Train_your_own_data_set_with_mnist.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 檢查 GPU 數量"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['/device:GPU:0', '/device:GPU:1']"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from tensorflow.python.client import device_lib\n",
    "\n",
    "def get_available_gpus():\n",
    "    local_device_protos = device_lib.list_local_devices()\n",
    "    return [x.name for x in local_device_protos if x.device_type == 'GPU']\n",
    "\n",
    "get_available_gpus()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "from tensorflow.contrib.learn.python.learn.datasets.mnist import extract_images, extract_labels\n",
    "\n",
    "# TensorFlow and tf.keras\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras.layers import Conv2D, MaxPooling2D, Dropout, Flatten, Dense\n",
    "# Commonly used modules\n",
    "import numpy as np\n",
    "import os\n",
    "import sys\n",
    "# Images, plots, display, and visualization\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import cv2\n",
    "import IPython\n",
    "from six.moves import urllib\n",
    "\n",
    "#input_data.read_data_sets('./newDataset')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 導入自己的 dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./newDataset/train-images-idx3-ubyte.gz\n",
      "Extracting ./newDataset/train-labels-idx1-ubyte.gz\n",
      "Extracting ./newDataset/t10k-images-idx3-ubyte.gz\n",
      "Extracting ./newDataset/t10k-labels-idx1-ubyte.gz\n"
     ]
    }
   ],
   "source": [
    "with open('./newDataset/train-images-idx3-ubyte.gz', 'rb') as f:\n",
    "  train_images_new = extract_images(f)\n",
    "  train_images_new = train_images_new.reshape(train_images_new.shape[0], 28, 28, 1)\n",
    "with open('./newDataset/train-labels-idx1-ubyte.gz', 'rb') as f:\n",
    "  train_labels_new = extract_labels(f)\n",
    "\n",
    "with open('./newDataset/t10k-images-idx3-ubyte.gz', 'rb') as f:\n",
    "  test_images_new = extract_images(f)\n",
    "  test_images_new = test_images_new.reshape(test_images_new.shape[0], 28, 28, 1)\n",
    "with open('./newDataset/t10k-labels-idx1-ubyte.gz', 'rb') as f:\n",
    "  test_labels_new = extract_labels(f)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 合併自己與 MNIST 的 Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Type is same, concat/merge datasets...\n",
      "Merge successed.\n"
     ]
    }
   ],
   "source": [
    "(train_images_old, train_labels_old), (test_images_old, test_labels_old) = keras.datasets.mnist.load_data()\n",
    "\n",
    "#https://stackoverflow.com/questions/43153076/how-to-concatenate-numpy-arrays-into-a-specific-shape\n",
    "#must have same dim\n",
    "train_images_old = train_images_old.reshape(train_images_old.shape[0], 28, 28, 1)\n",
    "test_images_old = test_images_old.reshape(test_images_old.shape[0], 28, 28, 1)\n",
    "\n",
    "train_images, train_labels, test_images, test_labels = [None, None, None, None]\n",
    "\n",
    "if type(train_images_old) == type(train_images_new):\n",
    "    print(\"Type is same, concat/merge datasets...\")\n",
    "    train_images = np.concatenate((train_images_new, train_images_old))\n",
    "    train_labels = np.concatenate((train_labels_new, train_labels_old))\n",
    "    \n",
    "    test_images = np.concatenate((test_images_new, test_images_old))\n",
    "    test_labels = np.concatenate((test_labels_new, test_labels_old))\n",
    "    print(\"Merge successed.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 前處理圖片並且打印出來看看"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def preprocess_images(imgs): # should work for both a single image and multiple images\n",
    "    sample_img = imgs if len(imgs.shape) == 2 else imgs[0]\n",
    "    assert sample_img.shape in [(28, 28, 1), (28, 28)], sample_img.shape # make sure images are 28x28 and single-channel (grayscale)\n",
    "    return imgs / 255.0\n",
    "\n",
    "train_images = preprocess_images(train_images)\n",
    "test_images = preprocess_images(test_images)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x144 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,2))\n",
    "for i in range(5):\n",
    "    plt.subplot(1,5,i+1)\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "    plt.grid(False)\n",
    "    plt.imshow(train_images[i].reshape(28, 28), cmap=plt.cm.binary)\n",
    "    plt.xlabel(train_labels[i])\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 模型參數設定"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.Sequential()\n",
    "# 32 convolution filters used each of size 5x5\n",
    "model.add(Conv2D(32, kernel_size=(5, 5), activation='relu', input_shape=(28, 28, 1)))\n",
    "# 64 convolution filters used each of size 5x5\n",
    "model.add(Conv2D(64, (5, 5), activation='relu'))\n",
    "# choose the best features via pooling\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "# randomly turn neurons on and off to improve convergence\n",
    "model.add(Dropout(0.25))\n",
    "# flatten since too many dimensions, we only want a classification output\n",
    "model.add(Flatten())\n",
    "# fully connected to get all relevant data\n",
    "model.add(Dense(128, activation='relu'))\n",
    "# one more dropout\n",
    "model.add(Dropout(0.5))\n",
    "# output a softmax to squash the matrix into output probabilities\n",
    "model.add(Dense(10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 設定模型優化器"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_2 (Conv2D)            (None, 24, 24, 32)        832       \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 20, 20, 64)        51264     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 10, 10, 64)        0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 10, 10, 64)        0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 6400)              0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 128)               819328    \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 10)                1290      \n",
      "=================================================================\n",
      "Total params: 872,714\n",
      "Trainable params: 872,714\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer=tf.keras.optimizers.Adam(lr=0.0001), \n",
    "              loss='sparse_categorical_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "model.summary() #shows summary"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 設定可以用 Tensorboard 看"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "tbCallBack = keras.callbacks.TensorBoard(log_dir='./Graph', histogram_freq=0, write_graph=True, write_images=True)\n",
    "#command: tensorboard --logdir path_to_current_dir/Graph "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 進行模型訓練"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/25\n",
      "60287/60287 [==============================] - 2s 35us/sample - loss: 2.1346 - acc: 0.3353\n",
      "Epoch 2/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 1.5568 - acc: 0.6111\n",
      "Epoch 3/25\n",
      "60287/60287 [==============================] - 2s 34us/sample - loss: 0.9872 - acc: 0.7214\n",
      "Epoch 4/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.6987 - acc: 0.7907\n",
      "Epoch 5/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.5612 - acc: 0.8320\n",
      "Epoch 6/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.4838 - acc: 0.8567\n",
      "Epoch 7/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.4330 - acc: 0.8714\n",
      "Epoch 8/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.3883 - acc: 0.8852\n",
      "Epoch 9/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.3512 - acc: 0.8977\n",
      "Epoch 10/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.3210 - acc: 0.9068\n",
      "Epoch 11/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.2948 - acc: 0.9142\n",
      "Epoch 12/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.2677 - acc: 0.9230\n",
      "Epoch 13/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.2465 - acc: 0.9284\n",
      "Epoch 14/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.2312 - acc: 0.9328\n",
      "Epoch 15/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.2168 - acc: 0.9368\n",
      "Epoch 16/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.2015 - acc: 0.9418\n",
      "Epoch 17/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.1898 - acc: 0.9461\n",
      "Epoch 18/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.1767 - acc: 0.9491\n",
      "Epoch 19/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.1735 - acc: 0.9500\n",
      "Epoch 20/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.1621 - acc: 0.9530\n",
      "Epoch 21/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.1567 - acc: 0.9548\n",
      "Epoch 22/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.1495 - acc: 0.9571\n",
      "Epoch 23/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.1464 - acc: 0.9578\n",
      "Epoch 24/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.1398 - acc: 0.9599\n",
      "Epoch 25/25\n",
      "60287/60287 [==============================] - 2s 33us/sample - loss: 0.1378 - acc: 0.9601\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(train_images, train_labels, epochs=25 ,batch_size = 4000,callbacks=[tbCallBack])\n",
    "#adjust epochs and batch_size by your self"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 測試 test 資料集總準確度"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10035/10035 [==============================] - 0s 44us/sample - loss: 0.0750 - acc: 0.9767\n",
      "Test accuracy: 0.9766816\n"
     ]
    }
   ],
   "source": [
    "test_loss, test_acc = model.evaluate(test_images, test_labels)\n",
    "\n",
    "print('Test accuracy:', test_acc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 儲存模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save('test_model.h5') #save model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 載入儲存的模型確認是否正常讀取"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "W0801 11:52:42.644349 140171754252096 deprecation.py:506] From /home/vk/.local/lib/python3.6/site-packages/tensorflow/python/ops/init_ops.py:97: calling GlorotUniform.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Call initializer instance with the dtype argument instead of passing it to the constructor\n",
      "W0801 11:52:42.645456 140171754252096 deprecation.py:506] From /home/vk/.local/lib/python3.6/site-packages/tensorflow/python/ops/init_ops.py:97: calling Zeros.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Call initializer instance with the dtype argument instead of passing it to the constructor\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_2 (Conv2D)            (None, 24, 24, 32)        832       \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 20, 20, 64)        51264     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 10, 10, 64)        0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 10, 10, 64)        0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 6400)              0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 128)               819328    \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 10)                1290      \n",
      "=================================================================\n",
      "Total params: 872,714\n",
      "Trainable params: 872,714\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "10035/10035 [==============================] - 0s 45us/sample - loss: 0.0750 - acc: 0.9767\n",
      "Restored model, accuracy: 97.67%\n"
     ]
    }
   ],
   "source": [
    "test_new_model = keras.models.load_model('test_model.h5') #load model test.\n",
    "test_new_model.summary()\n",
    "test_loss, test_acc = test_new_model.evaluate(test_images, test_labels)\n",
    "\n",
    "\n",
    "print(\"Restored model, accuracy: {:5.2f}%\".format(100*test_acc))"
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}