homework/Day_022_HW.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 『本次練習內容』\n",
    "#### 搭建一個能識別驗證碼的模型(CNN+CTC)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 『本次練習目的』\n",
    "  #### 了解如何結合CNN與CTC Loss\n",
    "  #### 有時間學員也可以延伸閱讀CNN+RNN+CTC模型\n",
    "  #### 沒有GPU的學員可以使用Colab，可以參考問答區的文章"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## 使用Colab可先run這裡的code\n",
    "#假如要連結自己的雲端硬碟再開啟\n",
    "##from google.colab import drive\n",
    "##drive.mount('/content/gdrive')\n",
    "#檢查GPU\n",
    "##!nvidia-smi\n",
    "##安裝產生驗證碼的套件"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 記得先 pip install captcha"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 136
    },
    "colab_type": "code",
    "id": "AxGi02yvx1Go",
    "outputId": "e7bf7eb2-6699-441b-d724-6b5635ffa2bc"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting captcha\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/90/fe/d4ddf1e6576073b5eaea76e9b2afa022c626212a30c871968480be3ccb7b/captcha-0.3-py3-none-any.whl (101kB)\n",
      "\r",
      "\u001b[K     |███▎                            | 10kB 28.6MB/s eta 0:00:01\r",
      "\u001b[K     |██████▌                         | 20kB 3.0MB/s eta 0:00:01\r",
      "\u001b[K     |█████████▊                      | 30kB 4.4MB/s eta 0:00:01\r",
      "\u001b[K     |█████████████                   | 40kB 2.9MB/s eta 0:00:01\r",
      "\u001b[K     |████████████████▏               | 51kB 3.6MB/s eta 0:00:01\r",
      "\u001b[K     |███████████████████▍            | 61kB 4.3MB/s eta 0:00:01\r",
      "\u001b[K     |██████████████████████▋         | 71kB 4.9MB/s eta 0:00:01\r",
      "\u001b[K     |█████████████████████████▉      | 81kB 5.5MB/s eta 0:00:01\r",
      "\u001b[K     |█████████████████████████████   | 92kB 4.4MB/s eta 0:00:01\r",
      "\u001b[K     |████████████████████████████████| 102kB 3.5MB/s \n",
      "\u001b[?25hRequirement already satisfied: Pillow in /usr/local/lib/python3.6/dist-packages (from captcha) (4.3.0)\n",
      "Requirement already satisfied: olefile in /usr/local/lib/python3.6/dist-packages (from Pillow->captcha) (0.46)\n",
      "Installing collected packages: captcha\n",
      "Successfully installed captcha-0.3\n"
     ]
    }
   ],
   "source": [
    "##!pip install captcha"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 產生驗證碼"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 254
    },
    "colab_type": "code",
    "id": "h2VVCBKwk6Zv",
    "outputId": "72cda421-1eca-46b1-d5bd-3b3d9e47eb08"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'GACT')"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 203,
       "width": 370
      },
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from captcha.image import ImageCaptcha\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import random\n",
    "\n",
    "%matplotlib inline\n",
    "%config InlineBackend.figure_format = 'retina'\n",
    "\n",
    "import string\n",
    "##驗證碼包含0-10數字以及26個英文字母\n",
    "characters = string.digits + string.ascii_uppercase\n",
    "print(characters)\n",
    "\n",
    "##設定產生圖片尺寸，以及總類別，n_class之所以要加一是為了留一個位置給Blank\n",
    "width, height, n_len, n_class = 170, 80, 4, len(characters)+1\n",
    "\n",
    "##設定產生驗證碼的generator\n",
    "generator = ImageCaptcha(width=width, height=height)\n",
    "\n",
    "##我們先練習固定長度4個字的驗證碼\n",
    "random_str = ''.join([random.choice(characters) for j in range(4)])\n",
    "img = generator.generate_image(random_str)\n",
    "\n",
    "plt.imshow(img)\n",
    "plt.title(random_str)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 自己搭建一個CNN模型使得output feature Map 高度被壓縮成小於4，8<寬度<20\n",
    "#### 解答的output feature Map 為 (batch_size,1,13,32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 136
    },
    "colab_type": "code",
    "id": "KX3Jq1sClGGn",
    "outputId": "ea40671a-fc3b-4758-8d29-4692252e840d"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:8: UserWarning: Update your `Conv2D` call to the Keras 2 API: `Conv2D(32, (3, 3), activation=\"relu\")`\n",
      "  \n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:10: UserWarning: Update your `Conv2D` call to the Keras 2 API: `Conv2D(32, (3, 3), activation=\"relu\")`\n",
      "  # Remove the CWD from sys.path while we load stuff.\n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:24: UserWarning: Update your `Dense` call to the Keras 2 API: `Dense(37, activation=\"softmax\", kernel_initializer=\"he_normal\")`\n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:26: UserWarning: Update your `Model` call to the Keras 2 API: `Model(inputs=Tensor(\"in..., outputs=Tensor(\"de...)`\n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:34: UserWarning: Update your `Model` call to the Keras 2 API: `Model(inputs=[<tf.Tenso..., outputs=[<tf.Tenso...)`\n"
     ]
    }
   ],
   "source": [
    "from keras.models import *\n",
    "from keras.layers import *\n",
    "rnn_size = 128\n",
    "\n",
    "input_tensor = Input((height,width, 3))\n",
    "x = input_tensor\n",
    "\n",
    "'''自己設計CNN層'''\n",
    "x = Convolution2D(32, 3, 3, activation='relu')(x)\n",
    "x = Convolution2D(32, 3, 3, activation='relu')(x)\n",
    "x = BatchNormalization(axis=-1)(x)\n",
    "x = MaxPooling2D((2,2))(x)\n",
    "\n",
    "x = Convolution2D(32, 3, 3, activation='relu')(x)\n",
    "x = Convolution2D(32, 3, 3, activation='relu')(x)\n",
    "x = BatchNormalization(axis=-1)(x)\n",
    "x = MaxPooling2D((2,2))(x)\n",
    "\n",
    "x = Convolution2D(32, 3, 3, activation='relu')(x)\n",
    "x = Convolution2D(32, 3, 3, activation='relu')(x)\n",
    "x = BatchNormalization(axis=-1)(x)\n",
    "x = MaxPooling2D((2,2))(x)\n",
    "\n",
    "x = Convolution2D(32, 3, 3, activation='relu')(x)\n",
    "x = Convolution2D(32, 3, 3, activation='relu')(x)\n",
    "x = BatchNormalization(axis=-1)(x)\n",
    "x = MaxPooling2D((2,1))(x)\n",
    "\n",
    "\n",
    "##記錄輸出CNN尺寸，loss部分需要這個資訊\n",
    "##conv_shape=(Batch_size,輸出高度,輸出寬度,輸出深度)\n",
    "conv_shape = x.get_shape()\n",
    "\n",
    "##從(Batch_size,輸出高度,輸出寬度,輸出深度)變成(Batch_size,輸出寬度,輸出深度*輸出高度)，以符合ctc loss需求\n",
    "x = Reshape(target_shape=(int(conv_shape[2]), int(conv_shape[1]*conv_shape[3])))(x)\n",
    "\n",
    "#x = Dense('自己設計', activation='relu')(x)\n",
    "x = Dense(32, activation='relu')(x)\n",
    "\n",
    "x = Dropout(0.25)(x)\n",
    "x = Dense(n_class, activation='softmax')(x)\n",
    "\n",
    "\n",
    "##包裝用來預測的model\n",
    "base_model = Model(input=input_tensor, output=x)\n",
    "\n",
    "\n",
    "##設定要給CTC Loss的資訊\n",
    "labels = Input(name='the_labels', shape=[n_len], dtype='float32')\n",
    "input_length = Input(name='input_length', shape=[1], dtype='int64')\n",
    "label_length = Input(name='label_length', shape=[1], dtype='int64')\n",
    "loss_out = Lambda(ctc_lambda_func, output_shape=(1,), \n",
    "                  name='ctc')([x, labels, input_length, label_length])\n",
    "\n",
    "##這裡的model是用來計算loss\n",
    "model = Model(input=[input_tensor, labels, input_length, label_length], output=[loss_out])\n",
    "\n",
    "##之所以要lambda y_true, y_pred: y_pred是因為我們的loss已經包在網路裡，會output:y_true, y_pred，而我們只需要y_pred\n",
    "model.compile(loss={'ctc': lambda y_true, y_pred: y_pred}, optimizer='SGD')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "id": "m1YRv0TVN_FL",
    "outputId": "c5b995c5-43b0-481b-8cd1-0fd5f85e4f03"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_2 (InputLayer)            (None, 80, 170, 3)   0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_9 (Conv2D)               (None, 78, 168, 32)  896         input_2[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_9 (BatchNor (None, 78, 168, 32)  128         conv2d_9[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_10 (Conv2D)              (None, 76, 166, 32)  9248        batch_normalization_9[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_10 (BatchNo (None, 76, 166, 32)  128         conv2d_10[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2D)  (None, 38, 83, 32)   0           batch_normalization_10[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_11 (Conv2D)              (None, 36, 81, 32)   9248        max_pooling2d_5[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_11 (BatchNo (None, 36, 81, 32)   128         conv2d_11[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_12 (Conv2D)              (None, 34, 79, 32)   9248        batch_normalization_11[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_12 (BatchNo (None, 34, 79, 32)   128         conv2d_12[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_6 (MaxPooling2D)  (None, 17, 39, 32)   0           batch_normalization_12[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_13 (Conv2D)              (None, 15, 37, 32)   9248        max_pooling2d_6[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_13 (BatchNo (None, 15, 37, 32)   128         conv2d_13[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_14 (Conv2D)              (None, 13, 35, 32)   9248        batch_normalization_13[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_14 (BatchNo (None, 13, 35, 32)   128         conv2d_14[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_7 (MaxPooling2D)  (None, 6, 17, 32)    0           batch_normalization_14[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_15 (Conv2D)              (None, 4, 15, 32)    9248        max_pooling2d_7[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_15 (BatchNo (None, 4, 15, 32)    128         conv2d_15[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_16 (Conv2D)              (None, 2, 13, 32)    9248        batch_normalization_15[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_16 (BatchNo (None, 2, 13, 32)    128         conv2d_16[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_8 (MaxPooling2D)  (None, 1, 13, 32)    0           batch_normalization_16[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "reshape_2 (Reshape)             (None, 13, 32)       0           max_pooling2d_8[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "dense_3 (Dense)                 (None, 13, 32)       1056        reshape_2[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_2 (Dropout)             (None, 13, 32)       0           dense_3[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_4 (Dense)                 (None, 13, 37)       1221        dropout_2[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "the_labels (InputLayer)         (None, 4)            0                                            \n",
      "__________________________________________________________________________________________________\n",
      "input_length (InputLayer)       (None, 1)            0                                            \n",
      "__________________________________________________________________________________________________\n",
      "label_length (InputLayer)       (None, 1)            0                                            \n",
      "__________________________________________________________________________________________________\n",
      "ctc (Lambda)                    (None, 1)            0           dense_4[0][0]                    \n",
      "                                                                 the_labels[0][0]                 \n",
      "                                                                 input_length[0][0]               \n",
      "                                                                 label_length[0][0]               \n",
      "==================================================================================================\n",
      "Total params: 68,933\n",
      "Trainable params: 68,421\n",
      "Non-trainable params: 512\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import backend as K\n",
    "import keras\n",
    "\n",
    "##CTC Loss需要四個資訊，分別是\n",
    "## Label\n",
    "## 預測\n",
    "## CNN OUTPUT寬度\n",
    "## 預測影像所包含文字長度\n",
    "def ctc_lambda_func(args):\n",
    "    y_pred, labels, input_length, label_length = args\n",
    "    return K.ctc_batch_cost(labels, y_pred, input_length, label_length)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "5Jzb1YnqlJug"
   },
   "outputs": [],
   "source": [
    "##設計generator產生training data\n",
    "## 產生包含要給loss的資訊\n",
    "## X=輸入影像\n",
    "## np.ones(batch_size)*int(conv_shape[2])=CNN輸出feature Map寬度\n",
    "## np.ones(batch_size)*n_len=字串長度(可浮動)\n",
    "\n",
    "def gen(batch_size=128):\n",
    "    X = np.zeros((batch_size,height, width, 3), dtype=np.uint8)\n",
    "    y = np.zeros((batch_size, n_len), dtype=np.uint8)\n",
    "    while True:\n",
    "        generator = ImageCaptcha(width=width, height=height)\n",
    "        for i in range(batch_size):\n",
    "            random_str = ''.join([random.choice(characters) for j in range(4)])\n",
    "            X[i] = np.array(generator.generate_image(random_str))\n",
    "            y[i] = [characters.find(x) for x in random_str]\n",
    "        yield [X, y,np.ones(batch_size)*int(conv_shape[2]), np.ones(batch_size)*n_len], np.ones(batch_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 225
    },
    "colab_type": "code",
    "id": "dfcFs6InVQaT",
    "outputId": "3f186d49-2e30-4efd-87c0-0670a58bc4c1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label:  [[ 0  6 32 30]]\n",
      "CNN輸出寬度:  [13.]\n",
      "字串長度(可浮動):  [4.]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 191,
       "width": 370
      },
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "next_ge=gen(batch_size=1)\n",
    "test_ge=next(next_ge)\n",
    "plt.imshow(test_ge[0][0][0])\n",
    "print('Label: ',test_ge[0][1])\n",
    "print('CNN輸出寬度: ',test_ge[0][2])\n",
    "print('字串長度(可浮動): ',test_ge[0][3])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 開始訓練"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "id": "fqdKnqollUKj",
    "outputId": "3da34151-a9df-4287-ead6-50c414d30e65"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "300/300 [==============================] - 36s 120ms/step - loss: 2.6483\n",
      "Epoch 2/30\n",
      "300/300 [==============================] - 36s 120ms/step - loss: 2.5941\n",
      "Epoch 3/30\n",
      "300/300 [==============================] - 36s 120ms/step - loss: 2.4702\n",
      "Epoch 4/30\n",
      "300/300 [==============================] - 36s 121ms/step - loss: 2.3983\n",
      "Epoch 5/30\n",
      "300/300 [==============================] - 37s 122ms/step - loss: 2.4439\n",
      "Epoch 6/30\n",
      "300/300 [==============================] - 36s 121ms/step - loss: 2.2228\n",
      "Epoch 7/30\n",
      "300/300 [==============================] - 36s 121ms/step - loss: 2.2304\n",
      "Epoch 8/30\n",
      "300/300 [==============================] - 36s 121ms/step - loss: 2.1833\n",
      "Epoch 9/30\n",
      "300/300 [==============================] - 36s 121ms/step - loss: 2.1137\n",
      "Epoch 10/30\n",
      "300/300 [==============================] - 36s 121ms/step - loss: 2.1068\n",
      "Epoch 11/30\n",
      "300/300 [==============================] - 36s 122ms/step - loss: 2.0592\n",
      "Epoch 12/30\n",
      "300/300 [==============================] - 37s 124ms/step - loss: 2.0204\n",
      "Epoch 13/30\n",
      "300/300 [==============================] - 37s 123ms/step - loss: 2.0545\n",
      "Epoch 14/30\n",
      "300/300 [==============================] - 37s 124ms/step - loss: 1.9768\n",
      "Epoch 15/30\n",
      "300/300 [==============================] - 37s 123ms/step - loss: 1.9763\n",
      "Epoch 16/30\n",
      "300/300 [==============================] - 37s 123ms/step - loss: 1.9368\n",
      "Epoch 17/30\n",
      "300/300 [==============================] - 38s 126ms/step - loss: 1.8706\n",
      "Epoch 18/30\n",
      "300/300 [==============================] - 37s 124ms/step - loss: 1.8781\n",
      "Epoch 19/30\n",
      "300/300 [==============================] - 36s 122ms/step - loss: 1.8408\n",
      "Epoch 20/30\n",
      "300/300 [==============================] - 36s 121ms/step - loss: 1.8981\n",
      "Epoch 21/30\n",
      "300/300 [==============================] - 37s 122ms/step - loss: 1.7787\n",
      "Epoch 22/30\n",
      "300/300 [==============================] - 37s 122ms/step - loss: 1.7409\n",
      "Epoch 23/30\n",
      "300/300 [==============================] - 37s 123ms/step - loss: 1.8090\n",
      "Epoch 24/30\n",
      "300/300 [==============================] - 37s 123ms/step - loss: 1.7667\n",
      "Epoch 25/30\n",
      "300/300 [==============================] - 37s 124ms/step - loss: 1.7412\n",
      "Epoch 26/30\n",
      "300/300 [==============================] - 36s 122ms/step - loss: 1.7469\n",
      "Epoch 27/30\n",
      "300/300 [==============================] - 37s 123ms/step - loss: 1.6994\n",
      "Epoch 28/30\n",
      "300/300 [==============================] - 37s 124ms/step - loss: 1.6732\n",
      "Epoch 29/30\n",
      "300/300 [==============================] - 36s 121ms/step - loss: 1.6230\n",
      "Epoch 30/30\n",
      "300/300 [==============================] - 36s 119ms/step - loss: 1.6077\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7f4694109b00>"
      ]
     },
     "execution_count": 123,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit_generator(gen(32), steps_per_epoch=300, epochs=60)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 預測"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "characters2 = characters + ' '\n",
    "[X_test, y_test, _, _], _  = next(gen(1))\n",
    "y_pred = base_model.predict(X_test)\n",
    "##用ctc_decode得到解答，自己寫可以參考下方\n",
    "out = K.get_value(K.ctc_decode(y_pred, input_length=np.ones(y_pred.shape[0])*y_pred.shape[1], )[0][0])\n",
    "out = ''.join([characters[x] for x in out[0]])\n",
    "y_true = ''.join([characters[x] for x in y_test[0]])\n",
    "\n",
    "plt.imshow(X_test[0])\n",
    "plt.title('pred:' + str(out) + '\\ntrue: ' + str(y_true))\n",
    "\n",
    "argmax = np.argmax(y_pred, axis=2)[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 自己寫 decode CTC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "## 其中0代表預測為空格，如果預測相同字符之間沒有空格要移除\n",
    "word=''\n",
    "n=''\n",
    "for single_result in y_pred[0].argmax(1):\n",
    "    if single_result != 36:\n",
    "        if n != single_result:\n",
    "            word+=characters[single_result]\n",
    "    n=single_result "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 額外參考：加入RNN的神經網路"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:8: UserWarning: Update your `Conv2D` call to the Keras 2 API: `Conv2D(32, (3, 3), activation=\"relu\")`\n",
      "  \n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:10: UserWarning: Update your `Conv2D` call to the Keras 2 API: `Conv2D(32, (3, 3), activation=\"relu\")`\n",
      "  # Remove the CWD from sys.path while we load stuff.\n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:23: UserWarning: Update your `GRU` call to the Keras 2 API: `GRU(128, return_sequences=True, name=\"gru1\", kernel_initializer=\"he_normal\")`\n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:25: UserWarning: Update your `GRU` call to the Keras 2 API: `GRU(128, return_sequences=True, go_backwards=True, name=\"gru1_b\", kernel_initializer=\"he_normal\")`\n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:28: UserWarning: Update your `GRU` call to the Keras 2 API: `GRU(128, return_sequences=True, name=\"gru2\", kernel_initializer=\"he_normal\")`\n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:30: UserWarning: Update your `GRU` call to the Keras 2 API: `GRU(128, return_sequences=True, go_backwards=True, name=\"gru2_b\", kernel_initializer=\"he_normal\")`\n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:34: UserWarning: Update your `Dense` call to the Keras 2 API: `Dense(37, activation=\"softmax\", kernel_initializer=\"he_normal\")`\n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:36: UserWarning: Update your `Model` call to the Keras 2 API: `Model(inputs=Tensor(\"in..., outputs=Tensor(\"de...)`\n",
      "/Users/chening/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:44: UserWarning: Update your `Model` call to the Keras 2 API: `Model(inputs=[<tf.Tenso..., outputs=[<tf.Tenso...)`\n"
     ]
    }
   ],
   "source": [
    "from keras.models import *\n",
    "from keras.layers import *\n",
    "rnn_size = 128\n",
    "\n",
    "input_tensor = Input((height,width, 3))\n",
    "x = input_tensor\n",
    "for i in range(4):\n",
    "    x = Convolution2D(32, 3, 3, activation='relu')(x)\n",
    "    x = keras.layers.BatchNormalization(axis=-1)(x)\n",
    "    x = Convolution2D(32, 3, 3, activation='relu')(x)\n",
    "    x = keras.layers.BatchNormalization(axis=-1)(x)\n",
    "    if i <3:\n",
    "        x = MaxPooling2D(pool_size=(2, 2))(x)\n",
    "    else:\n",
    "        x = MaxPooling2D(pool_size=(2, 1))(x)\n",
    "\n",
    "\n",
    "conv_shape = x.get_shape()\n",
    "x = Reshape(target_shape=(int(conv_shape[2]), int(conv_shape[1]*conv_shape[3])))(x)\n",
    "\n",
    "x = Dense(32, activation='relu')(x)\n",
    "\n",
    "gru_1 = GRU(rnn_size, return_sequences=True, init='he_normal', name='gru1')(x)\n",
    "gru_1b = GRU(rnn_size, return_sequences=True, go_backwards=True, \n",
    "             init='he_normal', name='gru1_b')(x)\n",
    "gru1_merged = keras.layers.add([gru_1, gru_1b])\n",
    "\n",
    "gru_2 = GRU(rnn_size, return_sequences=True, init='he_normal', name='gru2')(gru1_merged)\n",
    "gru_2b = GRU(rnn_size, return_sequences=True, go_backwards=True, \n",
    "             init='he_normal', name='gru2_b')(gru1_merged)\n",
    "x = keras.layers.Concatenate()([gru_2, gru_2b])\n",
    "\n",
    "x = Dropout(0.25)(x)\n",
    "x = Dense(n_class, init='he_normal', activation='softmax')(x)\n",
    "\n",
    "base_model = Model(input=input_tensor, output=x)\n",
    "\n",
    "labels = Input(name='the_labels', shape=[n_len], dtype='float32')\n",
    "input_length = Input(name='input_length', shape=[1], dtype='int64')\n",
    "label_length = Input(name='label_length', shape=[1], dtype='int64')\n",
    "loss_out = Lambda(ctc_lambda_func, output_shape=(1,), \n",
    "                  name='ctc')([x, labels, input_length, label_length])\n",
    "\n",
    "model = Model(input=[input_tensor, labels, input_length, label_length], output=[loss_out])\n",
    "model.compile(loss={'ctc': lambda y_true, y_pred: y_pred}, optimizer='Adam')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_4 (InputLayer)            (None, 80, 170, 3)   0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_25 (Conv2D)              (None, 78, 168, 32)  896         input_4[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_25 (BatchNo (None, 78, 168, 32)  128         conv2d_25[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_26 (Conv2D)              (None, 76, 166, 32)  9248        batch_normalization_25[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_26 (BatchNo (None, 76, 166, 32)  128         conv2d_26[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_13 (MaxPooling2D) (None, 38, 83, 32)   0           batch_normalization_26[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_27 (Conv2D)              (None, 36, 81, 32)   9248        max_pooling2d_13[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_27 (BatchNo (None, 36, 81, 32)   128         conv2d_27[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_28 (Conv2D)              (None, 34, 79, 32)   9248        batch_normalization_27[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_28 (BatchNo (None, 34, 79, 32)   128         conv2d_28[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_14 (MaxPooling2D) (None, 17, 39, 32)   0           batch_normalization_28[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_29 (Conv2D)              (None, 15, 37, 32)   9248        max_pooling2d_14[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_29 (BatchNo (None, 15, 37, 32)   128         conv2d_29[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_30 (Conv2D)              (None, 13, 35, 32)   9248        batch_normalization_29[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_30 (BatchNo (None, 13, 35, 32)   128         conv2d_30[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_15 (MaxPooling2D) (None, 6, 17, 32)    0           batch_normalization_30[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_31 (Conv2D)              (None, 4, 15, 32)    9248        max_pooling2d_15[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_31 (BatchNo (None, 4, 15, 32)    128         conv2d_31[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_32 (Conv2D)              (None, 2, 13, 32)    9248        batch_normalization_31[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_32 (BatchNo (None, 2, 13, 32)    128         conv2d_32[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_16 (MaxPooling2D) (None, 1, 13, 32)    0           batch_normalization_32[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "reshape_4 (Reshape)             (None, 13, 32)       0           max_pooling2d_16[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "dense_7 (Dense)                 (None, 13, 32)       1056        reshape_4[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "gru1 (GRU)                      (None, 13, 128)      61824       dense_7[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "gru1_b (GRU)                    (None, 13, 128)      61824       dense_7[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "add_2 (Add)                     (None, 13, 128)      0           gru1[0][0]                       \n",
      "                                                                 gru1_b[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "gru2 (GRU)                      (None, 13, 128)      98688       add_2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "gru2_b (GRU)                    (None, 13, 128)      98688       add_2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_2 (Concatenate)     (None, 13, 256)      0           gru2[0][0]                       \n",
      "                                                                 gru2_b[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "dropout_4 (Dropout)             (None, 13, 256)      0           concatenate_2[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "dense_8 (Dense)                 (None, 13, 37)       9509        dropout_4[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "the_labels (InputLayer)         (None, 4)            0                                            \n",
      "__________________________________________________________________________________________________\n",
      "input_length (InputLayer)       (None, 1)            0                                            \n",
      "__________________________________________________________________________________________________\n",
      "label_length (InputLayer)       (None, 1)            0                                            \n",
      "__________________________________________________________________________________________________\n",
      "ctc (Lambda)                    (None, 1)            0           dense_8[0][0]                    \n",
      "                                                                 the_labels[0][0]                 \n",
      "                                                                 input_length[0][0]               \n",
      "                                                                 label_length[0][0]               \n",
      "==================================================================================================\n",
      "Total params: 398,245\n",
      "Trainable params: 397,733\n",
      "Non-trainable params: 512\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Captcha.ipynb",
   "provenance": []
  },
  "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": "3.6.9"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}