siamese-nn-Recognize icebergs and ships_Solution.ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"name":"siamese-nn-Recognize icebergs and ships_Solution.ipynb","version":"0.3.2","provenance":[],"collapsed_sections":[]},"language_info":{"codemirror_mode":{"name":"ipython","version":3},"file_extension":".py","mimetype":"text/x-python","name":"python","nbconvert_exporter":"python","pygments_lexer":"ipython3","version":"3.6.8"},"kernelspec":{"name":"python3","display_name":"Python 3"},"accelerator":"GPU"},"cells":[{"cell_type":"markdown","metadata":{"_cell_guid":"608e43e3-8699-4299-9d18-9e9a6a3dd9bf","_uuid":"434ffd54c774edd4e8b851ab0a4b6675425e46da","id":"PiXnTZq6i1_j","colab_type":"text"},"source":["# Siamese Neural Networks\n","Recognize images of icebergs from images of ships. \n","These images were taken from space with a Sentinel-4 satellite. The goal was to create the most accurate neural network to differentiate them. However, neural networks usually only work well with A LOT of data, and I constrained you guys to 1000 examples to train on. That seems like a lot, but a CONVENTIONAL neural network needs more data. However, we are going to make a special kind of neural network: *Siamese Network*. This special network is used for face recognition and few-shot learning(learning from few examples). Let's get started by importing the usual stuff:"]},{"cell_type":"code","metadata":{"_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","id":"OGQZpMdYi1_k","colab_type":"code","colab":{}},"source":["from keras.layers import Input, Conv2D, Lambda, merge, Dense, Flatten,MaxPooling2D,Activation, Dropout\n","from keras.models import Model, Sequential\n","from keras.regularizers import l2\n","from keras import backend as K\n","from keras.optimizers import Adam\n","from skimage.io import imshow\n","import matplotlib.pyplot as plt\n","import numpy as np\n","import pandas as pd\n","import random"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"_cell_guid":"adbffb0a-1784-4595-b3f2-2ea930abc31f","_uuid":"b9200b733a896e4b6f261122fe9134c00955cc91","id":"9MQX4H9Di1_n","colab_type":"text"},"source":["## Load the Data\n"]},{"cell_type":"code","metadata":{"id":"jF4NMzrd6a1U","colab_type":"code","outputId":"3ad49feb-3637-4e86-e7fb-442fb243ad7f","executionInfo":{"status":"ok","timestamp":1566491441802,"user_tz":-330,"elapsed":1396,"user":{"displayName":"Ezhirko Arulmozhi","photoUrl":"https://lh3.googleusercontent.com/a-/AAuE7mA_nHqFdfM_zFBpb-BYHYJwL7O3uKBW_MNkY6jW=s64","userId":"16774972310128518178"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["from google.colab import drive\n","drive.mount('/content/drive/')"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Drive already mounted at /content/drive/; to attempt to forcibly remount, call drive.mount(\"/content/drive/\", force_remount=True).\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"HoFfcV-X8XA7","colab_type":"code","colab":{}},"source":["project_path = \"/content/drive/My Drive/AdvanceComputerVission/\""],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"ycefX0ib7piM","colab_type":"code","colab":{}},"source":["import os\n","os.chdir(project_path)\n"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"M2LV9OW57wPy","colab_type":"code","outputId":"8d8f4592-6fd3-45a6-ac4c-f48ea2e16e65","executionInfo":{"status":"ok","timestamp":1566491450497,"user_tz":-330,"elapsed":951,"user":{"displayName":"Ezhirko Arulmozhi","photoUrl":"https://lh3.googleusercontent.com/a-/AAuE7mA_nHqFdfM_zFBpb-BYHYJwL7O3uKBW_MNkY6jW=s64","userId":"16774972310128518178"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["pwd"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["'/content/drive/My Drive/AdvanceComputerVission'"]},"metadata":{"tags":[]},"execution_count":21}]},{"cell_type":"code","metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","id":"u5DhLB_8i1_n","colab_type":"code","outputId":"383be295-20c7-4755-c402-cb92b4bcd1ea","executionInfo":{"status":"ok","timestamp":1566491469716,"user_tz":-330,"elapsed":5613,"user":{"displayName":"Ezhirko Arulmozhi","photoUrl":"https://lh3.googleusercontent.com/a-/AAuE7mA_nHqFdfM_zFBpb-BYHYJwL7O3uKBW_MNkY6jW=s64","userId":"16774972310128518178"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"source":["npz = np.load(project_path+'input_data.npz')\n","X_train = npz['X_train']\n","Y_train = npz['Y_train']\n","del npz\n","print ('We have {} examples to work with'.format(Y_train.shape[0]-1000))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["We have 3113 examples to work with\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"_cell_guid":"ed28bc84-1a8a-472f-98ae-9553fe6d1097","_uuid":"b2e2306b8b16ae7f4f9b63249d1f2ff6e29559d3","id":"nWvj0qAci1_r","colab_type":"text"},"source":["## Checking the data"]},{"cell_type":"code","metadata":{"_cell_guid":"bc5a59a4-1eb3-4505-8cd2-28f738cfbd00","_uuid":"2127e653fc3aaa7bba73f6711bf592b1dded1f88","id":"PVMj0zwYi1_s","colab_type":"code","outputId":"b2bfb347-0bbb-4705-f7b7-ef9219dd7eb4","executionInfo":{"status":"ok","timestamp":1566491964234,"user_tz":-330,"elapsed":1632,"user":{"displayName":"Ezhirko Arulmozhi","photoUrl":"https://lh3.googleusercontent.com/a-/AAuE7mA_nHqFdfM_zFBpb-BYHYJwL7O3uKBW_MNkY6jW=s64","userId":"16774972310128518178"}},"colab":{"base_uri":"https://localhost:8080/","height":314}},"source":["ix = 100 #0-4112\n","imshow(np.squeeze(X_train[ix,:,:,2]))#Looking at the combined channel\n","plt.show()\n","labels = ['Ship','Iceberg']  #0 is no iceberg(ship) and 1 is iceberg\n","print ('This is:',labels[int(Y_train[ix])])"],"execution_count":0,"outputs":[{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAUcAAAEYCAYAAADPkTRJAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBAJD8P8Dga8NnaZaB4wAAAAASUVORK5CYII=\n","text/plain":["<Figure size 432x288 with 2 Axes>"]},"metadata":{"tags":[]}},{"output_type":"stream","text":["This is: Ship\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"W8bk_zYSi1_w","colab_type":"code","outputId":"40324305-3571-4480-f148-a5336ea4bfb8","executionInfo":{"status":"ok","timestamp":1566492355622,"user_tz":-330,"elapsed":2027,"user":{"displayName":"Ezhirko Arulmozhi","photoUrl":"https://lh3.googleusercontent.com/a-/AAuE7mA_nHqFdfM_zFBpb-BYHYJwL7O3uKBW_MNkY6jW=s64","userId":"16774972310128518178"}},"colab":{"base_uri":"https://localhost:8080/","height":314}},"source":["ix = 110 #0-4112\n","imshow(np.squeeze(X_train[ix,:,:,2]))#Looking at the combined channel\n","plt.show()\n","labels = ['Ship','Iceberg']  #0 is no iceberg(ship) and 1 is iceberg\n","print ('This is:',labels[int(Y_train[ix])])"],"execution_count":0,"outputs":[{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAUcAAAEYCAYAAADPkTRJAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n","text/plain":["<Figure size 432x288 with 2 Axes>"]},"metadata":{"tags":[]}},{"output_type":"stream","text":["This is: Iceberg\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"o2E4YUbbi1_z","colab_type":"text"},"source":["# Creating the structure of Siamese Network"]},{"cell_type":"markdown","metadata":{"id":"ch4ZWzvSi1_z","colab_type":"raw"},"source":["You just need to break it into steps:\n","1. 2 Inputs for each images\n","2. Creating a network which both images will go through individually\n","3. Couple the network to each input\n","4. Calculate the L1 distance between them. Just (x1,y1)-(x2,y2)\n","5. 1 Added layer that will say 1 if they are the same and 0 if they are different"]},{"cell_type":"markdown","metadata":{"id":"0GSrPsP8i1_0","colab_type":"text"},"source":["# Create two Inputs for each image - left_input and right_input"]},{"cell_type":"code","metadata":{"id":"wgm7Col0i1_1","colab_type":"code","outputId":"d319aae6-fc74-4729-e9f8-e7a6dd7a751f","executionInfo":{"status":"ok","timestamp":1566492283260,"user_tz":-330,"elapsed":1167,"user":{"displayName":"Ezhirko Arulmozhi","photoUrl":"https://lh3.googleusercontent.com/a-/AAuE7mA_nHqFdfM_zFBpb-BYHYJwL7O3uKBW_MNkY6jW=s64","userId":"16774972310128518178"}},"colab":{"base_uri":"https://localhost:8080/","height":121}},"source":["# We have 2 inputs, 1 for each picture\n","left_input = Input((75,75,3))\n","right_input = Input((75,75,3))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["WARNING: Logging before flag parsing goes to stderr.\n","W0822 16:44:43.709082 140194333845376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:74: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n","\n","W0822 16:44:43.755371 140194333845376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:517: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n","\n"],"name":"stderr"}]},{"cell_type":"markdown","metadata":{"id":"okB4EDSfi1_3","colab_type":"text"},"source":["# Create a base Network in which both images will go through individually."]},{"cell_type":"code","metadata":{"id":"sjoAeYv5i1_3","colab_type":"code","outputId":"b7f966a2-3c7e-43e3-9298-4bf6e8da1b3e","executionInfo":{"status":"ok","timestamp":1566492538174,"user_tz":-330,"elapsed":1624,"user":{"displayName":"Ezhirko Arulmozhi","photoUrl":"https://lh3.googleusercontent.com/a-/AAuE7mA_nHqFdfM_zFBpb-BYHYJwL7O3uKBW_MNkY6jW=s64","userId":"16774972310128518178"}},"colab":{"base_uri":"https://localhost:8080/","height":104}},"source":["# We will use 2 instances of 1 network for this task\n","base_network = Sequential([\n","    Conv2D(5,3, input_shape=(75,75,3)),\n","    Activation('relu'),\n","    MaxPooling2D(),\n","    Conv2D(5,3),\n","    Activation('relu'),\n","    MaxPooling2D(),\n","    Conv2D(7,2),\n","    Activation('relu'),\n","    MaxPooling2D(),\n","    Conv2D(7,2),\n","    Activation('relu'),\n","    Flatten(),\n","    Dense(18),\n","    Activation('sigmoid')\n","])"],"execution_count":0,"outputs":[{"output_type":"stream","text":["W0822 16:48:58.190411 140194333845376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4138: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n","\n","W0822 16:48:58.231856 140194333845376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3976: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.\n","\n"],"name":"stderr"}]},{"cell_type":"markdown","metadata":{"id":"zdqwmS_Ci1_5","colab_type":"text"},"source":["# Process the left and right inputs to the same Base Network(base_network)"]},{"cell_type":"code","metadata":{"id":"h4Kf9KHVi1_6","colab_type":"code","colab":{}},"source":["# Connect both left and right input to the same network\n","# Remember, they have the same weights\n","processed_a = base_network(left_input)\n","processed_b = base_network(right_input)"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"QhnuQPVoi1_8","colab_type":"text"},"source":["# Get the L1 Distance layer between the 2 processed encodings (processed_a and processed_b)"]},{"cell_type":"code","metadata":{"id":"lRzSYXmEi1_9","colab_type":"code","colab":{}},"source":["# Getting the L1 Distance between the 2 encodings\n","L1_layer = Lambda(lambda tensor:K.abs(tensor[0] - tensor[1]))\n","\n","\n","# Add the distance function to the network\n","L1_distance = L1_layer([processed_a, processed_b])\n"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"0n8XfxPhi1__","colab_type":"text"},"source":["# Add the prediction layer in the end and create the Siamese Network using Model function."]},{"cell_type":"code","metadata":{"id":"5-_vPiNmi2AA","colab_type":"code","colab":{}},"source":["prediction = Dense(1,activation='sigmoid')(L1_distance)\n","siamese_net = Model(inputs=[left_input,right_input],outputs=prediction)\n"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"9nPfRRpxi2AC","colab_type":"text"},"source":["# Define the Optimizer Adam and Compile the Siamese Model."]},{"cell_type":"code","metadata":{"_cell_guid":"25ce04b3-0c1e-4732-989f-a8a710cadeff","_uuid":"9adef076f173f6e666647cdc1470a128da820f9a","id":"HosCJgBpi2AC","colab_type":"code","outputId":"82949288-8b05-47d7-afea-18748452fbe1","executionInfo":{"status":"ok","timestamp":1566492947528,"user_tz":-330,"elapsed":1965,"user":{"displayName":"Ezhirko Arulmozhi","photoUrl":"https://lh3.googleusercontent.com/a-/AAuE7mA_nHqFdfM_zFBpb-BYHYJwL7O3uKBW_MNkY6jW=s64","userId":"16774972310128518178"}},"colab":{"base_uri":"https://localhost:8080/","height":154}},"source":["optimizer = Adam(0.001, decay=2.5e-4)\n","\n","siamese_net.compile(loss=\"binary_crossentropy\",optimizer=optimizer,metrics=['accuracy'])"],"execution_count":0,"outputs":[{"output_type":"stream","text":["W0822 16:55:47.218136 140194333845376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:790: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n","\n","W0822 16:55:47.226590 140194333845376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3376: The name tf.log is deprecated. Please use tf.math.log instead.\n","\n","W0822 16:55:47.233585 140194333845376 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/nn_impl.py:180: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n","Instructions for updating:\n","Use tf.where in 2.0, which has the same broadcast rule as np.where\n"],"name":"stderr"}]},{"cell_type":"markdown","metadata":{"_cell_guid":"dbba8e74-6b9c-4c84-96ed-55137265280b","_uuid":"28e3843cae1a192bb14eb04cc1fe1705ef69f405","id":"X-wdkdUai2AE","colab_type":"text"},"source":["# Show the Summary of your Siamese Network"]},{"cell_type":"code","metadata":{"id":"9F0kFEQvi2AE","colab_type":"code","outputId":"aedf44a6-fd94-46e7-b6f0-8787d5de4ed7","executionInfo":{"status":"ok","timestamp":1566492956222,"user_tz":-330,"elapsed":1745,"user":{"displayName":"Ezhirko Arulmozhi","photoUrl":"https://lh3.googleusercontent.com/a-/AAuE7mA_nHqFdfM_zFBpb-BYHYJwL7O3uKBW_MNkY6jW=s64","userId":"16774972310128518178"}},"colab":{"base_uri":"https://localhost:8080/","height":336}},"source":["siamese_net.summary()"],"execution_count":0,"outputs":[{"output_type":"stream","text":["__________________________________________________________________________________________________\n","Layer (type)                    Output Shape         Param #     Connected to                     \n","==================================================================================================\n","input_1 (InputLayer)            (None, 75, 75, 3)    0                                            \n","__________________________________________________________________________________________________\n","input_2 (InputLayer)            (None, 75, 75, 3)    0                                            \n","__________________________________________________________________________________________________\n","sequential_1 (Sequential)       (None, 18)           6912        input_1[0][0]                    \n","                                                                 input_2[0][0]                    \n","__________________________________________________________________________________________________\n","lambda_1 (Lambda)               (None, 18)           0           sequential_1[1][0]               \n","                                                                 sequential_1[2][0]               \n","__________________________________________________________________________________________________\n","dense_2 (Dense)                 (None, 1)            19          lambda_1[0][0]                   \n","==================================================================================================\n","Total params: 6,931\n","Trainable params: 6,931\n","Non-trainable params: 0\n","__________________________________________________________________________________________________\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"AVru6xrQi2AG","colab_type":"text"},"source":["# Creating the Pair of Images to train on"]},{"cell_type":"markdown","metadata":{"_cell_guid":"7080234f-01f4-4df1-88a7-459746335c50","_uuid":"4553890c8130dc76076252a09290497a4afc27e6","id":"6hFhg-hyi2AH","colab_type":"text"},"source":["It's not time to train though. We still have to create pairs of images to train on. There will be Positive(the same class) or Negative(different classes) for outputs. Let's construct this dataset. It is recommended that there are equal amounts positive and negative."]},{"cell_type":"code","metadata":{"_cell_guid":"32cb30db-144d-4f93-af05-6cb9a697c084","_uuid":"09a3e9d488acd34e99912c5f01474029e77637f5","id":"D-LRLBK8i2AH","colab_type":"code","colab":{}},"source":["# First let's separate the dataset from 1 matrix to a list of matricies\n","image_list = np.split(X_train[:1000],1000)\n","label_list = np.split(Y_train[:1000],1000)\n","\n","left_input = []\n","right_input = []\n","targets = []\n","\n","#Number of pairs per image\n","pairs = 5\n","#Let's create the new dataset to train on\n","for i in range(len(label_list)):\n","    for _ in range(pairs):\n","        compare_to = i\n","        while compare_to == i: #Make sure it's not comparing to itself\n","            compare_to = random.randint(0,999)\n","        left_input.append(image_list[i])\n","        right_input.append(image_list[compare_to])\n","        if label_list[i] == label_list[compare_to]:# They are the same\n","            targets.append(1.)\n","        else:# Not the same\n","            targets.append(0.)\n","            \n","left_input = np.squeeze(np.array(left_input))\n","right_input = np.squeeze(np.array(right_input))\n","targets = np.squeeze(np.array(targets))\n","\n","iceimage = X_train[101]\n","test_left = []\n","test_right = []\n","test_targets = []\n","\n","for i in range(Y_train.shape[0]-1000):\n","    test_left.append(iceimage)\n","    test_right.append(X_train[i+1000])\n","    test_targets.append(Y_train[i+1000])\n","\n","test_left = np.squeeze(np.array(test_left))\n","test_right = np.squeeze(np.array(test_right))\n","test_targets = np.squeeze(np.array(test_targets))"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"G4YObcyai2AK","colab_type":"text"},"source":["Now we have a lot more examples.\n","Now we have pairs x 1000 examples to train the network on. \n","Each side will have an input of an image and the output will be one if they are the same and zero if not."]},{"cell_type":"markdown","metadata":{"id":"GBICHx9pi2AK","colab_type":"text"},"source":["# Train the Siamese train on left and right input image pairs created and validate it by test pairs."]},{"cell_type":"code","metadata":{"_cell_guid":"39234b6f-5238-4e5e-b078-4bacda064611","_kg_hide-output":false,"_uuid":"040878576640646d05619511553f1e9af8ab0b29","id":"NEWKpa8Ni2AL","colab_type":"code","outputId":"4a4235e2-1985-4578-d809-7a56e65852e3","executionInfo":{"status":"ok","timestamp":1566493091157,"user_tz":-330,"elapsed":63242,"user":{"displayName":"Ezhirko Arulmozhi","photoUrl":"https://lh3.googleusercontent.com/a-/AAuE7mA_nHqFdfM_zFBpb-BYHYJwL7O3uKBW_MNkY6jW=s64","userId":"16774972310128518178"}},"colab":{"base_uri":"https://localhost:8080/","height":440}},"source":["\n","siamese_net.fit([left_input,right_input], targets,\n","          batch_size=16,\n","          epochs=10,\n","          verbose=1,\n","          validation_data=([test_left,test_right],test_targets))"],"execution_count":0,"outputs":[{"output_type":"stream","text":["W0822 16:57:09.874868 140194333845376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:986: The name tf.assign_add is deprecated. Please use tf.compat.v1.assign_add instead.\n","\n"],"name":"stderr"},{"output_type":"stream","text":["Train on 5000 samples, validate on 3113 samples\n","Epoch 1/10\n","5000/5000 [==============================] - 12s 2ms/step - loss: 0.6900 - acc: 0.5248 - val_loss: 0.6894 - val_acc: 0.5339\n","Epoch 2/10\n","5000/5000 [==============================] - 5s 1ms/step - loss: 0.6402 - acc: 0.6328 - val_loss: 0.6438 - val_acc: 0.6662\n","Epoch 3/10\n","5000/5000 [==============================] - 5s 1ms/step - loss: 0.5701 - acc: 0.7102 - val_loss: 0.5353 - val_acc: 0.7613\n","Epoch 4/10\n","5000/5000 [==============================] - 5s 1ms/step - loss: 0.5135 - acc: 0.7568 - val_loss: 0.5112 - val_acc: 0.7639\n","Epoch 5/10\n","5000/5000 [==============================] - 5s 1ms/step - loss: 0.4754 - acc: 0.7816 - val_loss: 0.5058 - val_acc: 0.7719\n","Epoch 6/10\n","5000/5000 [==============================] - 5s 1ms/step - loss: 0.4381 - acc: 0.8080 - val_loss: 0.5407 - val_acc: 0.7517\n","Epoch 7/10\n","5000/5000 [==============================] - 5s 1ms/step - loss: 0.4089 - acc: 0.8256 - val_loss: 0.5201 - val_acc: 0.7674\n","Epoch 8/10\n","5000/5000 [==============================] - 5s 1ms/step - loss: 0.3851 - acc: 0.8312 - val_loss: 0.5523 - val_acc: 0.7507\n","Epoch 9/10\n","5000/5000 [==============================] - 5s 1ms/step - loss: 0.3711 - acc: 0.8420 - val_loss: 0.5445 - val_acc: 0.7655\n","Epoch 10/10\n","5000/5000 [==============================] - 5s 1ms/step - loss: 0.3451 - acc: 0.8612 - val_loss: 0.5642 - val_acc: 0.7661\n"],"name":"stdout"},{"output_type":"execute_result","data":{"text/plain":["<keras.callbacks.History at 0x7f81300f6240>"]},"metadata":{"tags":[]},"execution_count":41}]},{"cell_type":"code","metadata":{"_uuid":"e572afcaaf7dc8632d813031ebd828d9eec0f9aa","id":"7EZHyFOdi2AP","colab_type":"code","colab":{}},"source":[""],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"vuiAiLWAi2AS","colab_type":"code","colab":{}},"source":[""],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"ILG5UzUVi2AU","colab_type":"code","colab":{}},"source":[""],"execution_count":0,"outputs":[]}]}