{"id":5163,"date":"2025-05-12T11:21:27","date_gmt":"2025-05-12T02:21:27","guid":{"rendered":"https:\/\/www.wetch.co.jp\/note\/?p=5163"},"modified":"2025-05-12T11:21:28","modified_gmt":"2025-05-12T02:21:28","slug":"%e3%83%a9%e3%83%99%e3%83%ab%e7%94%bb%e5%83%8f%e3%81%ae%e3%83%88%e3%83%aa%e3%83%9f%e3%83%b3%e3%82%b0%ef%bc%88%ef%bc%92%ef%bc%89","status":"publish","type":"post","link":"https:\/\/www.wetch.co.jp\/note\/%e3%83%a9%e3%83%99%e3%83%ab%e7%94%bb%e5%83%8f%e3%81%ae%e3%83%88%e3%83%aa%e3%83%9f%e3%83%b3%e3%82%b0%ef%bc%88%ef%bc%92%ef%bc%89\/","title":{"rendered":"\u30e9\u30d9\u30eb\u753b\u50cf\u306e\u30c8\u30ea\u30df\u30f3\u30b0\uff08\uff12\uff09"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">\u300c<a href=\"https:\/\/www.wetch.co.jp\/note\/%e3%83%a9%e3%83%99%e3%83%ab%e7%94%bb%e5%83%8f%e3%81%ae%e3%83%88%e3%83%aa%e3%83%9f%e3%83%b3%e3%82%b0%ef%bc%88%ef%bc%91%ef%bc%89\/\" target=\"_blank\" rel=\"noreferrer noopener\">\u30e9\u30d9\u30eb\u753b\u50cf\u306e\u30c8\u30ea\u30df\u30f3\u30b0\uff08\uff11\uff09<\/a>\u300d\u3067\u306fLabelMe\u3092\u4f7f\u7528\u3057\u3066\u30e9\u30d9\u30eb\u753b\u50cf\u306b\u5bfe\u3059\u308b\u30a2\u30ce\u30c6\u30fc\u30b7\u30e7\u30f3\u3092\u5b9f\u65bd\u3057\u307e\u3057\u305f\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u3053\u306e\u3088\u3046\u306a\u753b\u50cf\u3068\u30a2\u30ce\u30c6\u30fc\u30b7\u30e7\u30f3\u7d50\u679c\u306e\u7d44\u307f\u5408\u308f\u305b\uff08\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\uff09\u3092\u591a\u6570\u7528\u610f\u3059\u308b\u3053\u3068\u3067\u3001\u30bb\u30de\u30f3\u30c6\u30a3\u30c3\u30af\u30bb\u30b0\u30e1\u30f3\u30c6\u30fc\u30b7\u30e7\u30f3\u3092\u5b9f\u65bd\u3059\u308bAI\u30e2\u30c7\u30eb\u306e\u5b66\u7fd2\u304c\u53ef\u80fd\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u3068\u8a00\u3046\u3053\u3068\u3067\u3001AI\u30e2\u30c7\u30eb\u306e\u69cb\u7bc9\u3068\u5b66\u7fd2\u306b\u30c1\u30e3\u30ec\u30f3\u30b8\u3057\u3066\u307f\u307e\u3057\u305f\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u7d50\u679c\u304b\u3089\u5148\u306b\u8a00\u3063\u3066\u304a\u304f\u3068\u3001\u73fe\u6642\u70b9\u3067\u306f\u3001\u3042\u307e\u308a\u671f\u5f85\u901a\u308a\u306e\u6210\u679c\u306f\u5f97\u3089\u308c\u3066\u3044\u307e\u305b\u3093\u3002<br>\u8a66\u884c\u932f\u8aa4\u306e\u904e\u7a0b\u306b\u304a\u3051\u308b\u5931\u6557\u4f8b\u3082\u5f8c\u3005\u4f55\u304b\u306e\u53c2\u8003\u306b\u306a\u308b\u53ef\u80fd\u6027\u304c\u3042\u308a\u307e\u3059\u306e\u3067\u3001\u8a18\u9332\u306b\u6b8b\u3057\u3066\u304a\u3053\u3046\u304b\u3068\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u306a\u304a\u3001\u4ee5\u4e0b\u3067\u63d0\u793a\u3059\u308bPython\u306e\u5b9f\u88c5\u5185\u5bb9\u306f\u5168\u822c\u7684\u306bGemini\u5148\u751f\u4f5c\u3067\u3059\u3002<br>\u672c\u6765\u3067\u3042\u308c\u3070\u5185\u5bb9\u3092\u5341\u5206\u306b\u7406\u89e3\u3057\u305f\u4e0a\u3067\u5229\u7528\u3059\u3079\u304d\u304b\u3068\u601d\u3044\u307e\u3059\u304c\u3001\u307e\u305a\u306f\u305d\u308c\u3063\u307d\u3044\u51e6\u7406\u304c\u6700\u5f8c\u307e\u3067\u52d5\u4f5c\u3059\u308b\u3053\u3068\u3092\u4eca\u56de\u306e\u76ee\u6a19\u3068\u3057\u3066\u3044\u307e\u3059\u306e\u3067\u3001\u5b9f\u88c5\u304a\u3088\u3073\u52d5\u4f5c\u78ba\u8a8d\u3092\u512a\u5148\u3057\u307e\u3057\u305f\u3002<br>\u5185\u5bb9\u306e\u7406\u89e3\u3084\u30d6\u30e9\u30c3\u30b7\u30e5\u30a2\u30c3\u30d7\u306f\u4eca\u5f8c\u306e\u8ab2\u984c\u3068\u3057\u3066\u304a\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u30a2\u30ce\u30c6\u30fc\u30b7\u30e7\u30f3<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">\u307e\u305a\u306f\u3001\u5b66\u7fd2\u7528\u30c7\u30fc\u30bf\u3092\u751f\u6210\u3057\u307e\u3059\u3002<br>\u65e2\u5b58\u30e9\u30d9\u30eb\u753b\u50cf100\u70b9\u306b\u5bfe\u3057\u3066\u3001LabelMe\u3092\u4f7f\u7528\u3057\u3066\u30a2\u30ce\u30c6\u30fc\u30b7\u30e7\u30f3\u3092\u5b9f\u65bd\u3057\u3001\u7d50\u679c\uff08JSON\uff09\u3092\u751f\u6210\u3057\u307e\u3059\u3002<br>\u30e9\u30d9\u30eb\u753b\u50cf\u306f\u300c.\/input\/image\u300d\u306b\u3001JSON\u30d5\u30a1\u30a4\u30eb\u306f\u300c.\/input\/json\u300d\u306b\u683c\u7d0d\u3057\u3066\u304a\u304d\u307e\u3059\u3002<br>JSON\u30d5\u30a1\u30a4\u30eb\u306e\u540d\u79f0\u306f\u5143\u753b\u50cf\u306e\u62e1\u5f35\u5b50\uff08\u300c.jpg\u300d\u7b49\uff09\u3092\u300c.json\u300d\u306b\u5909\u3048\u305f\u3082\u306e\u3067\u3059\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u306a\u304a\u3001\u30e2\u30c7\u30eb\u306e\u5b66\u7fd2\u306e\u90fd\u5408\u4e0a\u3001\u5bfe\u8c61\u753b\u50cf\u306e\u30b5\u30a4\u30ba\u306f\u4e00\u8fba\u304c2\u306en\u4e57\u30d4\u30af\u30bb\u30eb\u306e\u6b63\u65b9\u5f62\u3067\u3042\u3063\u305f\u65b9\u304c\u826f\u3044\u3088\u3046\u306a\u306e\u3067\u3001\u3068\u308a\u3042\u3048\u305a\u65e2\u5b58\u753b\u50cf\u306e\u30a2\u30b9\u30da\u30af\u30c8\u6bd4\u3092\u7121\u8996\u3057\u3066\u3001512&#215;512\u306b\u30ea\u30b5\u30a4\u30ba\u3057\u305f\u3082\u306e\u3092\u4f7f\u7528\u3057\u307e\u3057\u305f\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u751f\u6210<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">\u30a2\u30ce\u30c6\u30fc\u30b7\u30e7\u30f3\u7d50\u679c\uff08JSON\uff09\u306b\u306f\u3001\u30e9\u30d9\u30eb\u90e8\u5206\u3092\u793a\u3059Polygon\uff08\u591a\u89d2\u5f62\uff09\u3092\u69cb\u6210\u3059\u308b\u5404\u30dd\u30a4\u30f3\u30c8\u306e\u5ea7\u6a19\u304c\u8a18\u9332\u3055\u308c\u3066\u3044\u307e\u3059\u3002<br>\u3053\u306e\u72b6\u614b\u306e\u307e\u307e\u3067\u306f\u30e2\u30c7\u30eb\u306e\u5b66\u7fd2\u306b\u306f\u4f7f\u7528\u3067\u304d\u306a\u3044\u305f\u3081\u3001\u3053\u308c\u3092\u753b\u50cf\u5316\u3057\u307e\u3059\u3002<br>\u753b\u50cf\u5316\u3067\u306f\u3001\u30e9\u30d9\u30eb\u90e8\u5206\u3092255\uff08\u767d\uff09\u305d\u306e\u4ed6\u306e\u90e8\u5206\u30920\uff08\u9ed2\uff09\u3068\u3059\u308b\u4e8c\u5024\u753b\u50cf\uff08\u30de\u30b9\u30af\u753b\u50cf\uff09\u3092\u751f\u6210\u3057\u307e\u3059\u3002<br>\u30de\u30b9\u30af\u753b\u50cf\u306e\u540d\u79f0\u306f\u30e9\u30d9\u30eb\u753b\u50cf\u3068\u540c\u3058\u540d\u524d\u306b\u3057\u3001\u30e9\u30d9\u30eb\u753b\u50cf\u306f\u300c.\/output\/images\u300d\u306b\u3001\u30de\u30b9\u30af\u753b\u50cf\u306f\u300c.\/output\/masks\u300d\u306b\u683c\u7d0d\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u4ee5\u4e0b\u304c\u540c\u51e6\u7406\u306e\u5b9f\u88c5\u7d50\u679c\u3067\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>import json\nimport numpy as np\nfrom PIL import Image, ImageDraw\nimport os\nimport cv2\n\ndef create_mask_from_json(json_path, image_size):\n    with open(json_path, 'r') as f:\n        data = json.load(f)\n\n    mask = Image.new('L', image_size, 0)\n    draw = ImageDraw.Draw(mask)\n\n    for shape in data&#91;'shapes']:\n        points = shape&#91;'points']\n        points = &#91;(int(x), int(y)) for x, y in shape&#91;'points']]\n        draw.polygon(points, fill=255)\n\n    return np.array(mask) \/ 255.0\n\ndef create_dataset(image_dir, json_dir, output_dir, image_size):\n    os.makedirs(os.path.join(output_dir, 'images'), exist_ok=True)\n    os.makedirs(os.path.join(output_dir, 'masks'), exist_ok=True)\n\n    image_files = os.listdir(image_dir)\n    for image_file in image_files:\n        if image_file.endswith(('.jpg', '.jpeg')):\n            image_path = os.path.join(image_dir, image_file)\n            json_path = os.path.join(json_dir, image_file.replace(os.path.splitext(image_file)&#91;1], '.json'))\n\n            if os.path.exists(json_path):\n                image = cv2.imread(image_path)\n                image = cv2.resize(image, image_size)\n                mask = create_mask_from_json(json_path, image_size)\n\n                cv2.imwrite(os.path.join(output_dir, 'images', image_file), image)\n                cv2.imwrite(os.path.join(output_dir, 'masks', image_file), mask * 255)\n\nimage_dir = \".\/input\/image\"\njson_dir = \".\/input\/json\"\noutput_dir = \".\/output\"\nimage_size = (512, 512)\ncreate_dataset(image_dir, json_dir, output_dir, image_size)<\/code><\/pre>\n\n\n\n<p class=\"wp-block-paragraph\">\u4e0a\u8a18\u51e6\u7406\u3092\u5b9f\u884c\u3057\u3001output\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u914d\u4e0b\u306b\u30e9\u30d9\u30eb\u753b\u50cf\u3068\u30de\u30b9\u30af\u753b\u50cf\u306e\u30bb\u30c3\u30c8\u3092\u751f\u6210\u3057\u3066\u304a\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">TensorFlow\u306eDataset\u751f\u6210<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">\u4eca\u56de\u306e\u30e2\u30c7\u30eb\u69cb\u7bc9\u3067\u306f\u4f8b\u306b\u3088\u3063\u3066TensorFlow\/Keras\u3092\u4f7f\u7528\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u30e2\u30c7\u30eb\u306e\u5b66\u7fd2\u6642\u306b\u306f\u524d\u8ff0\u306eoutput\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u914d\u4e0b\u306e\u753b\u50cf\u7fa4\u3092\u4f7f\u7528\u3059\u308b\u306e\u3067\u3059\u304c\u3001\u5b66\u7fd2\u51e6\u7406\u306e\u5165\u529b\u3068\u3059\u308b\u305f\u3081\u306b\u306fTensorflow(tf).data.Dataset\u30aa\u30d6\u30b8\u30a7\u30af\u30c8\u3068\u3057\u3066\u753b\u50cf\u30c7\u30fc\u30bf\u3092\u8aad\u307f\u8fbc\u3093\u3067\u304a\u304f\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002<br>\u307e\u305f\u3001\u753b\u50cf\u306e\u4e00\u90e8\u306f\u8a55\u4fa1\u7528\u3068\u3057\u3066\u4f7f\u7528\u3059\u308b\u3088\u3046\u3001\u5b66\u7fd2\u7528\u3068\u306f\u5225\u306eDataset\u3068\u3057\u3066\u30aa\u30d6\u30b8\u30a7\u30af\u30c8\u5316\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u4e0a\u8a18\u6a5f\u80fd\u3092\u95a2\u6570\u300ccreate_tf_dataset\u300d\u3068\u3057\u3066\u5b9f\u88c5\u3057\u307e\u3059\u3002<br>\u5b66\u7fd2\u7528Dataset\u300ctrain_dataset\u300d\u3001\u8a55\u4fa1\u7528Dataset\u300cval_dataset\u300d\u3092\u305d\u308c\u305e\u308c\u751f\u6210\u3057\u3001\u95a2\u6570\u5024\u3068\u3057\u3066\u8fd4\u3059\u4ed5\u69d8\u306b\u306a\u3063\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>import tensorflow as tf\nimport os\nimport random\n\ndef create_file_paths(image_dir, mask_dir):\n    image_files = sorted(os.listdir(image_dir))\n    mask_files = sorted(os.listdir(mask_dir))\n\n    image_paths = &#91;os.path.join(image_dir, file) for file in image_files]\n    mask_paths = &#91;os.path.join(mask_dir, file) for file in mask_files]\n\n    return image_paths, mask_paths\n\ndef load_image_and_mask(image_path, mask_path, image_size):\n    image = tf.io.read_file(image_path)\n    image = tf.image.decode_jpeg(image, channels=3)\n    image = tf.image.resize(image, image_size)\n    image = tf.cast(image, tf.float32) \/ 255.0\n\n    mask = tf.io.read_file(mask_path)\n    mask = tf.image.decode_png(mask, channels=1)\n    mask = tf.image.resize(mask, image_size, method='nearest')\n    mask = tf.cast(mask, tf.float32) \/ 255.0\n\n    return image, mask\n\ndef create_tf_dataset(image_dir, mask_dir, image_size, batch_size, validation_split=0.2, seed=None):\n    image_paths, mask_paths = create_file_paths(image_dir, mask_dir)\n    combined = list(zip(image_paths, mask_paths))\n    if seed is not None:\n        random.seed(seed)\n    random.shuffle(combined)\n    image_paths_tuple, mask_paths_tuple = zip(*combined)\n    image_paths = list(image_paths_tuple)\n    mask_paths = list(mask_paths_tuple)\n\n    dataset_size = len(image_paths)\n    validation_size = int(validation_split * dataset_size)\n\n    train_image_paths = image_paths&#91;validation_size:]\n    train_mask_paths = mask_paths&#91;validation_size:]\n    val_image_paths = image_paths&#91;:validation_size]\n    val_mask_paths = mask_paths&#91;:validation_size]\n\n    train_dataset = tf.data.Dataset.from_tensor_slices((train_image_paths, train_mask_paths))\n    train_dataset = train_dataset.map(\n         lambda image_path,\n         mask_path: load_image_and_mask(image_path, mask_path, image_size)\n    )\n    train_dataset = train_dataset.batch(batch_size)\n    train_dataset = train_dataset.prefetch(tf.data.AUTOTUNE)\n\n    val_dataset = tf.data.Dataset.from_tensor_slices((val_image_paths, val_mask_paths))\n    val_dataset = val_dataset.map(\n        lambda image_path,\n        mask_path: load_image_and_mask(image_path, mask_path, image_size)\n    )\n    val_dataset = val_dataset.batch(batch_size)\n    val_dataset = val_dataset.prefetch(tf.data.AUTOTUNE)\n\n    return train_dataset, val_dataset<\/code><\/pre>\n\n\n\n<h2 class=\"wp-block-heading\">\u30bb\u30b0\u30e1\u30f3\u30c6\u30fc\u30b7\u30e7\u30f3\u30e2\u30c7\u30eb\u306e\u69cb\u7bc9\uff08U-Net\u30e2\u30c7\u30eb\uff09<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b66\u7fd2\u306e\u6e96\u5099\u304c\u3067\u304d\u305f\u306e\u306f\u826f\u3044\u3068\u3057\u3066\u3001\u305d\u3082\u305d\u3082\u3069\u306e\u3088\u3046\u306aAI\u30e2\u30c7\u30eb\u3092\u69cb\u7bc9\u3059\u308b\u304b\u3068\u8a00\u3046\u70b9\u304c\u6700\u3082\u91cd\u8981\u306b\u306a\u308b\u306e\u3067\u3059\u304c\u3001\u30bb\u30de\u30f3\u30c6\u30a3\u30c3\u30af\u30bb\u30b0\u30e1\u30f3\u30c6\u30fc\u30b7\u30e7\u30f3\u7528\u306e\u4ee3\u8868\u7684\u306a\u30e2\u30c7\u30eb\u306e\u4e00\u3064\u3067\u3042\u308b\u300cU-Net\u300d\u304c\u826f\u3055\u3052\u3067\u3042\u308b\u3088\u3046\u306a\u306e\u3067\uff08Gemini\u5148\u751f\u8ac7\uff09\u3001\u4eca\u56de\u306f\u3053\u306e\u30e2\u30c7\u30eb\u3092\u69cb\u7bc9\u3057\u305f\u3044\u3068\u601d\u3044\u307e\u3059\u3002<br>\u3068\u8a00\u3046\u3053\u3068\u3067\u3001U-Net\u30e2\u30c7\u30eb\u3092\u5b9f\u88c5\u3057\u3066\u307f\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>import tensorflow as tf\nfrom tensorflow.keras import layers\n\ndef encoder(inputs, filters, max_pooling=True):\n    conv = layers.Conv2D(filters, 3, activation='relu', padding='same',\n                       kernel_regularizer=tf.keras.regularizers.l2(1e-4))(inputs)\n    conv = layers.BatchNormalization()(conv)\n    conv = layers.Conv2D(filters, 3, activation='relu', padding='same',\n                       kernel_regularizer=tf.keras.regularizers.l2(1e-4))(conv)\n    conv = layers.BatchNormalization()(conv)\n    conv = layers.Dropout(0.1)(conv)\n\n    if max_pooling:\n        next = layers.MaxPooling2D(pool_size=(2, 2))(conv)\n    else:\n        next = conv\n\n    return (next, conv)\n\ndef decoder(inputs, skip_connection, filters):\n    up = layers.Conv2DTranspose(filters, 2, strides=(2, 2), padding='same')(inputs)\n    merge = layers.concatenate(&#91;skip_connection, up], axis=3)\n    conv = layers.Conv2D(filters, 3, activation='relu', padding='same',\n                       kernel_regularizer=tf.keras.regularizers.l2(1e-4))(merge)\n    conv = layers.BatchNormalization()(conv)\n    conv = layers.Conv2D(filters, 3, activation='relu', padding='same',\n                       kernel_regularizer=tf.keras.regularizers.l2(1e-4))(conv)\n    conv = layers.BatchNormalization()(conv)\n    conv = layers.Dropout(0.1)(conv)\n\n    return conv\n\ndef build_unet(input_shape, num_filters=64):\n    inputs = layers.Input(input_shape)\n\n    # \u30a8\u30f3\u30b3\u30fc\u30c0\u30fc\n    enc1, skip1 = encoder(inputs, num_filters)\n    enc2, skip2 = encoder(enc1, num_filters * 2)\n    enc3, skip3 = encoder(enc2, num_filters * 4)\n    enc4, skip4 = encoder(enc3, num_filters * 8)\n\n    conv, _ = encoder(enc4, num_filters * 16, max_pooling=False)\n\n    # \u30c7\u30b3\u30fc\u30c0\u30fc\n    dec4 = decoder(conv, skip4, num_filters * 8)\n    dec3 = decoder(dec4, skip3, num_filters * 4)\n    dec2 = decoder(dec3, skip2, num_filters * 2)\n    dec1 = decoder(dec2, skip1, num_filters)\n\n    outputs = layers.Conv2D(1, 1, activation='sigmoid')(dec1)\n\n    return tf.keras.Model(inputs=inputs, outputs=outputs)<\/code><\/pre>\n\n\n\n<p class=\"wp-block-paragraph\">\u5185\u5bb9\u306b\u95a2\u3057\u3066\u306f\u3001\u307b\u307c\u30c1\u30f3\u30d7\u30f3\u30ab\u30f3\u30d7\u30f3\u3067\u3059\u3002<br>\u3068\u308a\u3042\u3048\u305a\u3001\u4e0a\u8a18\u300cbuild_unet\u300d\u95a2\u6570\u3092\u547c\u3073\u51fa\u3059\u3053\u3068\u3067\u3001U-Net\u30e2\u30c7\u30eb\u304c\u69cb\u7bc9\u3055\u308c\u308b\uff08\u95a2\u6570\u5024\u3068\u3057\u3066\u8fd4\u3055\u308c\u308b\uff09\u306e\u3067\u3001\u305d\u308c\u306b\u5bfe\u3057\u3066\u5b66\u7fd2\u3092\u5b9f\u65bd\u3059\u308b\u3053\u3068\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u30e2\u30c7\u30eb\u306e\u5b66\u7fd2<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">\u4e0a\u8a18\u307e\u3067\u3067\u3001\u5b66\u7fd2\u7528\u306e\u30c7\u30fc\u30bf\u306e\u6e96\u5099\u3068\u30e2\u30c7\u30eb\u306e\u751f\u6210\u304c\u3067\u304d\u305d\u3046\u306a\u306e\u3067\u3001\u3044\u3088\u3044\u3088\u30e2\u30c7\u30eb\u306e\u5b66\u7fd2\u51e6\u7406\u3092\u5b9f\u88c5\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>import tensorflow as tf\nfrom tensorflow.keras import backend as K\n\ndef dice_coef(y_true, y_pred, smooth=1):\n    intersection = K.sum(y_true * y_pred, axis=&#91;1, 2, 3])\n    union = K.sum(y_true, axis=&#91;1, 2, 3]) + K.sum(y_pred, axis=&#91;1, 2, 3])\n    dice = K.mean((2. * intersection + smooth) \/ (union + smooth), axis=0)\n    return dice\n\ndef dice_loss(y_true, y_pred):\n    return 1 - dice_coef(y_true, y_pred)\n\ndef train_model(model, dataset, epochs, batch_size, validation_data=None):\n    model.compile(optimizer='adam', loss=dice_loss, metrics=&#91;dice_coef])\n    if validation_data is not None:\n        model.fit(dataset, epochs=epochs, batch_size=batch_size, validation_data=validation_data)\n    else:\n        model.fit(dataset, epochs=epochs, batch_size=batch_size)<\/code><\/pre>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b66\u7fd2\u51e6\u7406\u306f\u5b9f\u306f\u5358\u7d14\u3067\u3001\u4ee5\u4e0b\u306e2\u3064\u306e\u51e6\u7406\u306e\u307f\u3067\u69cb\u6210\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>model.compile : \u5b66\u7fd2\u6642\u306b\u4f7f\u7528\u3059\u308b\u5404\u7a2e\u51e6\u7406\uff08\u95a2\u6570\uff09\u3092\u8a2d\u5b9a\u3059\u308b<\/li>\n\n\n\n<li>model.fit : \u5b66\u7fd2\u3092\u5b9f\u884c\u3059\u308b<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">compile\u306e\u76ee\u7684\u306f\u5b9f\u88c5\u5185\u5bb9\u306b\u3042\u308b\u3088\u3046\u306b\u30aa\u30d7\u30c6\u30a3\u30de\u30a4\u30b6\uff08optimizer\uff09\u3001\u640d\u5931\u95a2\u6570\uff08loss\uff09\u3001\u8a55\u4fa1\u6307\u6a19\uff08metrics\uff09\u3092\u8a2d\u5b9a\u3059\u308b\u3053\u3068\u306e\u3088\u3046\u3067\u3059\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u30aa\u30d7\u30c6\u30a3\u30de\u30a4\u30b6\u3067\u306f\u30e2\u30c7\u30eb\u306e\u91cd\u307f\u3092\u66f4\u65b0\u3059\u308b\u305f\u3081\u306e\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u3092\u6307\u5b9a\u3059\u308b\u3088\u3046\u3067\u3001\u300cAdam (Adaptive Moment Estimation)\u300d\u306f\u300c\u52fe\u914d\u306e\u5c65\u6b74\u306b\u57fa\u3065\u3044\u3066\u5b66\u7fd2\u7387\u3092\u9069\u5fdc\u7684\u306b\u8abf\u6574\u3059\u308b\u9ad8\u5ea6\u306a\u30aa\u30d7\u30c6\u30a3\u30de\u30a4\u30b6\u300d\u3068\u306e\u3053\u3068\u3067\u3059\u3002<br>\u8981\u306f\u30e2\u30c7\u30eb\u306e\u7cbe\u5ea6\u304c\u5411\u4e0a\u3059\u308b\u3088\u3046\u306b\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u66f4\u65b0\u3059\u308b\u3001\u307e\u3055\u306b\u5b66\u7fd2\u306e\u4e2d\u6838\u3068\u306a\u308b\u6a5f\u80fd\u3068\u8a00\u3048\u307e\u3059\uff08\u591a\u5206\uff09\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u640d\u5931\u95a2\u6570\u306f\u30e2\u30c7\u30eb\u306e\u4e88\u6e2c\u3068\u671f\u5f85\u3055\u308c\u305f\u5024\uff08\u6b63\u89e3\uff09\u3068\u306e\u9593\u306e\u8aa4\u5dee\u3092\u5b9a\u91cf\u7684\u306b\u8a55\u4fa1\u3059\u308b\u305f\u3081\u306e\u95a2\u6570\u3067\u3001\u5b66\u7fd2\u306b\u304a\u3044\u3066\u306f\u3001\u3053\u306e\u640d\u5931\u95a2\u6570\u306e\u5024\u3092\u5c0f\u3055\u304f\u3059\u308b\u3053\u3068\u304c\u76ee\u6a19\u306b\u306a\u308a\u307e\u3059\u3002<br>\u306a\u304a\u3001\u30bb\u30de\u30f3\u30c6\u30a3\u30c3\u30af\u30bb\u30b0\u30e1\u30f3\u30c6\u30fc\u30b7\u30e7\u30f3\u306e\u5b66\u7fd2\u306b\u304a\u3044\u3066\u306f\u4e0a\u8a18\u306b\u3042\u308b\u3088\u3046\u306aDice\u640d\u5931\uff08Dice Loss\uff09\u3092\u4f7f\u7528\u3059\u308b\u3068\u826f\u3044\u3088\u3046\u3067\u3059\uff08Gemini\u5148\u751f\u8ac7\uff09\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u8a55\u4fa1\u6307\u6a19\u306f\u3001\u5b66\u7fd2\u4e2d\u3084\u8a55\u4fa1\u4e2d\u306b\u30e2\u30c7\u30eb\u306e\u6027\u80fd\u3092\u8ffd\u8de1\u304a\u3088\u3073\u8a55\u4fa1\u3059\u308b\u305f\u3081\u306b\u4f7f\u7528\u3055\u308c\u308b\u95a2\u6570\u3067\u3001\u640d\u5931\u95a2\u6570\u3068\u306f\u7570\u306a\u308a\u76f4\u63a5\u7684\u306b\u30e2\u30c7\u30eb\u306e\u91cd\u307f\u3092\u66f4\u65b0\u3059\u308b\u305f\u3081\u306b\u4f7f\u7528\u3055\u308c\u308b\u308f\u3051\u3067\u306f\u306a\u3044\u3088\u3046\u3067\u3059\u3002<br>\u5b66\u7fd2\u904e\u7a0b\u3067\u306f\u640d\u5931\u95a2\u6570\u306e\u5024\u3082\u8868\u793a\u3055\u308c\u307e\u3059\u3057\u3001\u524d\u8ff0\u306e\u3088\u3046\u306b\u5b66\u7fd2\u306e\u76ee\u6a19\u306f\u640d\u5931\u95a2\u6570\u306e\u5024\u3092\u5c0f\u3055\u304f\u3059\u308b\u3053\u3068\u306a\u306e\u3067\u3001\u305d\u3061\u3089\u306e\u5909\u5316\u3060\u3051\u3092\u898b\u3066\u3044\u3051\u3070\u826f\u3044\u306e\u3067\u306f\uff1f\u3068\u601d\u3063\u305f\u308a\u3057\u307e\u3059\u304c\u3001\u305d\u306e\u8fba\u306f\u4eca\u5f8c\u78ba\u8a8d\u3057\u3066\u3044\u304d\u305f\u3044\u3068\u601d\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u4e00\u65b9\u3001fit\u306e\u76ee\u7684\u306f\u5b9f\u969b\u306b\u5b66\u7fd2\u3092\u5b9f\u884c\u3059\u308b\u3053\u3068\u3067\u3042\u308a\u3001\u4e8b\u524d\u306b\u6e96\u5099\u3057\u305f\u5b66\u7fd2\u7528Dataset\u3001\u8a55\u4fa1\u7528Dataset\u3092\u6307\u5b9a\u3059\u308b\u3053\u3068\u3067\u3001\u305d\u308c\u3089\u3092\u4f7f\u7528\u3057\u305f\u5b66\u7fd2\u304c\u884c\u308f\u308c\u307e\u3059\u3002<br>\u307e\u305f\u3001\u5f15\u6570\u3067\u306fepochs\uff08\u5b66\u7fd2\u56de\u6570\uff09\u3068batch_size\uff08\u30d0\u30c3\u30c1\u30b5\u30a4\u30ba\uff09\u3082\u6307\u5b9a\u3057\u3066\u3044\u307e\u3059\u3002<br>epochs\u306e\u65b9\u306f\u4e00\u56de\u306efit\u306e\u547c\u3073\u51fa\u3057\u3067\u4f55\u56de\u306e\u5b66\u7fd2\u3092\u5b9f\u884c\u3059\u308b\u304b\u3068\u8a00\u3046\u5206\u304b\u308a\u3084\u3059\u3044\u3082\u306e\u3067\u3059\u3002<br>batch_size\u306b\u95a2\u3057\u3066\u306f\u2026\u826f\u304f\u5206\u304b\u308a\u307e\u305b\u3093\u3002\u5bbf\u984c\u3068\u3057\u3066\u6b8b\u3057\u3066\u304a\u3044\u3066\u3001\u5148\u306b\u9032\u307f\u307e\u3057\u3087\u3046\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u4e0a\u8a18\u307e\u3067\u3067\u3001\u5b66\u7fd2\u7528Dataset\u306e\u751f\u6210\uff08create_tf_dataset\uff09\u3001\u30e2\u30c7\u30eb\u306e\u69cb\u7bc9\uff08build_unet\uff09\u3001\u5b66\u7fd2\u306e\u5b9f\u884c\uff08train_model\uff09\u306b\u95a2\u3059\u308b\u95a2\u6570\u304c\u6e96\u5099\u3067\u304d\u307e\u3057\u305f\u306e\u3067\u3001\u3053\u308c\u3089\u3092\u547c\u3073\u51fa\u3057\u3066\u5b9f\u969b\u306b\u5b66\u7fd2\u3092\u5b9f\u65bd\u3059\u308b\u51e6\u7406\u3092\u5b9f\u88c5\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>import tensorflow as tf\nfrom create_tf_dataset import create_tf_dataset\nfrom build_unet import build_unet\nfrom train_model import train_model\nfrom train_model import dice_loss\nfrom train_model import dice_coef\nimport os\n\nIMAGE_DIR = \".\/output\/images\"\nMASK_DIR = \".\/output\/masks\"\nIMAGE_SIZE = (512, 512)\nBATCH_SIZE = 32\nINPUT_SHAPE = (512, 512, 3)\nEPOCHS = 10\nMODEL_PATH = \".\/saved_model\/unet.keras\"\nVALIDATION_SPLIT = 0.2\nRANDOM_SEED = 42\n\nif os.path.exists(MODEL_PATH):\n    model = tf.keras.models.load_model(\n        MODEL_PATH,\n        custom_objects={'dice_loss': dice_loss, 'dice_coef': dice_coef}\n    )\nelse:\n    model = build_unet(INPUT_SHAPE)\n\ntrain_dataset, val_dataset = create_tf_dataset(\n    IMAGE_DIR, MASK_DIR, IMAGE_SIZE, BATCH_SIZE,\n    validation_split=VALIDATION_SPLIT, seed=RANDOM_SEED\n)\n\ntrain_model(model, train_dataset, epochs=EPOCHS, batch_size=BATCH_SIZE, validation_data=val_dataset)\n\nmodel.save(MODEL_PATH)<\/code><\/pre>\n\n\n\n<p class=\"wp-block-paragraph\">\u306a\u304a\u3001\u30e2\u30c7\u30eb\u306e\u5b66\u7fd2\u3092\u5b9f\u65bd\u3057\u305f\u3068\u3057\u3066\u3001\u305d\u306e\u7d50\u679c\u306f\u57fa\u672c\u7684\u306b\u306f\u30e1\u30e2\u30ea\u4e0a\u306b\u3057\u304b\u5b58\u5728\u3057\u306a\u3044\u305f\u3081\u3001\u30d7\u30ed\u30bb\u30b9\u304c\u7d42\u4e86\u3057\u3066\u3057\u307e\u3048\u3070\u6d88\u3048\u3066\u3057\u307e\u3044\u307e\u3059\u3002<br>\u305d\u308c\u3067\u306f\u610f\u5473\u304c\u306a\u3044\u305f\u3081\u3001\u300cmodel.save\u300d\u306b\u3088\u308a\u5b66\u7fd2\u7d50\u679c\u3092MODEL_PATH\u3067\u6307\u5b9a\u3057\u305f\u30d5\u30a1\u30a4\u30eb\u306b\u4fdd\u5b58\u3059\u308b\u3088\u3046\u306b\u3057\u3066\u3044\u307e\u3059\u3002<br>\u307e\u305f\u3001\u4fdd\u5b58\u3055\u308c\u305f\u30e2\u30c7\u30eb\u3092\u547c\u3073\u51fa\u3057\u3066\u8ffd\u52a0\u5b66\u7fd2\u3067\u304d\u308b\u3088\u3046\u3001MODEL_PATH\u3067\u6307\u5b9a\u3057\u305f\u30d5\u30a1\u30a4\u30eb\u304c\u5b58\u5728\u3059\u308c\u3070\u3001\u305d\u308c\u3092\u8aad\u307f\u8fbc\u3093\u3060\u4e0a\u3067\u518d\u5ea6\u5b66\u7fd2\u3092\u5b9f\u884c\u3059\u308b\u3088\u3046\u306b\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u30bb\u30b0\u30e1\u30f3\u30c6\u30fc\u30b7\u30e7\u30f3\u306e\u5b9f\u884c<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">\u524d\u8ff0\u306e\u51e6\u7406\u3067\u751f\u6210\u3057\u305f\u5b66\u7fd2\u6e08\u307f\u30e2\u30c7\u30eb\u3092\u4f7f\u7528\u3057\u3066\u3001\u5b9f\u969b\u306b\u30bb\u30b0\u30e1\u30f3\u30c6\u30fc\u30b7\u30e7\u30f3\u3092\u5b9f\u884c\u3057\u3066\u307f\u307e\u3057\u3087\u3046\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5177\u4f53\u7684\u306b\u306f\u4ee5\u4e0b\u306e\u3088\u3046\u306a\u51e6\u7406\u3092\u5b9f\u884c\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>import cv2\nimport numpy as np\nimport tensorflow as tf\nimport os\n\nfrom train_model import dice_loss\nfrom train_model import dice_coef\n\ndef preprocess_image(image_path, image_size):\n    try:\n        image = cv2.imread(image_path)\n        if image is None:\n            raise ValueError(f\"Failed to read image: {image_path}\")\n        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n        image = cv2.resize(image, image_size)\n        image = image \/ 255.0\n        return image.astype(np.float32)\n    except Exception as e:\n        print(f\"Error preprocessing image {image_path}: {e}\")\n        return None\n\ndef infer_mask(model, image):\n    try:\n        image = np.expand_dims(image, axis=0)\n        prediction = model.predict(image)\n        mask = prediction.squeeze()\n        return mask\n    except Exception as e:\n        print(f\"Error inferring mask: {e}\")\n        return None\n\ndef postprocess_equalization(mask):\n    mask_scaled = (mask * 255).astype(np.uint8)\n    equalized_mask = cv2.equalizeHist(mask_scaled)\n    normalized_mask = equalized_mask \/ 255.0\n    return normalized_mask\n\ndef postprocess_mask(mask, threshold=0.5):\n    return (mask &gt; threshold).astype(np.uint8)\n\ndef save_mask(mask, save_path, format='png'):\n    try:\n        if format == 'png':\n            cv2.imwrite(save_path, mask * 255)\n        elif format == 'jpeg':\n            cv2.imwrite(save_path, mask * 255)\n        else:\n            raise ValueError(f\"Unsupported format: {format}\")\n    except Exception as e:\n        print(f\"Error saving mask: {e}\")\n\nIMAGE_DIR = \".\/output\/images\"\nIMAGE_SIZE = (512, 512)\nOUTPUT_DIR = \".\/predict\"\nTHRESHOLD = 0.5\nMODEL_PATH = \".\/saved_model\/unet.keras\"\n\nos.makedirs(OUTPUT_DIR, exist_ok=True)\n\ntry:\n    loaded_model = tf.keras.models.load_model(\n        MODEL_PATH,\n        custom_objects={'dice_loss': dice_loss, 'dice_coef': dice_coef}\n    )\nexcept Exception as e:\n    print(f\"Error loading model: {e}\")\n    exit()\n\nimage_files = os.listdir(IMAGE_DIR)\nfor image_file in image_files:\n    image_path = os.path.join(IMAGE_DIR, image_file)\n    save_path = os.path.join(OUTPUT_DIR, os.path.splitext(image_file)&#91;0] + '.png')\n\n    image = preprocess_image(image_path, IMAGE_SIZE)\n    if image is None:\n        continue\n\n    mask = infer_mask(loaded_model, image)\n    if mask is None:\n        continue\n\n    mask = postprocess_equalization(mask)\n\n    #mask = postprocess_mask(mask, THRESHOLD)\n    save_mask(mask, save_path)<\/code><\/pre>\n\n\n\n<p class=\"wp-block-paragraph\">\u30e2\u30c7\u30eb\u3092\u30ed\u30fc\u30c9\u3057\u3001\u753b\u50cf\u3092\u8aad\u307f\u8fbc\u307f\u3001\u305d\u306e\u753b\u50cf\u306b\u5bfe\u3059\u308b\u30e9\u30d9\u30eb\u90e8\u5206\u306e\u4e88\u6e2c\uff08model.predict\uff09\u3092\u5b9f\u65bd\u3057\u3001\u7d50\u679c\u3092png\u30d5\u30a1\u30a4\u30eb\u3068\u3057\u3066\u4fdd\u5b58\u3059\u308b\u3068\u8a00\u3046\u306e\u304c\u5927\u307e\u304b\u306a\u6d41\u308c\u3067\u3059\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u672c\u6765\u3067\u3042\u308c\u3070\u300cpostprocess_mask\u300d\u3067\u4e8c\u5024\u5316\u3059\u308b\u306e\u3067\u3059\u304c\u3001\u30e2\u30c7\u30eb\u306e\u7d20\u306e\u4e88\u6e2c\u7d50\u679c\uff08\u5404\u30d4\u30af\u30bb\u30eb\u3092\u30e9\u30d9\u30eb\u90e8\u5206\u3067\u3042\u308b\u53ef\u80fd\u6027\u306e\u9ad8\u3055\u306b\u6e96\u3058\u3066255\u304b\u30890\u307e\u3067\u306e\u5024\u3067\u8868\u73fe\u3057\u305f\u753b\u50cf\uff09\u3092\u898b\u3066\u307f\u305f\u304b\u3063\u305f\u306e\u3067\u3001\u73fe\u5728\u306f\u30b3\u30e1\u30f3\u30c8\u30a2\u30a6\u30c8\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u307e\u305f\u3001\u300cpostprocess_equalization\u300d\u3067\u306f\u4e88\u6e2c\u7d50\u679c\u306b\u5bfe\u3057\u3066\u30d2\u30b9\u30c8\u30b0\u30e9\u30e0\u5e73\u5766\u5316\u3092\u884c\u306a\u3063\u3066\u3044\u307e\u3059\u3002<br>\u3053\u308c\u306f\u4e88\u6e2c\u7d50\u679c\u304c\u6975\u7aef\u306b\u767d\uff08\u30e9\u30d9\u30eb\u3063\u307d\u3044\uff09\u5bc4\u308a\u3067\u3042\u3063\u305f\u308a\u3001\u9006\u306b\u9ed2\uff08\u30e9\u30d9\u30eb\u3063\u307d\u304f\u306a\u3044\uff09\u5bc4\u308a\u306b\u306a\u308a\u3001\u305d\u306e\u307e\u307e\u7279\u5b9a\u306e\u95be\u5024\uff08\u4e0a\u8a18\u5b9f\u88c5\u3067\u306f0.5\uff09\u3067\u4e8c\u5024\u5316\u3057\u3066\u3082\u5358\u306a\u308b\u771f\u3063\u767d\uff08\u771f\u3063\u9ed2\uff09\u306a\u753b\u50cf\u306b\u306a\u308b\u3060\u3051\u3068\u3044\u3046\u30b1\u30fc\u30b9\u304c\u591a\u304b\u3063\u305f\u305f\u3081\u3067\u3059\u3002<br>\u5f37\u5236\u7684\u306b\u8272\u306e\u7bc4\u56f2\u3092\u767d\u304b\u3089\u9ed2\u306b\u5e45\u5e83\u304f\u5206\u6563\u3059\u308b\u3088\u3046\u306b\u3057\u3066\u3001\u95be\u50240.5\u3067\u3082\u305d\u308c\u3063\u307d\u304f\u4e8c\u5024\u5316\u3055\u308c\u308b\u3088\u3046\u306b\u3057\u305f\u8a33\u3067\u3059\u304c\u3001\u305d\u3082\u305d\u3082\u305d\u306e\u3088\u3046\u306a\u52a0\u5de5\u3092\u3057\u306a\u3051\u308c\u3070\u306a\u3089\u306a\u3044\u6642\u70b9\u3067\u4e88\u6e2c\u306e\u7cbe\u5ea6\u306b\u304b\u306a\u308a\u306e\u4e0d\u5b89\u304c\u3042\u308a\u307e\u3059\u304c\u2026<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u307e\u3068\u3081<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">\u3068\u308a\u3042\u3048\u305a\u3001\u95a2\u9023\u51e6\u7406\u3092\u3072\u305f\u3059\u3089\u5b9f\u88c5\u3057\u3066\u307f\u305f\u3068\u8a00\u3046\u306e\u304c\u4eca\u56de\u306e\u5185\u5bb9\u3067\u3059\u3002<br>\u4e0a\u8a18\u5b9f\u88c5\u5185\u5bb9\u306f\u5b9f\u969b\u306b\u52d5\u4f5c\u3057\u3001\u4e88\u6e2c\u753b\u50cf\u306e\u51fa\u529b\u3082\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u307e\u3057\u305f\u3002<br>\u305f\u3060\u3001\u7d50\u679c\u304c\u4eca\u3072\u3068\u3064\u306a\u3093\u3067\u3059\u3088\u306d\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u305d\u306e\u8fba\u306b\u95a2\u3057\u3066\u3001\u6b21\u56de\u4ee5\u964d\u3067\u89e6\u308c\u3066\u3044\u304d\u305f\u3044\u3068\u601d\u3044\u307e\u3059\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u300c\u30e9\u30d9\u30eb\u753b\u50cf\u306e\u30c8\u30ea\u30df\u30f3\u30b0\uff08\uff11\uff09\u300d\u3067\u306fLabelMe\u3092\u4f7f\u7528\u3057\u3066\u30e9\u30d9\u30eb\u753b\u50cf\u306b\u5bfe\u3059\u308b\u30a2\u30ce\u30c6\u30fc\u30b7\u30e7\u30f3\u3092\u5b9f\u65bd\u3057\u307e\u3057\u305f\u3002 \u3053\u306e\u3088\u3046\u306a\u753b\u50cf\u3068\u30a2\u30ce\u30c6\u30fc\u30b7\u30e7\u30f3\u7d50\u679c\u306e\u7d44\u307f\u5408\u308f\u305b\uff08\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\uff09\u3092\u591a\u6570\u7528\u610f\u3059\u308b\u3053\u3068\u3067\u3001\u30bb\u30de\u30f3\u30c6\u30a3\u30c3\u30af\u30bb\u30b0\u30e1\u30f3\u30c6\u30fc [&hellip;]<\/p>\n","protected":false},"author":6,"featured_media":5177,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":"","_wp_rev_ctl_limit":""},"categories":[210],"tags":[],"class_list":["post-5163","post","type-post","status-publish","format-standard","has-post-thumbnail","category-210"],"_links":{"self":[{"href":"https:\/\/www.wetch.co.jp\/note\/wp-json\/wp\/v2\/posts\/5163","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wetch.co.jp\/note\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wetch.co.jp\/note\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wetch.co.jp\/note\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wetch.co.jp\/note\/wp-json\/wp\/v2\/comments?post=5163"}],"version-history":[{"count":0,"href":"https:\/\/www.wetch.co.jp\/note\/wp-json\/wp\/v2\/posts\/5163\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.wetch.co.jp\/note\/wp-json\/wp\/v2\/media\/5177"}],"wp:attachment":[{"href":"https:\/\/www.wetch.co.jp\/note\/wp-json\/wp\/v2\/media?parent=5163"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wetch.co.jp\/note\/wp-json\/wp\/v2\/categories?post=5163"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wetch.co.jp\/note\/wp-json\/wp\/v2\/tags?post=5163"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}