{"id":3853,"date":"2019-03-29T13:24:04","date_gmt":"2019-03-29T04:24:04","guid":{"rendered":"https:\/\/www.ecomottblog.com\/?p=3853"},"modified":"2019-03-29T16:20:22","modified_gmt":"2019-03-29T07:20:22","slug":"%e3%82%ab%e3%83%a9%e3%82%b9%e3%81%8c%e5%ab%8c%e3%81%84%e3%81%aa%e5%8f%8b%e9%81%94%e3%81%ae%e3%81%9f%e3%82%81%e3%81%ab%e7%94%bb%e5%83%8f%e5%88%a4%e5%ae%9aai%e3%82%92%e4%bd%bf%e3%81%a3%e3%81%a6%e3%80%81","status":"publish","type":"post","link":"https:\/\/test.ecomottblog.com\/?p=3853","title":{"rendered":"\u30ab\u30e9\u30b9\u304c\u5acc\u3044\u306a\u53cb\u9054\u306e\u305f\u3081\u306b\u753b\u50cf\u5224\u5b9aAI\u3092\u4f7f\u3063\u3066\u3001Slack\u306b\u901a\u77e5"},"content":{"rendered":"

\u306f\u3058\u3081\u306b<\/h2>\n

\u30ab\u30e9\u30b9\u304c\u5acc\u3044\u306a\u53cb\u9054\u306e\u305f\u3081\u306b\u3001\u753b\u50cf\u5224\u5b9aAI\u3092\u4f7f\u3063\u3066\u3001\u30ab\u30e9\u30b9\u3092\u767a\u898b\u3057\u305f\u3089slack\u306b\u300c\u30ab\u30e9\u30b9\u6765\u305f\u305e\uff5e\uff01\uff01\uff01\u300d\u3068\u901a\u77e5\u3057\u3066\u304f\u308c\u308b\u3001\u7279\u306b\u5f79\u7acb\u305f\u306a\u305d\u3046\u306a\u3082\u306e\u3092\u4f5c\u308a\u307e\u3057\u305f\u3002<\/p>\n

\"\"<\/p>\n

\u30b4\u30df\u6368\u3066\u5834\u306a\u3069\u306b\u30ab\u30e9\u30b9\u304c\u305f\u307e\u3063\u3066\u3044\u305f\u3089\u300c\u30ab\u30e9\u30b9\u304c\u3044\u308b\uff01\u300d\u3068\u3044\u3046\u3053\u3068\u3092slack\u3067\u6559\u3048\u3066\u304f\u308c\u3001\u3054\u307f\u6368\u3066\u306b\u884c\u304f\u5fc3\u306e\u6e96\u5099\u304c\u3067\u304d\u308b\u3082\u306e\u3002\u304f\u3089\u3044\u306b\u8003\u3048\u3066\u3044\u307e\u3059\uff08\u7b11\uff09<\/p>\n

\u672c\u5f53\u306f\u3001\u5b9f\u969b\u306b\u30b9\u30c8\u30ea\u30fc\u30df\u30f3\u30b0\u3092\u3068\u3063\u3066\u3044\u308b\u30ab\u30e1\u30e9\u304c\u3001\u30ab\u30e9\u30b9\u3092\u767a\u898b\u3057\u305f\u3089slack\u306b\u901a\u77e5\u3059\u308b\u3068\u3044\u3046\u6d41\u308c\u3067\u3084\u308a\u305f\u304b\u3063\u305f\u3067\u3059\u304c\u3001\u30ab\u30e1\u30e9\u306e\u90e8\u5206\u306f\u307e\u305f\u306e\u6a5f\u4f1a\u306b\u3068\u3044\u3046\u3053\u3068\u3067\u30fb\u30fb\u30fb\u3002\u4f59\u88d5\u304c\u3042\u308c\u3070\u3084\u308a\u307e\u3059\u3002\u30ab\u30e1\u30e9\u306e\u65b9\u3082\u3002<\/strong><\/p>\n

Udemy\u306e\u3010\u753b\u50cf\u5224\u5b9aAI\u81ea\u4f5c\u306b\u30c1\u30e3\u30ec\u30f3\u30b8\uff01\u3011TensorFlow\u30fbKeras\u30fbPython\u30fbFlask\u3067\u4f5c\u308b\u6a5f\u68b0\u5b66\u7fd2\u30a2\u30d7\u30ea\u958b\u767a\u5165\u9580<\/a>\u3092\u53c2\u8003\u306b\u5206\u985e\u5668\u3092\u4f5c\u308a\u3001\u8a55\u4fa1\u306e\u5024\u306b\u30ab\u30e9\u30b9\u3092\u691c\u51fa\u3057\u305f\u3089slack\u901a\u77e5\u3092\u3059\u308b\u3068\u3044\u3046\u69cb\u9020\u306b\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n

CNN\u306b\u95a2\u3059\u308b\u5185\u5bb9\u306f\u3001Udemy\u306e\u3010\u753b\u50cf\u5224\u5b9aAI\u81ea\u4f5c\u306b\u30c1\u30e3\u30ec\u30f3\u30b8\uff01\u3011TensorFlow\u30fbKeras\u30fbPython\u30fbFlask\u3067\u4f5c\u308b\u6a5f\u68b0\u5b66\u7fd2\u30a2\u30d7\u30ea\u958b\u767a\u5165\u9580<\/a>\u3092\u53c2\u8003\u306b\u3057\u3066\u3044\u305f\u3060\u304f\u304b\u3001
\n\u300c\u30a8\u30a2\u30b3\u30f3\u6d88\u3057\u305f\u3063\u3051\uff1f\u300d\u3092AI\u3067\u306a\u304f\u305d\u3046\uff01<\/a>\u3092\u3054\u53c2\u8003\u304f\u3060\u3055\u3044\u3002<\/p>\n

\u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001\u8a55\u4fa1\u7d50\u679c\u304b\u3089slack\u306b\u6d41\u3059\u90e8\u5206\u3092\u30e1\u30a4\u30f3\u306b\u8a18\u8ff0\u3057\u3066\u304a\u308a\u307e\u3059\u3002<\/p>\n

<\/p>\n

\u3061\u306a\u307f\u306b<\/h3>\n

Udemy\u306e\u3010\u753b\u50cf\u5224\u5b9aAI\u81ea\u4f5c\u306b\u30c1\u30e3\u30ec\u30f3\u30b8\uff01\u3011TensorFlow\u30fbKeras\u30fbPython\u30fbFlask\u3067\u4f5c\u308b\u6a5f\u68b0\u5b66\u7fd2\u30a2\u30d7\u30ea\u958b\u767a\u5165\u9580<\/a>\u3092\u304a\u4f5c\u308a\u3044\u305f\u3060\u3044\u305f\u4e95\u4e0a \u535a\u6a39\u3055\u3093\u306b\u611f\u8b1d\u3067\u3059\u3002\u304a\u52e7\u3081\u3067\u3059\u3002<\/p>\n

 <\/p>\n

\u5185\u5bb9<\/h2>\n

\u753b\u50cf\u3092\u4e8b\u524d\u306b\u5b66\u7fd2\u3055\u305b\u305f\u306e\u3061\u306b\u3001\u5206\u985e\u5668\u306b\u304b\u3051\u3001\u305d\u306e\u753b\u50cf\u304c\u30ab\u30e9\u30b9\u304b\u3069\u3046\u304b\u3092\u5224\u5b9a\u3059\u308b\u3068\u3044\u3046\u5185\u5bb9\u3067\u3059\u3002<\/p>\n

\u5185\u90e8<\/h3>\n

\u25a0\u30e2\u30c7\u30eb\u306e\u5f62\u6210\u3068\u30ed\u30fc\u30c9<\/h4>\n

model_train()\u3067\u3001\u30e2\u30c7\u30eb\u3092\u5f62\u6210\u3057\u3066\u300ccnn.h5\u300d\u3068\u3044\u3046\u540d\u524d\u3067\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3057\u307e\u3059\u3002<\/p>\n

def model_train(X, y):\n\n\u00a0 \u00a0 model = Sequential()\n\u00a0 \u00a0 model.add(Conv2D(32, (3, 3),padding='same',input_shape=X.shape[1:]))\n\u00a0 \u00a0 model.add(Activation('relu'))\n\u00a0 \u00a0 model.add(Conv2D(32, (3, 3)))\n\u00a0 \u00a0 model.add(Activation('relu'))\n\u00a0 \u00a0 model.add(MaxPooling2D(pool_size=(2, 2)))\n\u00a0 \u00a0 model.add(Dropout(0.25))\n\n\u00a0 \u00a0 model.add(Conv2D(64, (3, 3), padding='same'))\n\u00a0 \u00a0 model.add(Activation('relu'))\n\u00a0 \u00a0 model.add(Conv2D(64, (3, 3)))\n\u00a0 \u00a0 model.add(Activation('relu'))\n\u00a0 \u00a0 model.add(MaxPooling2D(pool_size=(2, 2)))\n\u00a0 \u00a0 model.add(Dropout(0.25))\n\n\u00a0 \u00a0 model.add(Flatten())\n\u00a0 \u00a0 model.add(Dense(512))\n\u00a0 \u00a0 model.add(Activation('relu'))\n\u00a0 \u00a0 model.add(Dropout(0.5))\n\u00a0 \u00a0 model.add(Dense(num_classes))\n\u00a0 \u00a0 model.add(Activation('softmax'))\n\n\u00a0 \u00a0 opt = keras.optimizers.rmsprop(lr=0.0001, decay=1e-6)\n\n\u00a0 \u00a0 # Let's train the model using RMSprop\n\u00a0 \u00a0 model.compile(loss='categorical_crossentropy',optimizer=opt,metrics=['accuracy'])\n\n\u00a0 \u00a0 model.fit(X, y, batch_size=32, epochs=100)\n\n\u00a0 \u00a0 # \u30e2\u30c7\u30eb\u306e\u4fdd\u5b58\n\u00a0 \u00a0 model.save('.\/cnn.h5')\n\n\u00a0 \u00a0 return model\n<\/code><\/pre>\n
\u4ee5\u4e0b\u306eload_model()\u3067\u3001\u4fdd\u5b58\u3057\u305f\u30e2\u30c7\u30eb\u3092\u30ed\u30fc\u30c9\u3057\u307e\u3059\u3002<\/p>\n
def load_model():\n\u00a0 \u00a0 # \u30e2\u30c7\u30eb\u306e\u30ed\u30fc\u30c9\n\u00a0 \u00a0 return load_model('.\/cnn.h5')\n<\/code><\/pre>\n
\u25a0\u5206\u985e\u5668\u306e\u7d50\u679c\u3092slack\u306b\u901a\u77e5<\/h4>\n
slack\u3078\u306e\u901a\u77e5\u65b9\u6cd5\u306f\u30b0\u30b0\u308c\u3070\u51fa\u3066\u304f\u308b\u306e\u3067\u3001\u3053\u3053\u3067\u306f\u7701\u7565\u3057\u307e\u3059\u3002<\/p>\n
# slack\u3078\u901a\u77e5\ndef notify_slack(urlToHook, message):\n\u00a0 \u00a0 slack = slackweb.Slack(url = urlToHook)\n\u00a0 \u00a0 slack.notify(text = message)\n\ndef main():\n\u00a0 \u00a0 # slack\u3078\u901a\u77e5\u3059\u308b\u305f\u3081\u306eWebhook\u306eURL\n\u00a0 \u00a0 url_to_hook = \"hogehogehogehogehoge\"\n\n\u00a0 \u00a0 image = Image.open(sys.argv[1])\n\u00a0 \u00a0 image = image.convert('RGB')\n\u00a0 \u00a0 image = image.resize((image_size, image_size))\n\u00a0 \u00a0 data = np.asarray(image)\/255\n\u00a0 \u00a0 X = []\n\u00a0 \u00a0 X.append(data)\n\u00a0 \u00a0 X = np.array(X)\n\u00a0 \u00a0 model = load_model()\n\n\u00a0 \u00a0 result = model.predict([X])[0]\n\u00a0 \u00a0 predicted = result.argmax()\n\u00a0 \u00a0 percentage = int(result[predicted] * 100)\n\u00a0 \u00a0 # 80%\u4ee5\u4e0a\u4f3c\u3066\u3044\u308b\u306a\u3089\u30ab\u30e9\u30b9\u3067\u901a\u77e5\n\u00a0 \u00a0 if percentage > 80 and classes[predicted]==\"crow\":\n\u00a0 \u00a0 \u00a0 \u00a0 message = \"\u305d\u308c\u306f\u30ab\u30e9\u30b9\u3060\u7387\" + str(percentage) + \" %\" + \"\u3064\u307e\u308a\u30ab\u30e9\u30b9\u3060\u305d\u3044\u3064\u306f\"\n\u00a0 \u00a0 else:\n\u00a0 \u00a0 \u00a0 \u00a0 message = \"\u30ab\u30e9\u30b9\u3058\u3083\u306d\u30fc\u3002\u5b89\u5fc3\u3057\u308d\"\n\n\u00a0 \u00a0 notify_slack(url_to_hook, message)\n<\/code><\/pre>\n
\u5b9f\u9a13<\/h3>\n
\u3067\u306f\u65e9\u901f\u3001\u30ab\u30e9\u30b9\u3092\u5206\u985e\u5668\u306b\u5165\u308c\u3066\u307f\u3066\u3001\u30ab\u30e9\u30b9\u304b\u3069\u3046\u304b\u6559\u3048\u3066\u3082\u3089\u3044\u307e\u3059\u3002<\/p>\n
\"\"<\/p>\n
\u3068\u3057\u3066\u3001\u30ab\u30e9\u30b9\u306e\u753b\u50cf\u3092\u5f15\u6570\u306b\u3057\u3066\u3001\u5206\u985e\u5668\u306b\u6295\u3052\u307e\u3059\u3002<\/p>\n
\"\"<\/p>\n
 <\/p>\n
\u304a\u3002\u3046\u307e\u304f\u3067\u304d\u3066\u307e\u3059\u306d\u3002
\n\u30ab\u30e9\u30b9\u3089\u3057\u3044\u3067\u3059\u3002
\n\u3061\u306a\u307f\u306b\u5224\u5b9a\u306b\u4f7f\u3063\u305f\u753b\u50cf\u306f\u3053\u3093\u306a\u611f\u3058\u3067\u3059\u3002<\/p>\n
\"\"<\/p>\n
\u3067\u306f\u3001\u307e\u3063\u305f\u304f\u9055\u3046\u30b5\u30eb\u306e\u753b\u50cf\u3067\u306f\uff1f<\/p>\n
\"\"<\/p>\n
 <\/p>\n
\u3068\u3057\u3066\u3001\u30b5\u30eb\u306e\u753b\u50cf\u3092\u5f15\u6570\u306b\u3057\u3066\u3001\u5206\u985e\u5668\u306b\u6295\u3052\u307e\u3059\u3002<\/p>\n
\"\"<\/p>\n
\u304a\u3002\u3046\u307e\u304f\u3067\u304d\u3066\u307e\u3059\u306d\u3002
\n\u307e\u3042\u30b4\u30df\u6368\u3066\u5834\u306b\u30b5\u30eb\u3044\u305f\u65b9\u304c\u5b89\u5fc3\u3067\u304d\u307e\u305b\u3093\u304c\u306d\u3002
\n\u3061\u306a\u307f\u306b\u5224\u5b9a\u306b\u4f7f\u3063\u305f\u753b\u50cf\u306f\u3053\u3093\u306a\u611f\u3058\u3067\u3059\u3002<\/p>\n
\"\"<\/p>\n
\u3058\u3083\u3042\u30ab\u30e9\u30b9\u306b\u4f3c\u3066\u3044\u308b\u3084\u3064\u306f\uff1f<\/p>\n
\"\"<\/p>\n
 <\/p>\n
\u3068\u3057\u3066\u3001\u9df9\uff1f\u9df2\uff1f\u306e\u753b\u50cf\u3092\u5f15\u6570\u306b\u3057\u3066\u3001\u5206\u985e\u5668\u306b\u6295\u3052\u307e\u3059\u3002<\/p>\n
\"\"<\/p>\n
\u3093\uff1f\u30ab\u30e9\u30b9\u3067\u306f\u306a\u3044\u306e\u3060\u304c\u30fb\u30fb\u30fb\u3002
\n\u3061\u306a\u307f\u306b\u753b\u50cf\u306f\u30fb\u30fb\u30fb\u306a\u3093\u3060\u308d\u3046\u3001\u9df9\u304b\u306a\uff1f\u9df2\uff1f
\n\u3061\u306a\u307f\u306b\u5224\u5b9a\u306b\u4f7f\u3063\u305f\u753b\u50cf\u306f\u3053\u3093\u306a\u611f\u3058\u3067\u3059\u3002
\n\u4e0b\u306e\u753b\u50cf\u306f\u304a\u3058\u3055\u3093\u304c\u3044\u307e\u3059\u304c\u3001\u304a\u3058\u3055\u3093\u304c\u3044\u306a\u3044\u4e0b\u306e\u3088\u3046\u306a\u9df9\u3092\u4f7f\u3044\u307e\u3057\u305f\u3002<\/p>\n
\"\"<\/p>\n
\u3057\u304b\u3082\u3001\u30d5\u30a1\u30a4\u30eb\u540d\u306bbird3\u3063\u3066\u30fb\u30fb\u30fb
\n\u9ce5\u5927\u597d\u304d\u306a\u4eba\u306b\u6012\u3089\u308c\u305d\u3046\u306a\u30d5\u30a1\u30a4\u30eb\u540d\u306e\u4ed8\u3051\u65b9\u3067\u3057\u305f\u3002
\n\u3059\u307f\u307e\u305b\u3093\u3002<\/p>\n
\u7d50\u8ad6<\/h2>\n
\u3068\u3044\u3046\u3053\u3068\u3067\u3001\u30e2\u30c7\u30eb\u3092\u5b9a\u7fa9\u3059\u308b\u524d\u51e6\u7406\u3068\u753b\u50cf\u306e\u679a\u6570\u3084\u305d\u306e\u8cea\u306b\u3088\u3063\u3066\u3001\u5024\u304c\u3060\u3044\u3076\u5909\u308f\u308b\u307f\u305f\u3044\u3067\u3059\u3002
\n\u307e\u3060\u307e\u3060\u81ea\u5206\u3082\u52c9\u5f37\u4e2d\u3067\u3059\u306e\u3067\u3001\u7cbe\u5ea6\u3092\u9ad8\u304f\u3057\u3066\u9811\u5f35\u308a\u307e\u3059\u3002
\n\u3055\u3089\u306b\u3001\u5b9f\u969b\u306e\u30ab\u30e1\u30e9\u304b\u3089\u5373\u6642\u306b\u5224\u5b9a\u3057\u3066Slack\u901a\u77e5\u3068\u304b\u3084\u308c\u305f\u3089\u3084\u308a\u307e\u3059\u3002<\/p>\n
\u53c2\u8003<\/h2>\n
\u30fb\u3010\u753b\u50cf\u5224\u5b9aAI\u81ea\u4f5c\u306b\u30c1\u30e3\u30ec\u30f3\u30b8\uff01\u3011TensorFlow\u30fbKeras\u30fbPython\u30fbFlask\u3067\u4f5c\u308b\u6a5f\u68b0\u5b66\u7fd2\u30a2\u30d7\u30ea\u958b\u767a\u5165\u9580<\/a>
\n\u30fb\u300c\u30a8\u30a2\u30b3\u30f3\u6d88\u3057\u305f\u3063\u3051\uff1f\u300d\u3092AI\u3067\u306a\u304f\u305d\u3046\uff01<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"
\u306f\u3058\u3081\u306b \u30ab\u30e9\u30b9\u304c\u5acc\u3044\u306a\u53cb\u9054\u306e\u305f\u3081\u306b\u3001\u753b\u50cf\u5224\u5b9aAI\u3092\u4f7f\u3063\u3066\u3001\u30ab\u30e9\u30b9\u3092\u767a\u898b\u3057\u305f\u3089slack\u306b\u300c\u30ab\u30e9\u30b9\u6765\u305f\u305e\uff5e\uff01\uff01\uff01\u300d\u3068\u901a\u77e5\u3057\u3066\u304f\u308c\u308b\u3001\u7279\u306b\u5f79\u7acb\u305f\u306a\u305d\u3046\u306a\u3082\u306e\u3092\u4f5c\u308a\u307e\u3057\u305f\u3002 \u30b4\u30df\u6368\u3066\u5834\u306a\u3069\u306b\u30ab\u30e9\u30b9\u304c\u305f\u307e\u3063\u3066\u3044\u305f\u3089\u300c\u30ab\u30e9\u30b9\u304c\u3044 […]<\/p>\n","protected":false},"author":34,"featured_media":3869,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[27],"tags":[45,51,50,48,49,31,52],"_links":{"self":[{"href":"https:\/\/test.ecomottblog.com\/index.php?rest_route=\/wp\/v2\/posts\/3853"}],"collection":[{"href":"https:\/\/test.ecomottblog.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/test.ecomottblog.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/test.ecomottblog.com\/index.php?rest_route=\/wp\/v2\/users\/34"}],"replies":[{"embeddable":true,"href":"https:\/\/test.ecomottblog.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=3853"}],"version-history":[{"count":11,"href":"https:\/\/test.ecomottblog.com\/index.php?rest_route=\/wp\/v2\/posts\/3853\/revisions"}],"predecessor-version":[{"id":3855,"href":"https:\/\/test.ecomottblog.com\/index.php?rest_route=\/wp\/v2\/posts\/3853\/revisions\/3855"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/test.ecomottblog.com\/index.php?rest_route=\/wp\/v2\/media\/3869"}],"wp:attachment":[{"href":"https:\/\/test.ecomottblog.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=3853"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/test.ecomottblog.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=3853"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/test.ecomottblog.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=3853"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}