have to read the json data using pd.read_json( ) and we see it has two bands or channels and 1604 images in the training data .  The data can be downloaded form this [link](https://www.kaggle.com/c/statoil-iceberg-classifier-challenge/data) .  "
+    },
+    {
+      "metadata": {
+        "_cell_guid": "cfc162c8-8dc7-44ec-b50b-8f7cf6e737b1",
+        "_uuid": "c8c595a2cdfba21f1617a0116bc42babf4de3370",
+        "trusted": true
+      },
+      "cell_type": "code",
+      "source": "train_df = pd.read_json('../input/train.json')\ntest_df = pd.read_json('../input/test.json')\ntrain_df.head()",
+      "execution_count": 2,
+      "outputs": [
+        {
+          "output_type": "execute_result",
+          "execution_count": 2,
+          "data": {
+            "text/plain": "                                              band_1  \\\n0  [-27.878360999999998, -27.15416, -28.668615, -...   \n1  [-12.242375, -14.920304999999999, -14.920363, ...   \n2  [-24.603676, -24.603714, -24.871029, -23.15277...   \n3  [-22.454607, -23.082819, -23.998013, -23.99805...   \n4  [-26.006956, -23.164886, -23.164886, -26.89116...   \n\n                                              band_2        id inc_angle  \\\n0  [-27.154118, -29.537888, -31.0306, -32.190483,...  dfd5f913   43.9239   \n1  [-31.506321, -27.984554, -26.645678, -23.76760...  e25388fd   38.1562   \n2  [-24.870956, -24.092632, -20.653963, -19.41104...  58b2aaa0   45.2859   \n3  [-27.889421, -27.519794, -27.165262, -29.10350...  4cfc3a18   43.8306   \n4  [-27.206915, -30.259186, -30.259186, -23.16495...  271f93f4   35.6256   \n\n   is_iceberg  \n0           0  \n1           0  \n2           1  \n3           0  \n4           0  ",
+            "text/html": "<div>\n<style>\n    .dataframe thead tr:only-child th {\n        text-align: right;\n    }\n\n    .dataframe thead th {\n        text-align: left;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>band_1</th>\n      <th>band_2</th>\n      <th>id</th>\n      <th>inc_angle</th>\n      <th>is_iceberg</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>[-27.878360999999998, -27.15416, -28.668615, -...</td>\n      <td>[-27.154118, -29.537888, -31.0306, -32.190483,...</td>\n      <td>dfd5f913</td>\n      <td>43.9239</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>[-12.242375, -14.920304999999999, -14.920363, ...</td>\n      <td>[-31.506321, -27.984554, -26.645678, -23.76760...</td>\n      <td>e25388fd</td>\n      <td>38.1562</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>[-24.603676, -24.603714, -24.871029, -23.15277...</td>\n      <td>[-24.870956, -24.092632, -20.653963, -19.41104...</td>\n      <td>58b2aaa0</td>\n      <td>45.2859</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>[-22.454607, -23.082819, -23.998013, -23.99805...</td>\n      <td>[-27.889421, -27.519794, -27.165262, -29.10350...</td>\n      <td>4cfc3a18</td>\n      <td>43.8306</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>[-26.006956, -23.164886, -23.164886, -26.89116...</td>\n      <td>[-27.206915, -30.259186, -30.259186, -23.16495...</td>\n      <td>271f93f4</td>\n      <td>35.6256</td>\n      <td>0</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
+          },
+          "metadata": {}
+        }
+      ]
+    },
+    {
+      "metadata": {
+        "_uuid": "10a60a58d768649672587a8e17cf1a99461f892a"
+      },
+      "cell_type": "markdown",
+      "source": "We need to preprocess the data to reshape it to make it in shape (1604,75,75,2) . We get the the training data in X_band ."
+    },
+    {
+      "metadata": {
+        "_cell_guid": "c0cf60e9-f6fa-46a8-b209-584cd17e58e3",
+        "_uuid": "9448794c7a226f1c61c75b31178121782babb47d",
+        "trusted": true,
+        "collapsed": true
+      },
+      "cell_type": "code",
+      "source": "X_band_1=np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in train_df[\"band_1\"]])\nX_band_2=np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in train_df[\"band_2\"]])",
+      "execution_count": 3,
+      "outputs": []
+    },
+    {
+      "metadata": {
+        "_cell_guid": "1359c2ae-d42c-4071-bbc0-077a0a1276e0",
+        "_uuid": "06e801c07554537979ffc9648de9da34f5bb77c6",
+        "trusted": true,
+        "collapsed": true
+      },
+      "cell_type": "code",
+      "source": "X_band = np.zeros([1604,75,75,2])\nfor t in range(1604):\n    X_band[t,:,:,0] = X_band_1[t]\n    X_band[t,:,:,1] = X_band_2[t]",
+      "execution_count": 4,
+      "outputs": []
+    },
+    {
+      "metadata": {
+        "_uuid": "7055906ebdc1fe1cc5a8858b7f5be07554f1fd70"
+      },
+      "cell_type": "markdown",
+      "source": "Now we import all required keras packages . "
+    },
+    {
+      "metadata": {
+        "_cell_guid": "8e29480a-875f-41ab-8c33-baf597221998",
+        "_uuid": "e1cc3d6f0cd75b6b715f66901f20a7a0e9beb186",
+        "trusted": true
+      },
+      "cell_type": "code",
+      "source": "from keras import layers\nfrom keras.layers import Input, Dense, Activation, ZeroPadding2D, BatchNormalization, Flatten, Conv2D\nfrom keras.layers import AveragePooling2D, MaxPooling2D, Dropout, GlobalMaxPooling2D, GlobalAveragePooling2D\nfrom keras.models import Model\nfrom keras.preprocessing import image\nfrom keras.utils import layer_utils\nfrom keras.utils.data_utils import get_file\nfrom keras.applications.imagenet_utils import preprocess_input\n\nfrom IPython.display import SVG\nfrom keras.utils.vis_utils import model_to_dot\nfrom keras.utils import plot_model\n\nimport keras.backend as K\nK.set_image_data_format('channels_last')\nimport matplotlib.pyplot as plt\nfrom matplotlib.pyplot import imshow",
+      "execution_count": 5,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "text": "Using TensorFlow backend.\n/opt/conda/lib/python3.6/importlib/_bootstrap.py:219: RuntimeWarning: compiletime version 3.5 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.6\n  return f(*args, **kwds)\n",
+          "name": "stderr"
+        }
+      ]
+    },
+    {
+      "metadata": {
+        "_uuid": "95ef7d7e362fffffaacebafebcf70a662650a726"
+      },
+      "cell_type": "markdown",
+      "source": "The deep NN model in keras can be built in 2 ways . One is using Sequential API and the other is Model API . I am using Model API . So we built the model thatconsistes of 3 layers of convolution layer and two fully connected layer . Each conv layer is made of (Conv2D,BatchNormalization,ReLU,MaxPooling2D) block. "
+    },
+    {
+      "metadata": {
+        "_cell_guid": "cfc32109-12f5-471f-8fad-462cd91ee1ea",
+        "_uuid": "7eb2c775a705d742e8d27b8d34b29eca2583aa3c",
+        "collapsed": true,
+        "trusted": true
+      },
+      "cell_type": "code",
+      "source": "def Iceberg_model(input_shape):\n    X_in = Input(input_shape)\n    \n    X = Conv2D(10,kernel_size=(5,5),input_shape=(75,75,2))(X_in)\n    X = BatchNormalization()(X)\n    X = Activation('relu')(X)\n    X = MaxPooling2D(pool_size=(2,2))(X)\n    \n    X = Conv2D(10,kernel_size=(5,5))(X)\n    X = BatchNormalization()(X)\n    X = Activation('relu')(X)\n    X = MaxPooling2D(pool_size=(2,2))(X)\n    \n    X = Conv2D(10,kernel_size=(5,5))(X)\n    X = BatchNormalization()(X)\n    X = Activation('relu')(X)\n    X = MaxPooling2D(pool_size=(2,2))(X)\n    \n    X = Flatten()(X)\n    X = Dense(50)(X)\n    X = Activation('relu')(X)\n    \n    X = Dense(1)(X)\n    X = Activation('sigmoid')(X)\n    \n    model = Model(inputs=X_in,outputs=X,name='Iceberg_model')\n    return model",
+      "execution_count": 6,
+      "outputs": []
+    },
+    {
+      "metadata": {
+        "_uuid": "b1d8e2c16c34c34b6252503e7afab3fb9cd083d2"
+      },
+      "cell_type": "markdown",
+      "source": "In Keras we have follow 4 simple steps to built a NN model . \n1.  create the model .\n2.  compile the model .\n3.  fit the model on the training data. \n4.  evaluate the model on the test data . "
+    },
+    {
+      "metadata": {
+        "_uuid": "be0e3034710115db7a088842375f1735f9a813b5"
+      },
+      "cell_type": "markdown",
+      "source": "For  creating the model we provide the shape of each training data (75,75,2) ."
+    },
+    {
+      "metadata": {
+        "_cell_guid": "5f738ef2-a41b-49d7-a462-6ed8d4f4005f",
+        "_uuid": "7ae316e5773a2ac422d73e6e38e08080528fc443",
+        "trusted": true,
+        "collapsed": true
+      },
+      "cell_type": "code",
+      "source": "IcebergModel = Iceberg_model((75,75,2))",
+      "execution_count": 15,
+      "outputs": []
+    },
+    {
+      "metadata": {
+        "_uuid": "858018401b56401390458f143ed6508b462e34e7"
+      },
+      "cell_type": "markdown",
+      "source": "Compile the model using 'adam' optimizer , 'binary_crossentropy' loss and accuracy metric ."
+    },
+    {
+      "metadata": {
+        "_cell_guid": "c263a7ab-8cfd-47ca-993f-13041f44e925",
+        "_uuid": "ce1bdee690b623e2ffe95b5a67bea56745ba8caf",
+        "collapsed": true,
+        "trusted": true
+      },
+      "cell_type": "code",
+      "source": "IcebergModel.compile(optimizer='adam',loss='binary_crossentropy',metrics=['accuracy'])",
+      "execution_count": 16,
+      "outputs": []
+    },
+    {
+      "metadata": {
+        "_uuid": "cafe997e2d89ea1ff13b22a63f10dd31117cae83"
+      },
+      "cell_type": "markdown",
+      "source": "Now we fit the model with the training data having epochs = 20 and random batch_size = 128 . "
+    },
+    {
+      "metadata": {
+        "_cell_guid": "09cd8290-8bc4-4568-b366-03286d2bdc47",
+        "_uuid": "6c45909f5afa43d1f791f76ac9bdb03a5a31ad7b",
+        "trusted": true
+      },
+      "cell_type": "code",
+      "source": "target = train_df['is_iceberg'].values\nIcebergModel.fit(x=X_band,y=target,epochs=20,batch_size=128)",
+      "execution_count": 17,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "text": "Epoch 1/20\n1604/1604 [==============================] - 13s 8ms/step - loss: 0.6654 - acc: 0.6241\nEpoch 2/20\n1604/1604 [==============================] - 12s 7ms/step - loss: 0.4797 - acc: 0.7550\nEpoch 3/20\n1604/1604 [==============================] - 11s 7ms/step - loss: 0.3634 - acc: 0.8416\nEpoch 4/20\n1604/1604 [==============================] - 11s 7ms/step - loss: 0.3080 - acc: 0.8622\nEpoch 5/20\n1604/1604 [==============================] - 12s 7ms/step - loss: 0.2684 - acc: 0.8840\nEpoch 6/20\n1604/1604 [==============================] - 11s 7ms/step - loss: 0.2278 - acc: 0.9146\nEpoch 7/20\n1604/1604 [==============================] - 12s 7ms/step - loss: 0.1981 - acc: 0.9296\nEpoch 8/20\n1604/1604 [==============================] - 12s 7ms/step - loss: 0.1671 - acc: 0.9458\nEpoch 9/20\n1604/1604 [==============================] - 12s 7ms/step - loss: 0.1536 - acc: 0.9445\nEpoch 10/20\n1604/1604 [==============================] - 12s 7ms/step - loss: 0.1179 - acc: 0.9688\nEpoch 11/20\n1604/1604 [==============================] - 12s 8ms/step - loss: 0.1008 - acc: 0.9713\nEpoch 12/20\n1604/1604 [==============================] - 12s 7ms/step - loss: 0.0986 - acc: 0.9782\nEpoch 13/20\n1604/1604 [==============================] - 11s 7ms/step - loss: 0.0722 - acc: 0.9869\nEpoch 14/20\n1604/1604 [==============================] - 12s 7ms/step - loss: 0.0492 - acc: 0.9963\nEpoch 15/20\n1604/1604 [==============================] - 11s 7ms/step - loss: 0.0361 - acc: 0.9988\nEpoch 16/20\n1604/1604 [==============================] - 13s 8ms/step - loss: 0.0293 - acc: 0.9988\nEpoch 17/20\n1604/1604 [==============================] - 15s 9ms/step - loss: 0.0219 - acc: 1.0000\nEpoch 18/20\n1604/1604 [==============================] - 13s 8ms/step - loss: 0.0166 - acc: 1.0000\nEpoch 19/20\n1604/1604 [==============================] - 12s 7ms/step - loss: 0.0132 - acc: 1.0000\nEpoch 20/20\n1604/1604 [==============================] - 12s 7ms/step - loss: 0.0110 - acc: 1.0000\n",
+          "name": "stdout"
+        },
+        {
+          "output_type": "execute_result",
+          "execution_count": 17,
+          "data": {
+            "text/plain": "<keras.callbacks.History at 0x7fc375d37da0>"
+          },
+          "metadata": {}
+        }
+      ]
+    },
+    {
+      "metadata": {
+        "_uuid": "b5066ec82f92036add6e54bf9a64442c3b2ca513"
+      },
+      "cell_type": "markdown",
+      "source": "Evaluating the model on the traning data and it gets a 96.5 % accuarcy . "
+    },
+    {
+      "metadata": {
+        "_cell_guid": "cbbb7dd6-4e55-4eac-931a-91e7ceea04d1",
+        "_uuid": "2cb5a693c5c4d90f265aafec81cf2f5b91386512",
+        "trusted": true
+      },
+      "cell_type": "code",
+      "source": "IcebergModel.evaluate(x=X_band,y=target)",
+      "execution_count": 18,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "text": "1604/1604 [==============================] - 6s 4ms/step\n",
+          "name": "stdout"
+        },
+        {
+          "output_type": "execute_result",
+          "execution_count": 18,
+          "data": {
+            "text/plain": "[0.090052791218507913, 0.96508728179551118]"
+          },
+          "metadata": {}
+        }
+      ]
+    },
+    {
+      "metadata": {
+        "_uuid": "73b609c2f543f0893ab9f2b82ed52c6950b5c485"
+      },
+      "cell_type": "markdown",
+      "source": "Classification report on the training data .  "
+    },
+    {
+      "metadata": {
+        "trusted": true,
+        "_uuid": "2893f9043ead5d440d43e1e67fc028d3949a562d"
+      },
+      "cell_type": "code",
+      "source": "from sklearn.metrics import classification_report\npred_label = IcebergModel.predict(x=X_band)\npred_label[pred_label>0.5]=1\npred_label[pred_label<=0.5]=0\nprint(classification_report(target,pred_label))",
+      "execution_count": 25,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "text": "             precision    recall  f1-score   support\n\n          0       1.00      0.93      0.97       851\n          1       0.93      1.00      0.96       753\n\navg / total       0.97      0.97      0.97      1604\n\n",
+          "name": "stdout"
+        }
+      ]
+    },
+    {
+      "metadata": {
+        "_cell_guid": "0a55a653-c526-4b4f-b477-66d03ce2dfc7",
+        "_uuid": "9ca13c023d0c9846fc25b30ae80c2b9e1bc5d8dd",
+        "collapsed": true,
+        "trusted": true
+      },
+      "cell_type": "code",
+      "source": "",
+      "execution_count": null,
+      "outputs": []
+    }
+  ],
+  "metadata": {
+    "language_info": {
+      "name": "python",
+      "version": "3.6.3",
+      "mimetype": "text/x-python",
+      "codemirror_mode": {
+        "name": "ipython",
+        "version": 3
+      },
+      "pygments_lexer": "ipython3",
+      "nbconvert_exporter": "python",
+      "file_extension": ".py"
+    },
+    "kernelspec": {
+      "display_name": "Python 3",
+      "language": "python",
+      "name": "python3"
+    }
+  },
+  "nbformat": 4,
+  "nbformat_minor": 1
+}