Fine_Tuning_VGG19.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "a0cbb497",
   "metadata": {
    "id": "72ab04b1-ceeb-42f9-bd60-b47ab0b27d09"
   },
   "source": [
    "## Import"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "be650862",
   "metadata": {
    "executionInfo": {
     "elapsed": 4220,
     "status": "ok",
     "timestamp": 1652256032258,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "5b9f99a6-eeeb-4286-822e-545cdba4d2b0"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-05-12 16:13:55.871585: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n"
     ]
    }
   ],
   "source": [
    "import os \n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from glob import glob\n",
    "from tqdm.auto import tqdm\n",
    "import cv2\n",
    "\n",
    "from tensorflow.keras.applications import VGG19\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.models import Sequential, Model\n",
    "from tensorflow.keras.layers import Flatten, Dense, Dropout, GlobalAveragePooling2D, Input\n",
    "import tensorflow.keras as keras\n",
    "from tensorflow.data import Dataset\n",
    "\n",
    "\n",
    "from tensorflow.keras.optimizers import Adam, RMSprop\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1a584a3b",
   "metadata": {
    "id": "7aaad7d5-643e-4ccd-ae32-920a9553b630"
   },
   "source": [
    "## Data Pre-processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "16fecf17",
   "metadata": {
    "executionInfo": {
     "elapsed": 415,
     "status": "ok",
     "timestamp": 1652256061293,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "343093ae-587c-4e26-8774-2cab3de75add"
   },
   "outputs": [],
   "source": [
    "def get_train_data(data_dir):\n",
    "    img_path_list = []\n",
    "    label_list = []\n",
    "    for case_name in os.listdir(data_dir):\n",
    "        current_path = os.path.join(data_dir, case_name)\n",
    "        if os.path.isdir(current_path):\n",
    "            # get image path\n",
    "            img_path_list.extend(glob(os.path.join(current_path, 'image', '*.jpg')))\n",
    "            img_path_list.extend(glob(os.path.join(current_path, 'image', '*.png')))\n",
    "            \n",
    "            # get label\n",
    "            label_df = pd.read_csv(current_path+'/label-{}.csv'.format(str(int(case_name[-2:]))))\n",
    "            label_list.extend(label_df['leaf_weight'])\n",
    "                \n",
    "    return img_path_list, label_list\n",
    "\n",
    "def get_test_data(data_dir):\n",
    "    # get image path\n",
    "    img_path_list = glob(os.path.join(data_dir, 'image', '*.jpg'))\n",
    "    img_path_list.extend(glob(os.path.join(data_dir, 'image', '*.png')))\n",
    "    img_path_list.sort(key=lambda x:int(x.split('/')[-1].split('.')[0]))\n",
    "    return img_path_list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "7d9e0bb9",
   "metadata": {
    "executionInfo": {
     "elapsed": 416,
     "status": "ok",
     "timestamp": 1652256063966,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "Bc7vEgitz93P"
   },
   "outputs": [],
   "source": [
    "def load_img_numpy(img_dir, target_size) :\n",
    "    img = cv2.imread(img_dir)\n",
    "    img = cv2.resize(img, dsize=target_size)\n",
    "    img = img.astype(np.float32) / 255.0 # normalize \n",
    "    return img"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "5a214e76",
   "metadata": {
    "executionInfo": {
     "elapsed": 1,
     "status": "ok",
     "timestamp": 1652256064629,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "rw1af4pzz-4P"
   },
   "outputs": [],
   "source": [
    "def create_dataset(img_dir, label, target_size, batch_size=None) :    \n",
    "    img = [] # 원본이미지\n",
    "    \n",
    "    ## 최적화가 필요한 부분\n",
    "    for i in img_dir:\n",
    "        image = load_img_numpy(i, target_size)\n",
    "        img.append(image)\n",
    "        \n",
    "    x = Dataset.from_tensor_slices(img)\n",
    "    y = Dataset.from_tensor_slices(label)\n",
    "    dataset = Dataset.zip((x, y))\n",
    "    \n",
    "    if batch_size is not None:\n",
    "        dataset = dataset.batch(batch_size, drop_remainder=True).repeat()\n",
    "   \n",
    "    return dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "ca101e7f",
   "metadata": {
    "executionInfo": {
     "elapsed": 2,
     "status": "ok",
     "timestamp": 1652256066346,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "QF8v7HJy0AD-"
   },
   "outputs": [],
   "source": [
    "def load_test_img_numpy(img_dir, target_size) :\n",
    "    img = cv2.imread(img_dir)\n",
    "    img = cv2.resize(img, dsize=target_size)\n",
    "    img = img.astype(np.float32) / 255.0\n",
    "    return img\n",
    "\n",
    "def create_test_dataset(img_dir, target_size, batch_size=None) :\n",
    "    file_name = [x.split('/')[-1][:-4] for x in img_dir]\n",
    "    x_img = []\n",
    "\n",
    "    for i in tqdm(img_dir,\n",
    "                total=len(img_dir),\n",
    "                position=0,\n",
    "                leave=True):\n",
    "        img = load_test_img_numpy(i, target_size)\n",
    "        x_img.append(img)\n",
    "\n",
    "    dataset = Dataset.from_tensor_slices(x_img)\n",
    "    \n",
    "    if batch_size is not None :\n",
    "        dataset = dataset.batch(batch_size, drop_remainder=True).repeat()\n",
    "       \n",
    "    return dataset, file_name"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "19a5417d",
   "metadata": {
    "id": "Zxt-_rQT1Eck"
   },
   "source": [
    "## Load Data Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "180fd716",
   "metadata": {
    "executionInfo": {
     "elapsed": 79659,
     "status": "ok",
     "timestamp": 1652256147539,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "14655d2b-3c47-409d-9c1e-8de4753950bd"
   },
   "outputs": [],
   "source": [
    "IMAGE_SIZE = 224\n",
    "BATCH_SIZE = 8\n",
    "LEARNING_RATE = 2e-4\n",
    "\n",
    "all_img_path, all_label = get_train_data('/home/lab05/AI_proj_Team2/AI_proj_Team2/data/train/')\n",
    "test_img_path = get_test_data('/home/lab05/AI_proj_Team2/AI_proj_Team2/data/test')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7ca032e8",
   "metadata": {
    "id": "68577959-091a-46cc-9c9b-18e2404f5de6"
   },
   "source": [
    "## Train / Validation Split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "99437b1f",
   "metadata": {
    "executionInfo": {
     "elapsed": 523,
     "status": "ok",
     "timestamp": 1652256154595,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "31f5a9d4-7af5-466c-b1ac-1480f285dd9c"
   },
   "outputs": [],
   "source": [
    "# Train : Validation = 0.8 : 0.2 Split\n",
    "train_len = int(len(all_img_path)*0.8)\n",
    "\n",
    "train_img_path = all_img_path[:train_len]\n",
    "train_label = all_label[:train_len]\n",
    "\n",
    "vali_img_path = all_img_path[train_len:]\n",
    "vali_label = all_label[train_len:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "76aa5d81",
   "metadata": {
    "executionInfo": {
     "elapsed": 1988962,
     "status": "ok",
     "timestamp": 1652258145694,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "oe3_r_Dy1OEs"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-05-12 16:18:29.596014: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 766488576 exceeds 10% of free system memory.\n"
     ]
    }
   ],
   "source": [
    "train_dataset = create_dataset(train_img_path, train_label, (IMAGE_SIZE, IMAGE_SIZE), BATCH_SIZE)\n",
    "valid_dataset = create_dataset(vali_img_path, vali_label, (IMAGE_SIZE, IMAGE_SIZE), BATCH_SIZE )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "8a4eb2b3",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 450,
     "status": "ok",
     "timestamp": 1652258455903,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "L9bGKCFl1Ox5",
    "outputId": "490eda3b-95ba-47e0-9e6b-4108deb87a85"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<RepeatDataset shapes: ((8, 224, 224, 3), (8,)), types: (tf.float32, tf.float32)>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "6a369a24",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 481,
     "status": "ok",
     "timestamp": 1652258458239,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "UJwaF5Fg1RE9",
    "outputId": "3741f516-0877-47f7-9d16-d714cfc44796"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<RepeatDataset shapes: ((8, 224, 224, 3), (8,)), types: (tf.float32, tf.float32)>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "valid_dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a764e1cf",
   "metadata": {
    "id": "Gy0tYGA968OQ"
   },
   "source": [
    "## VGG16 Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "adca6a20",
   "metadata": {
    "colab": {
     "background_save": true,
     "base_uri": "https://localhost:8080/"
    },
    "id": "1DMQNS6C67IQ"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"vgg19\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         [(None, 224, 224, 3)]     0         \n",
      "_________________________________________________________________\n",
      "block1_conv1 (Conv2D)        (None, 224, 224, 64)      1792      \n",
      "_________________________________________________________________\n",
      "block1_conv2 (Conv2D)        (None, 224, 224, 64)      36928     \n",
      "_________________________________________________________________\n",
      "block1_pool (MaxPooling2D)   (None, 112, 112, 64)      0         \n",
      "_________________________________________________________________\n",
      "block2_conv1 (Conv2D)        (None, 112, 112, 128)     73856     \n",
      "_________________________________________________________________\n",
      "block2_conv2 (Conv2D)        (None, 112, 112, 128)     147584    \n",
      "_________________________________________________________________\n",
      "block2_pool (MaxPooling2D)   (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "block3_conv1 (Conv2D)        (None, 56, 56, 256)       295168    \n",
      "_________________________________________________________________\n",
      "block3_conv2 (Conv2D)        (None, 56, 56, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_conv3 (Conv2D)        (None, 56, 56, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_conv4 (Conv2D)        (None, 56, 56, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_pool (MaxPooling2D)   (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "block4_conv1 (Conv2D)        (None, 28, 28, 512)       1180160   \n",
      "_________________________________________________________________\n",
      "block4_conv2 (Conv2D)        (None, 28, 28, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block4_conv3 (Conv2D)        (None, 28, 28, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block4_conv4 (Conv2D)        (None, 28, 28, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block4_pool (MaxPooling2D)   (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "block5_conv1 (Conv2D)        (None, 14, 14, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv2 (Conv2D)        (None, 14, 14, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv3 (Conv2D)        (None, 14, 14, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv4 (Conv2D)        (None, 14, 14, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block5_pool (MaxPooling2D)   (None, 7, 7, 512)         0         \n",
      "=================================================================\n",
      "Total params: 20,024,384\n",
      "Trainable params: 20,024,384\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "base_model = VGG19(\n",
    "    weights='imagenet',\n",
    "    include_top = False,\n",
    "    input_shape = (IMAGE_SIZE,IMAGE_SIZE) + (3,)\n",
    ")\n",
    "base_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "3758c5ea",
   "metadata": {
    "executionInfo": {
     "elapsed": 5338,
     "status": "ok",
     "timestamp": 1652238481780,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "VfW-vOg3BqMG"
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.regularizers import *\n",
    "\n",
    "LAMBDA_VAL = 0.001\n",
    "\n",
    "# base_model.trainable = False\n",
    "# model = Sequential()\n",
    "# model.add(base_model)\n",
    "# model.add(GlobalAveragePooling2D()) # Flatten??\n",
    "# model.add(Dropout(0.3))\n",
    "# model.add(Dense(64, activation='relu'))\n",
    "# model.add(Dense(1, activation='linear', kernel_regularizer=L2(l2=LAMBDA_VAL)))\n",
    "\n",
    "base_model.trainable = False\n",
    "model = Sequential()\n",
    "\n",
    "model.add(base_model)\n",
    "\n",
    "model.add(Flatten())\n",
    "\n",
    "model.add(Dense(units=256,\n",
    "                activation='relu'))\n",
    "\n",
    "model.add(Dropout(rate=0.5))\n",
    "model.add(Dense(units=128,\n",
    "                activation='relu'))\n",
    "\n",
    "model.add(Dense(units=64,\n",
    "                activation='relu'))\n",
    "\n",
    "model.add(Dropout(rate=0.5))\n",
    "\n",
    "model.add(Dense(units=1,\n",
    "                activation='linear',  kernel_regularizer=L2(l2=LAMBDA_VAL)))\n",
    "\n",
    "\n",
    "# Early Stopping\n",
    "es = EarlyStopping(monitor='val_loss',\n",
    "                   mode='auto',\n",
    "                   patience=5,\n",
    "                   verbose=1)\n",
    "\n",
    "# Checkpoint\n",
    "# model_checkpoint = './{epoch:06d}-{val_acc:0.6f}-{acc:0.6f}.ckpt'\n",
    "\n",
    "# checkpointer = ModelCheckpoint(\n",
    "#     filepath=model_checkpoint,\n",
    "#     verbose=1,\n",
    "#     period=2,\n",
    "#     save_best_weights=True,\n",
    "#     mode='auto',\n",
    "#     monitor='val_acc'\n",
    "# )\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "17686c3b",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1843743,
     "status": "ok",
     "timestamp": 1652240326409,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "YQOu60ItBwB7",
    "outputId": "703a3bff-4457-48a7-a1c3-546dc0b27c54"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-05-12 16:21:21.216962: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 766488576 exceeds 10% of free system memory.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-05-12 16:21:22.376047: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:116] None of the MLIR optimization passes are enabled (registered 2)\n",
      "2022-05-12 16:21:22.397950: I tensorflow/core/platform/profile_utils/cpu_utils.cc:112] CPU Frequency: 2499995000 Hz\n",
      "2022-05-12 16:21:22.493480: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10\n",
      "2022-05-12 16:21:22.718181: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "159/159 [==============================] - 14s 71ms/step - loss: 15402.0079 - mae: 78.0685 - val_loss: 10195.2920 - val_mae: 74.1492\n",
      "Epoch 2/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 10369.7982 - mae: 72.9319 - val_loss: 10173.6641 - val_mae: 72.1017\n",
      "Epoch 3/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 9326.7200 - mae: 69.3372 - val_loss: 10161.9482 - val_mae: 71.6011\n",
      "Epoch 4/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 8653.1208 - mae: 65.9489 - val_loss: 10146.9580 - val_mae: 70.8243\n",
      "Epoch 5/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 8800.7460 - mae: 66.4469 - val_loss: 10132.7119 - val_mae: 71.4767\n",
      "Epoch 6/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 8539.6886 - mae: 65.5547 - val_loss: 10049.8633 - val_mae: 71.9018\n",
      "Epoch 7/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 8241.5557 - mae: 64.5412 - val_loss: 10191.3369 - val_mae: 69.3855\n",
      "Epoch 8/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 8430.9889 - mae: 65.4706 - val_loss: 10063.4033 - val_mae: 70.3317\n",
      "Epoch 9/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 7607.9199 - mae: 62.6535 - val_loss: 9934.0215 - val_mae: 71.4821\n",
      "Epoch 10/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 7874.4159 - mae: 63.2964 - val_loss: 9915.3877 - val_mae: 71.8154\n",
      "Epoch 11/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 7227.7542 - mae: 62.0816 - val_loss: 10017.2881 - val_mae: 70.8005\n",
      "Epoch 12/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 6932.4043 - mae: 60.2623 - val_loss: 9926.1650 - val_mae: 71.2583\n",
      "Epoch 13/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 6517.0435 - mae: 57.6646 - val_loss: 9906.4990 - val_mae: 71.9723\n",
      "Epoch 14/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 6683.2344 - mae: 58.8235 - val_loss: 9835.8145 - val_mae: 72.4876\n",
      "Epoch 15/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 6364.4312 - mae: 57.5751 - val_loss: 9762.1611 - val_mae: 74.0786\n",
      "Epoch 16/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 6345.3428 - mae: 56.9685 - val_loss: 9818.9082 - val_mae: 72.7341\n",
      "Epoch 17/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 5997.9976 - mae: 55.9108 - val_loss: 9764.3467 - val_mae: 73.1735\n",
      "Epoch 18/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 6323.4403 - mae: 56.5769 - val_loss: 9836.6855 - val_mae: 74.0578\n",
      "Epoch 19/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 6645.3855 - mae: 57.1067 - val_loss: 9718.9756 - val_mae: 75.4124\n",
      "Epoch 20/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 6221.8274 - mae: 56.0193 - val_loss: 9803.1445 - val_mae: 72.0770\n",
      "Epoch 21/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 5691.4154 - mae: 53.5743 - val_loss: 9723.9668 - val_mae: 74.4172\n",
      "Epoch 22/30\n",
      "159/159 [==============================] - 11s 70ms/step - loss: 6333.0670 - mae: 56.1970 - val_loss: 9731.4844 - val_mae: 74.7149\n",
      "Epoch 23/30\n",
      "159/159 [==============================] - 11s 70ms/step - loss: 6141.7292 - mae: 54.8604 - val_loss: 9758.8779 - val_mae: 75.6214\n",
      "Epoch 24/30\n",
      "159/159 [==============================] - 11s 70ms/step - loss: 6175.0327 - mae: 54.5812 - val_loss: 9810.3770 - val_mae: 73.8087\n",
      "Epoch 25/30\n",
      "159/159 [==============================] - 11s 70ms/step - loss: 5834.0379 - mae: 54.8031 - val_loss: 9668.4883 - val_mae: 73.9486\n",
      "Epoch 26/30\n",
      "159/159 [==============================] - 11s 71ms/step - loss: 5563.8835 - mae: 53.3092 - val_loss: 9609.4453 - val_mae: 76.1285\n",
      "Epoch 27/30\n",
      "159/159 [==============================] - 11s 71ms/step - loss: 5497.2123 - mae: 51.7249 - val_loss: 9778.9873 - val_mae: 78.5664\n",
      "Epoch 28/30\n",
      "159/159 [==============================] - 11s 71ms/step - loss: 5348.3192 - mae: 52.4972 - val_loss: 9611.9199 - val_mae: 77.2956\n",
      "Epoch 29/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 5609.4736 - mae: 52.5582 - val_loss: 9726.0703 - val_mae: 73.6699\n",
      "Epoch 30/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 5356.7993 - mae: 52.6197 - val_loss: 9811.2656 - val_mae: 74.4214\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer=Adam(learning_rate=LEARNING_RATE),\n",
    "              loss='mse',\n",
    "              metrics=['mae'])\n",
    "\n",
    "\n",
    "EPOCHS = 30\n",
    "STEPS_PER_EPOCH = len(train_img_path) // BATCH_SIZE\n",
    "VAL_STEPS = len(vali_img_path) // BATCH_SIZE\n",
    "\n",
    "history = model.fit(train_dataset, epochs=EPOCHS,\n",
    "                    steps_per_epoch=STEPS_PER_EPOCH,\n",
    "                    validation_data=valid_dataset,\n",
    "                    validation_steps = VAL_STEPS,\n",
    "                    # callbacks=[es, checkpointer],\n",
    "                    verbose=1)\n",
    "\n",
    "model.save('./pretrained_model_efficentnetb7.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "d05c2f22",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 517
    },
    "executionInfo": {
     "elapsed": 867,
     "status": "ok",
     "timestamp": 1652241805175,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "9rY03zPESR6c",
    "outputId": "d9f7bef9-d6f9-4016-e553-228a36235197"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Training and Validation MAE')"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 8))\n",
    "plt.plot(history.history['mae'], label='Training MAE')\n",
    "plt.plot(history.history['val_mae'], label='Validation MAE')\n",
    "plt.legend(loc='lower right')\n",
    "plt.title('Training and Validation MAE')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "13123672",
   "metadata": {
    "id": "Q0LFMCpiSXIt"
   },
   "source": [
    "## Fine-tuning\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "9f1d5b46",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 257,
     "status": "ok",
     "timestamp": 1652242129680,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "pKL31JZJSWQB",
    "outputId": "13018c04-2722-464c-a4ff-a2b943eee41c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of layers in the base model:  22\n"
     ]
    }
   ],
   "source": [
    "base_model.trainable=True\n",
    "print(\"Number of layers in the base model: \", len(base_model.layers))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "a196a5ed",
   "metadata": {
    "executionInfo": {
     "elapsed": 3,
     "status": "ok",
     "timestamp": 1652242130775,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "2zl7ppAjTbM9"
   },
   "outputs": [],
   "source": [
    "fine_tune_at= int(len(base_model.layers)*0.9)\n",
    "\n",
    "for i, layer in enumerate(base_model.layers[:fine_tune_at]):\n",
    "    if 'bn' not in layer.name :\n",
    "        layer.trainable = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "be5f2c75",
   "metadata": {
    "executionInfo": {
     "elapsed": 2,
     "status": "ok",
     "timestamp": 1652242131848,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "YY_tvdjuTdDZ"
   },
   "outputs": [],
   "source": [
    "model.compile(optimizer=Adam(learning_rate=LEARNING_RATE/10),\n",
    "              loss='mse',\n",
    "              metrics=['mae'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "f8ecd09d",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1903664,
     "status": "ok",
     "timestamp": 1652244036465,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "3K_0LwBMTd8j",
    "outputId": "c8712863-a6d5-497a-d6fd-4de838bd19c1"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-05-12 16:27:11.812368: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 766488576 exceeds 10% of free system memory.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "159/159 [==============================] - 31s 183ms/step - loss: 9060.3352 - mae: 64.2138 - val_loss: 11579.2637 - val_mae: 71.5470\n",
      "Epoch 2/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 8556.9802 - mae: 63.8506 - val_loss: 10451.7285 - val_mae: 72.7835\n",
      "Epoch 3/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 7378.4606 - mae: 61.0359 - val_loss: 9879.1201 - val_mae: 74.9753\n",
      "Epoch 4/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 6121.6101 - mae: 55.3408 - val_loss: 10429.1875 - val_mae: 73.6690\n",
      "Epoch 5/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 7078.5073 - mae: 60.4650 - val_loss: 10278.4795 - val_mae: 72.9374\n",
      "Epoch 6/30\n",
      "159/159 [==============================] - 29s 186ms/step - loss: 5391.0802 - mae: 52.5889 - val_loss: 9708.0537 - val_mae: 72.5191\n",
      "Epoch 7/30\n",
      "159/159 [==============================] - 30s 186ms/step - loss: 5525.1322 - mae: 52.3110 - val_loss: 10171.4277 - val_mae: 73.1205\n",
      "Epoch 8/30\n",
      "159/159 [==============================] - 29s 186ms/step - loss: 4764.9260 - mae: 49.7390 - val_loss: 10063.9307 - val_mae: 73.4542\n",
      "Epoch 9/30\n",
      "159/159 [==============================] - 30s 186ms/step - loss: 4797.4152 - mae: 48.8421 - val_loss: 10520.6035 - val_mae: 71.1677\n",
      "Epoch 10/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 4037.3955 - mae: 44.8901 - val_loss: 10405.8711 - val_mae: 72.6801\n",
      "Epoch 11/30\n",
      "159/159 [==============================] - 29s 186ms/step - loss: 4690.1459 - mae: 46.5216 - val_loss: 10538.2979 - val_mae: 72.4393\n",
      "Epoch 12/30\n",
      "159/159 [==============================] - 29s 186ms/step - loss: 3470.1276 - mae: 41.5978 - val_loss: 10861.2080 - val_mae: 75.8787\n",
      "Epoch 13/30\n",
      "159/159 [==============================] - 30s 186ms/step - loss: 4312.7461 - mae: 45.4251 - val_loss: 10301.4551 - val_mae: 71.4906\n",
      "Epoch 14/30\n",
      "159/159 [==============================] - 30s 186ms/step - loss: 3887.0147 - mae: 43.2393 - val_loss: 10904.1152 - val_mae: 73.6532\n",
      "Epoch 15/30\n",
      "159/159 [==============================] - 29s 186ms/step - loss: 3932.6593 - mae: 42.7741 - val_loss: 9813.9795 - val_mae: 72.3750\n",
      "Epoch 16/30\n",
      "159/159 [==============================] - 29s 186ms/step - loss: 4225.8707 - mae: 43.0102 - val_loss: 10772.1943 - val_mae: 70.1954\n",
      "Epoch 17/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 3889.3488 - mae: 42.3056 - val_loss: 10209.1768 - val_mae: 69.5906\n",
      "Epoch 18/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 4025.6494 - mae: 42.4590 - val_loss: 10824.2197 - val_mae: 75.0134\n",
      "Epoch 19/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 3621.0928 - mae: 41.1103 - val_loss: 10839.6045 - val_mae: 72.8431\n",
      "Epoch 20/30\n",
      "159/159 [==============================] - 29s 186ms/step - loss: 4472.7601 - mae: 46.1770 - val_loss: 10415.6982 - val_mae: 69.9247\n",
      "Epoch 21/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 3158.9480 - mae: 38.1063 - val_loss: 10022.6455 - val_mae: 72.5554\n",
      "Epoch 22/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 4161.7772 - mae: 42.8631 - val_loss: 10138.1123 - val_mae: 71.3296\n",
      "Epoch 23/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 3818.8234 - mae: 42.8819 - val_loss: 10721.9512 - val_mae: 71.5559\n",
      "Epoch 24/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 3764.8721 - mae: 41.8709 - val_loss: 10700.4082 - val_mae: 73.6733\n",
      "Epoch 25/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 3103.1039 - mae: 38.0019 - val_loss: 10526.7891 - val_mae: 68.8836\n",
      "Epoch 26/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 3508.6160 - mae: 41.1667 - val_loss: 11117.9668 - val_mae: 67.4141\n",
      "Epoch 27/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 4066.1143 - mae: 42.6713 - val_loss: 11680.2451 - val_mae: 72.6691\n",
      "Epoch 28/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 2611.3389 - mae: 35.7928 - val_loss: 11389.0537 - val_mae: 75.4962\n",
      "Epoch 29/30\n",
      "159/159 [==============================] - 29s 186ms/step - loss: 3267.0329 - mae: 37.7799 - val_loss: 11011.9590 - val_mae: 74.3355\n",
      "Epoch 30/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 3279.9918 - mae: 39.0334 - val_loss: 11637.3779 - val_mae: 69.6243\n"
     ]
    }
   ],
   "source": [
    "FINE_TUNE_EPOCHS = 30 \n",
    "history = model.fit(train_dataset, epochs=FINE_TUNE_EPOCHS,\n",
    "                    steps_per_epoch=STEPS_PER_EPOCH,\n",
    "                    validation_data=valid_dataset,\n",
    "                    validation_steps = VAL_STEPS,\n",
    "                    # callbacks=[es, checkpointer],\n",
    "                    verbose=1)\n",
    "\n",
    "model.save('./fine_tuned_model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "10206bc5",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 517
    },
    "executionInfo": {
     "elapsed": 832,
     "status": "ok",
     "timestamp": 1652244431542,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "JUN1s9zxcJUt",
    "outputId": "0bbb0968-b61d-45ad-bbef-b7ceca953a8e"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Training and Validation fine-tuning MAE')"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 8))\n",
    "plt.plot(history.history['mae'], label='Training MAE')\n",
    "plt.plot(history.history['val_mae'], label='Validation MAE')\n",
    "plt.legend(loc='lower right')\n",
    "plt.title('Training and Validation fine-tuning MAE')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9f1fde47",
   "metadata": {
    "id": "5af7e345-7445-4717-909c-2a2e0a4e8993"
   },
   "source": [
    "## CustomDataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1a4c3a35",
   "metadata": {
    "id": "XKy0OCPdjh7j"
   },
   "source": [
    "---\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c8a55fb0",
   "metadata": {
    "id": "vNGSAvhjjjpe"
   },
   "source": [
    "## 규제 없는 모델\n",
    "&loss = mae"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "0cc61f4b",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 248
    },
    "executionInfo": {
     "elapsed": 490,
     "status": "error",
     "timestamp": 1652253855855,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "h6Nu-dSbjy0O",
    "outputId": "b831c399-ad7e-46c0-a518-aa32d677b460"
   },
   "outputs": [],
   "source": [
    "\n",
    "base_model.trainable = False\n",
    "model2 = Sequential()\n",
    "\n",
    "model2.add(base_model)\n",
    "\n",
    "model2.add(Flatten())\n",
    "\n",
    "model2.add(Dense(units=256,\n",
    "                activation='relu'))\n",
    "\n",
    "model2.add(Dropout(rate=0.5))\n",
    "model2.add(Dense(units=128,\n",
    "                activation='relu'))\n",
    "\n",
    "model2.add(Dense(units=64,\n",
    "                activation='relu'))\n",
    "\n",
    "model2.add(Dropout(rate=0.5))\n",
    "\n",
    "model2.add(Dense(units=1,\n",
    "                activation='linear'))\n",
    "\n",
    "\n",
    "checkpoint_filepath = './no_regul_pretrain_best_model.h5'\n",
    "checkpointer = ModelCheckpoint(\n",
    "    filepath=checkpoint_filepath,\n",
    "    monitor='val_mae',\n",
    "    save_best_only=True,\n",
    "    verbose=1\n",
    ")\n",
    "\n",
    "model2.compile(optimizer=Adam(learning_rate=LEARNING_RATE),\n",
    "              loss='mae',\n",
    "              metrics=['mae'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "04887ae0",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1851169,
     "status": "ok",
     "timestamp": 1652252230179,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "gScaMCKijz8F",
    "outputId": "378db086-9170-48ff-d743-2e9ee5faa90a",
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-05-12 16:43:30.034771: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 766488576 exceeds 10% of free system memory.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "159/159 [==============================] - 12s 71ms/step - loss: 77.2141 - mae: 77.2141 - val_loss: 69.4449 - val_mae: 69.4449\n",
      "\n",
      "Epoch 00001: val_mae improved from inf to 69.44485, saving model to ./no_regul_pretrain_best_model.h5\n",
      "Epoch 2/30\n",
      "159/159 [==============================] - 11s 70ms/step - loss: 70.0335 - mae: 70.0335 - val_loss: 67.5610 - val_mae: 67.5610\n",
      "\n",
      "Epoch 00002: val_mae improved from 69.44485 to 67.56102, saving model to ./no_regul_pretrain_best_model.h5\n",
      "Epoch 3/30\n",
      "159/159 [==============================] - 11s 70ms/step - loss: 65.2288 - mae: 65.2288 - val_loss: 66.6582 - val_mae: 66.6582\n",
      "\n",
      "Epoch 00003: val_mae improved from 67.56102 to 66.65819, saving model to ./no_regul_pretrain_best_model.h5\n",
      "Epoch 4/30\n",
      "159/159 [==============================] - 11s 70ms/step - loss: 58.7927 - mae: 58.7927 - val_loss: 66.2924 - val_mae: 66.2924\n",
      "\n",
      "Epoch 00004: val_mae improved from 66.65819 to 66.29240, saving model to ./no_regul_pretrain_best_model.h5\n",
      "Epoch 5/30\n",
      "159/159 [==============================] - 11s 71ms/step - loss: 55.1110 - mae: 55.1110 - val_loss: 66.7375 - val_mae: 66.7375\n",
      "\n",
      "Epoch 00005: val_mae did not improve from 66.29240\n",
      "Epoch 6/30\n",
      "159/159 [==============================] - 11s 71ms/step - loss: 49.7041 - mae: 49.7041 - val_loss: 68.3732 - val_mae: 68.3732\n",
      "\n",
      "Epoch 00006: val_mae did not improve from 66.29240\n",
      "Epoch 7/30\n",
      "159/159 [==============================] - 11s 71ms/step - loss: 46.3836 - mae: 46.3836 - val_loss: 70.0130 - val_mae: 70.0130\n",
      "\n",
      "Epoch 00007: val_mae did not improve from 66.29240\n",
      "Epoch 8/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 45.7989 - mae: 45.7989 - val_loss: 69.8311 - val_mae: 69.8311\n",
      "\n",
      "Epoch 00008: val_mae did not improve from 66.29240\n",
      "Epoch 9/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 43.4609 - mae: 43.4609 - val_loss: 71.8777 - val_mae: 71.8777\n",
      "\n",
      "Epoch 00009: val_mae did not improve from 66.29240\n",
      "Epoch 10/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 42.5054 - mae: 42.5054 - val_loss: 72.4307 - val_mae: 72.4307\n",
      "\n",
      "Epoch 00010: val_mae did not improve from 66.29240\n",
      "Epoch 11/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 43.4981 - mae: 43.4981 - val_loss: 74.9231 - val_mae: 74.9231\n",
      "\n",
      "Epoch 00011: val_mae did not improve from 66.29240\n",
      "Epoch 12/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 40.1085 - mae: 40.1085 - val_loss: 73.9458 - val_mae: 73.9458\n",
      "\n",
      "Epoch 00012: val_mae did not improve from 66.29240\n",
      "Epoch 13/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 41.0295 - mae: 41.0295 - val_loss: 73.9179 - val_mae: 73.9179\n",
      "\n",
      "Epoch 00013: val_mae did not improve from 66.29240\n",
      "Epoch 14/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 39.1221 - mae: 39.1221 - val_loss: 74.8112 - val_mae: 74.8112\n",
      "\n",
      "Epoch 00014: val_mae did not improve from 66.29240\n",
      "Epoch 15/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 38.3583 - mae: 38.3583 - val_loss: 74.2705 - val_mae: 74.2705\n",
      "\n",
      "Epoch 00015: val_mae did not improve from 66.29240\n",
      "Epoch 16/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 37.0890 - mae: 37.0890 - val_loss: 77.1901 - val_mae: 77.1901\n",
      "\n",
      "Epoch 00016: val_mae did not improve from 66.29240\n",
      "Epoch 17/30\n",
      "159/159 [==============================] - 12s 72ms/step - loss: 37.2364 - mae: 37.2364 - val_loss: 73.4377 - val_mae: 73.4377\n",
      "\n",
      "Epoch 00017: val_mae did not improve from 66.29240\n",
      "Epoch 18/30\n",
      "159/159 [==============================] - 12s 73ms/step - loss: 37.5164 - mae: 37.5164 - val_loss: 73.6379 - val_mae: 73.6379\n",
      "\n",
      "Epoch 00018: val_mae did not improve from 66.29240\n",
      "Epoch 19/30\n",
      "159/159 [==============================] - 12s 73ms/step - loss: 37.6957 - mae: 37.6957 - val_loss: 73.9518 - val_mae: 73.9518\n",
      "\n",
      "Epoch 00019: val_mae did not improve from 66.29240\n",
      "Epoch 20/30\n",
      "159/159 [==============================] - 12s 73ms/step - loss: 36.1970 - mae: 36.1970 - val_loss: 75.2477 - val_mae: 75.2477\n",
      "\n",
      "Epoch 00020: val_mae did not improve from 66.29240\n",
      "Epoch 21/30\n",
      "159/159 [==============================] - 12s 73ms/step - loss: 36.4031 - mae: 36.4031 - val_loss: 74.0539 - val_mae: 74.0539\n",
      "\n",
      "Epoch 00021: val_mae did not improve from 66.29240\n",
      "Epoch 22/30\n",
      "159/159 [==============================] - 12s 73ms/step - loss: 36.1147 - mae: 36.1147 - val_loss: 74.6609 - val_mae: 74.6609\n",
      "\n",
      "Epoch 00022: val_mae did not improve from 66.29240\n",
      "Epoch 23/30\n",
      "159/159 [==============================] - 12s 73ms/step - loss: 35.1042 - mae: 35.1042 - val_loss: 73.4333 - val_mae: 73.4333\n",
      "\n",
      "Epoch 00023: val_mae did not improve from 66.29240\n",
      "Epoch 24/30\n",
      "159/159 [==============================] - 12s 74ms/step - loss: 35.8304 - mae: 35.8304 - val_loss: 74.9657 - val_mae: 74.9657\n",
      "\n",
      "Epoch 00024: val_mae did not improve from 66.29240\n",
      "Epoch 25/30\n",
      "159/159 [==============================] - 12s 74ms/step - loss: 37.3609 - mae: 37.3609 - val_loss: 72.8169 - val_mae: 72.8169\n",
      "\n",
      "Epoch 00025: val_mae did not improve from 66.29240\n",
      "Epoch 26/30\n",
      "159/159 [==============================] - 12s 73ms/step - loss: 36.3175 - mae: 36.3175 - val_loss: 73.2825 - val_mae: 73.2825\n",
      "\n",
      "Epoch 00026: val_mae did not improve from 66.29240\n",
      "Epoch 27/30\n",
      "159/159 [==============================] - 11s 72ms/step - loss: 36.0879 - mae: 36.0879 - val_loss: 74.2474 - val_mae: 74.2474\n",
      "\n",
      "Epoch 00027: val_mae did not improve from 66.29240\n",
      "Epoch 28/30\n",
      "159/159 [==============================] - 11s 70ms/step - loss: 35.5675 - mae: 35.5675 - val_loss: 75.7341 - val_mae: 75.7341\n",
      "\n",
      "Epoch 00028: val_mae did not improve from 66.29240\n",
      "Epoch 29/30\n",
      "159/159 [==============================] - 11s 70ms/step - loss: 37.0959 - mae: 37.0959 - val_loss: 74.7552 - val_mae: 74.7552\n",
      "\n",
      "Epoch 00029: val_mae did not improve from 66.29240\n",
      "Epoch 30/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 33.4595 - mae: 33.4595 - val_loss: 73.3468 - val_mae: 73.3468\n",
      "\n",
      "Epoch 00030: val_mae did not improve from 66.29240\n"
     ]
    }
   ],
   "source": [
    "history2 = model2.fit(train_dataset, epochs=EPOCHS,\n",
    "                    steps_per_epoch=STEPS_PER_EPOCH,\n",
    "                    validation_data=valid_dataset,\n",
    "                    validation_steps = VAL_STEPS,\n",
    "                    callbacks=[checkpointer],\n",
    "                    verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "61fa7989",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 517
    },
    "executionInfo": {
     "elapsed": 913,
     "status": "ok",
     "timestamp": 1652252361496,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "XCYbOGyVj0gR",
    "outputId": "1041a412-c6bb-4012-d400-64582a05114d"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Training and Validation MAE')"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 8))\n",
    "plt.plot(history2.history['mae'], label='Training MAE')\n",
    "plt.plot(history2.history['val_mae'], label='Validation MAE')\n",
    "plt.legend(loc='lower right')\n",
    "plt.title('Training and Validation MAE')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6d92c502",
   "metadata": {
    "id": "-Wyqc55DrUEW"
   },
   "source": [
    "## fine tune"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "e5a69133",
   "metadata": {
    "executionInfo": {
     "elapsed": 288,
     "status": "ok",
     "timestamp": 1652252365245,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "CVGw0lRwrSro"
   },
   "outputs": [],
   "source": [
    "base_model.trainable=True\n",
    "\n",
    "fine_tune_at = int(len(base_model.layers) * 0.9)\n",
    "for layer in base_model.layers[:fine_tune_at] :\n",
    "    layer.trainable = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "21d73e47",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 248
    },
    "executionInfo": {
     "elapsed": 549,
     "status": "error",
     "timestamp": 1652253823083,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "nhMj2JjrrXTk",
    "outputId": "7bdc943d-94b6-45e0-e0be-9f59ef646996"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-05-12 16:49:21.939568: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 766488576 exceeds 10% of free system memory.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "159/159 [==============================] - 28s 174ms/step - loss: 39.6794 - mae: 39.6794 - val_loss: 79.5630 - val_mae: 79.5630\n",
      "\n",
      "Epoch 00001: val_mae improved from inf to 79.56303, saving model to ./no_regul_finetune_best_model.h5\n",
      "Epoch 2/30\n",
      "159/159 [==============================] - 27s 173ms/step - loss: 44.5723 - mae: 44.5723 - val_loss: 72.8138 - val_mae: 72.8138\n",
      "\n",
      "Epoch 00002: val_mae improved from 79.56303 to 72.81383, saving model to ./no_regul_finetune_best_model.h5\n",
      "Epoch 3/30\n",
      "159/159 [==============================] - 27s 173ms/step - loss: 47.0900 - mae: 47.0900 - val_loss: 76.5105 - val_mae: 76.5105\n",
      "\n",
      "Epoch 00003: val_mae did not improve from 72.81383\n",
      "Epoch 4/30\n",
      "159/159 [==============================] - 28s 173ms/step - loss: 42.1234 - mae: 42.1234 - val_loss: 79.2941 - val_mae: 79.2941\n",
      "\n",
      "Epoch 00004: val_mae did not improve from 72.81383\n",
      "Epoch 5/30\n",
      "159/159 [==============================] - 28s 174ms/step - loss: 39.9008 - mae: 39.9008 - val_loss: 77.2580 - val_mae: 77.2580\n",
      "\n",
      "Epoch 00005: val_mae did not improve from 72.81383\n",
      "Epoch 6/30\n",
      "159/159 [==============================] - 28s 175ms/step - loss: 37.4859 - mae: 37.4859 - val_loss: 75.6273 - val_mae: 75.6273\n",
      "\n",
      "Epoch 00006: val_mae did not improve from 72.81383\n",
      "Epoch 7/30\n",
      "159/159 [==============================] - 28s 176ms/step - loss: 39.2531 - mae: 39.2531 - val_loss: 72.6055 - val_mae: 72.6055\n",
      "\n",
      "Epoch 00007: val_mae improved from 72.81383 to 72.60551, saving model to ./no_regul_finetune_best_model.h5\n",
      "Epoch 8/30\n",
      "159/159 [==============================] - 28s 177ms/step - loss: 37.9580 - mae: 37.9580 - val_loss: 76.7889 - val_mae: 76.7889\n",
      "\n",
      "Epoch 00008: val_mae did not improve from 72.60551\n",
      "Epoch 9/30\n",
      "159/159 [==============================] - 29s 180ms/step - loss: 37.0772 - mae: 37.0772 - val_loss: 73.6084 - val_mae: 73.6084\n",
      "\n",
      "Epoch 00009: val_mae did not improve from 72.60551\n",
      "Epoch 10/30\n",
      "159/159 [==============================] - 29s 183ms/step - loss: 37.2342 - mae: 37.2342 - val_loss: 73.6394 - val_mae: 73.6394\n",
      "\n",
      "Epoch 00010: val_mae did not improve from 72.60551\n",
      "Epoch 11/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 37.4201 - mae: 37.4201 - val_loss: 76.7255 - val_mae: 76.7255\n",
      "\n",
      "Epoch 00011: val_mae did not improve from 72.60551\n",
      "Epoch 12/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 37.4093 - mae: 37.4093 - val_loss: 83.5987 - val_mae: 83.5987\n",
      "\n",
      "Epoch 00012: val_mae did not improve from 72.60551\n",
      "Epoch 13/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 34.3409 - mae: 34.3409 - val_loss: 74.3796 - val_mae: 74.3796\n",
      "\n",
      "Epoch 00013: val_mae did not improve from 72.60551\n",
      "Epoch 14/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 33.8939 - mae: 33.8939 - val_loss: 72.5283 - val_mae: 72.5283\n",
      "\n",
      "Epoch 00014: val_mae improved from 72.60551 to 72.52826, saving model to ./no_regul_finetune_best_model.h5\n",
      "Epoch 15/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 34.3066 - mae: 34.3066 - val_loss: 74.6523 - val_mae: 74.6523\n",
      "\n",
      "Epoch 00015: val_mae did not improve from 72.52826\n",
      "Epoch 16/30\n",
      "159/159 [==============================] - 30s 186ms/step - loss: 35.3049 - mae: 35.3049 - val_loss: 74.4264 - val_mae: 74.4264\n",
      "\n",
      "Epoch 00016: val_mae did not improve from 72.52826\n",
      "Epoch 17/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 34.6464 - mae: 34.6464 - val_loss: 76.2117 - val_mae: 76.2117\n",
      "\n",
      "Epoch 00017: val_mae did not improve from 72.52826\n",
      "Epoch 18/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 32.8970 - mae: 32.8970 - val_loss: 78.0069 - val_mae: 78.0069\n",
      "\n",
      "Epoch 00018: val_mae did not improve from 72.52826\n",
      "Epoch 19/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 34.7167 - mae: 34.7167 - val_loss: 74.7290 - val_mae: 74.7290\n",
      "\n",
      "Epoch 00019: val_mae did not improve from 72.52826\n",
      "Epoch 20/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 35.3147 - mae: 35.3147 - val_loss: 70.2595 - val_mae: 70.2595\n",
      "\n",
      "Epoch 00020: val_mae improved from 72.52826 to 70.25954, saving model to ./no_regul_finetune_best_model.h5\n",
      "Epoch 21/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 34.3698 - mae: 34.3698 - val_loss: 74.2782 - val_mae: 74.2782\n",
      "\n",
      "Epoch 00021: val_mae did not improve from 70.25954\n",
      "Epoch 22/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 33.3264 - mae: 33.3264 - val_loss: 75.7438 - val_mae: 75.7438\n",
      "\n",
      "Epoch 00022: val_mae did not improve from 70.25954\n",
      "Epoch 23/30\n",
      "159/159 [==============================] - 29s 185ms/step - loss: 30.2728 - mae: 30.2728 - val_loss: 71.3078 - val_mae: 71.3078\n",
      "\n",
      "Epoch 00023: val_mae did not improve from 70.25954\n",
      "Epoch 24/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 34.1048 - mae: 34.1048 - val_loss: 70.8774 - val_mae: 70.8774\n",
      "\n",
      "Epoch 00024: val_mae did not improve from 70.25954\n",
      "Epoch 25/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 32.5606 - mae: 32.5606 - val_loss: 73.5039 - val_mae: 73.5039\n",
      "\n",
      "Epoch 00025: val_mae did not improve from 70.25954\n",
      "Epoch 26/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 33.7423 - mae: 33.7423 - val_loss: 72.1516 - val_mae: 72.1516\n",
      "\n",
      "Epoch 00026: val_mae did not improve from 70.25954\n",
      "Epoch 27/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 31.3994 - mae: 31.3994 - val_loss: 71.5841 - val_mae: 71.5841\n",
      "\n",
      "Epoch 00027: val_mae did not improve from 70.25954\n",
      "Epoch 28/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 31.2797 - mae: 31.2797 - val_loss: 71.3348 - val_mae: 71.3348\n",
      "\n",
      "Epoch 00028: val_mae did not improve from 70.25954\n",
      "Epoch 29/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 30.6241 - mae: 30.6241 - val_loss: 73.1727 - val_mae: 73.1727\n",
      "\n",
      "Epoch 00029: val_mae did not improve from 70.25954\n",
      "Epoch 30/30\n",
      "159/159 [==============================] - 29s 184ms/step - loss: 32.4703 - mae: 32.4703 - val_loss: 69.0366 - val_mae: 69.0366\n",
      "\n",
      "Epoch 00030: val_mae improved from 70.25954 to 69.03661, saving model to ./no_regul_finetune_best_model.h5\n"
     ]
    }
   ],
   "source": [
    "model2.compile(optimizer=Adam(learning_rate=LEARNING_RATE/10),\n",
    "              loss='mae',\n",
    "              metrics=['mae'])\n",
    "\n",
    "checkpoint_filepath = './no_regul_finetune_best_model.h5'\n",
    "checkpointer = ModelCheckpoint(\n",
    "    filepath=checkpoint_filepath,\n",
    "    monitor='val_mae',\n",
    "    save_best_only=True,\n",
    "    verbose=1\n",
    ")\n",
    "\n",
    "history2 = model2.fit(train_dataset, epochs=EPOCHS,\n",
    "                    steps_per_epoch=STEPS_PER_EPOCH,\n",
    "                    validation_data=valid_dataset,\n",
    "                    validation_steps = VAL_STEPS,\n",
    "                    callbacks=[checkpointer],\n",
    "                    verbose=1)\n",
    "\n",
    "model2.save('./by_chance_noregul_finetunemoedl.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "aac5efc2",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 266
    },
    "executionInfo": {
     "elapsed": 271,
     "status": "error",
     "timestamp": 1652250294400,
     "user": {
      "displayName": "유영재",
      "userId": "06640138067198506258"
     },
     "user_tz": -540
    },
    "id": "4uFocaZeyq1-",
    "outputId": "148b2d62-cd79-403a-e46e-706ba28373ae"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Training and Validation MAE')"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 8))\n",
    "plt.plot(history2.history['mae'], label='Training MAE')\n",
    "plt.plot(history2.history['val_mae'], label='Validation MAE')\n",
    "plt.legend(loc='lower right')\n",
    "plt.title('Training and Validation MAE')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "02331091",
   "metadata": {
    "id": "36_VDyOy56FW"
   },
   "source": [
    "## learning rate 변경"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "1e621e9f",
   "metadata": {
    "id": "-VGCnTwv58YU"
   },
   "outputs": [],
   "source": [
    " base_model = VGG19(\n",
    "    weights='imagenet',\n",
    "    include_top = False,\n",
    "    input_shape = (IMAGE_SIZE,IMAGE_SIZE) + (3,)\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "fba41ada",
   "metadata": {
    "id": "sr5nENK56BLa"
   },
   "outputs": [],
   "source": [
    "\n",
    "LEARNING_RATE = 1e-4\n",
    "\n",
    "base_model.trainable = False\n",
    "model3 = Sequential()\n",
    "\n",
    "model3.add(base_model)\n",
    "\n",
    "model3.add(Flatten())\n",
    "\n",
    "model3.add(Dense(units=256,\n",
    "                activation='relu'))\n",
    "\n",
    "model3.add(Dropout(rate=0.5))\n",
    "model3.add(Dense(units=128,\n",
    "                activation='relu'))\n",
    "\n",
    "model3.add(Dense(units=64,\n",
    "                activation='relu'))\n",
    "\n",
    "model3.add(Dropout(rate=0.5))\n",
    "\n",
    "model3.add(Dense(units=1,\n",
    "                activation='linear'))\n",
    "\n",
    "checkpoint_filepath = './no_regul_pretrain_best_model_lr.h5'\n",
    "\n",
    "checkpointer = ModelCheckpoint(\n",
    "    filepath=checkpoint_filepath,\n",
    "    monitor='val_mae',\n",
    "    save_best_only=True,\n",
    "    verbose=1\n",
    ")\n",
    "\n",
    "model3.compile(optimizer=Adam(learning_rate=LEARNING_RATE),\n",
    "              loss='mae',\n",
    "              metrics=['mae'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "a9477372",
   "metadata": {
    "id": "k6j6V3_k6CPv",
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "159/159 [==============================] - 12s 73ms/step - loss: 77.9060 - mae: 77.9060 - val_loss: 69.6926 - val_mae: 69.6926\n",
      "\n",
      "Epoch 00001: val_mae improved from inf to 69.69263, saving model to ./no_regul_pretrain_best_model_lr.h5\n",
      "Epoch 2/30\n",
      "159/159 [==============================] - 11s 71ms/step - loss: 71.6517 - mae: 71.6517 - val_loss: 68.6489 - val_mae: 68.6489\n",
      "\n",
      "Epoch 00002: val_mae improved from 69.69263 to 68.64894, saving model to ./no_regul_pretrain_best_model_lr.h5\n",
      "Epoch 3/30\n",
      "159/159 [==============================] - 11s 70ms/step - loss: 69.8354 - mae: 69.8354 - val_loss: 67.6897 - val_mae: 67.6897\n",
      "\n",
      "Epoch 00003: val_mae improved from 68.64894 to 67.68970, saving model to ./no_regul_pretrain_best_model_lr.h5\n",
      "Epoch 4/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 67.2082 - mae: 67.2082 - val_loss: 67.4329 - val_mae: 67.4329\n",
      "\n",
      "Epoch 00004: val_mae improved from 67.68970 to 67.43293, saving model to ./no_regul_pretrain_best_model_lr.h5\n",
      "Epoch 5/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 65.6222 - mae: 65.6222 - val_loss: 67.2265 - val_mae: 67.2265\n",
      "\n",
      "Epoch 00005: val_mae improved from 67.43293 to 67.22648, saving model to ./no_regul_pretrain_best_model_lr.h5\n",
      "Epoch 6/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 64.2389 - mae: 64.2389 - val_loss: 66.9562 - val_mae: 66.9562\n",
      "\n",
      "Epoch 00006: val_mae improved from 67.22648 to 66.95618, saving model to ./no_regul_pretrain_best_model_lr.h5\n",
      "Epoch 7/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 63.0802 - mae: 63.0802 - val_loss: 66.2604 - val_mae: 66.2604\n",
      "\n",
      "Epoch 00007: val_mae improved from 66.95618 to 66.26039, saving model to ./no_regul_pretrain_best_model_lr.h5\n",
      "Epoch 8/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 61.5162 - mae: 61.5162 - val_loss: 66.1830 - val_mae: 66.1830\n",
      "\n",
      "Epoch 00008: val_mae improved from 66.26039 to 66.18302, saving model to ./no_regul_pretrain_best_model_lr.h5\n",
      "Epoch 9/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 63.2182 - mae: 63.2182 - val_loss: 66.0256 - val_mae: 66.0256\n",
      "\n",
      "Epoch 00009: val_mae improved from 66.18302 to 66.02563, saving model to ./no_regul_pretrain_best_model_lr.h5\n",
      "Epoch 10/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 60.3663 - mae: 60.3663 - val_loss: 66.1182 - val_mae: 66.1182\n",
      "\n",
      "Epoch 00010: val_mae did not improve from 66.02563\n",
      "Epoch 11/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 61.2711 - mae: 61.2711 - val_loss: 66.2028 - val_mae: 66.2028\n",
      "\n",
      "Epoch 00011: val_mae did not improve from 66.02563\n",
      "Epoch 12/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 61.9559 - mae: 61.9559 - val_loss: 66.0501 - val_mae: 66.0501\n",
      "\n",
      "Epoch 00012: val_mae did not improve from 66.02563\n",
      "Epoch 13/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 61.3023 - mae: 61.3023 - val_loss: 66.0598 - val_mae: 66.0598\n",
      "\n",
      "Epoch 00013: val_mae did not improve from 66.02563\n",
      "Epoch 14/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 59.2212 - mae: 59.2212 - val_loss: 66.0866 - val_mae: 66.0866\n",
      "\n",
      "Epoch 00014: val_mae did not improve from 66.02563\n",
      "Epoch 15/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 59.1315 - mae: 59.1315 - val_loss: 66.1859 - val_mae: 66.1859\n",
      "\n",
      "Epoch 00015: val_mae did not improve from 66.02563\n",
      "Epoch 16/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 57.9881 - mae: 57.9881 - val_loss: 66.2372 - val_mae: 66.2372\n",
      "\n",
      "Epoch 00016: val_mae did not improve from 66.02563\n",
      "Epoch 17/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 59.5475 - mae: 59.5475 - val_loss: 66.3480 - val_mae: 66.3480\n",
      "\n",
      "Epoch 00017: val_mae did not improve from 66.02563\n",
      "Epoch 18/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 57.5293 - mae: 57.5293 - val_loss: 66.3829 - val_mae: 66.3829\n",
      "\n",
      "Epoch 00018: val_mae did not improve from 66.02563\n",
      "Epoch 19/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 57.0970 - mae: 57.0970 - val_loss: 66.3139 - val_mae: 66.3139\n",
      "\n",
      "Epoch 00019: val_mae did not improve from 66.02563\n",
      "Epoch 20/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 58.3546 - mae: 58.3546 - val_loss: 66.7961 - val_mae: 66.7961\n",
      "\n",
      "Epoch 00020: val_mae did not improve from 66.02563\n",
      "Epoch 21/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 56.5696 - mae: 56.5696 - val_loss: 66.1442 - val_mae: 66.1442\n",
      "\n",
      "Epoch 00021: val_mae did not improve from 66.02563\n",
      "Epoch 22/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 55.7321 - mae: 55.7321 - val_loss: 66.6083 - val_mae: 66.6083\n",
      "\n",
      "Epoch 00022: val_mae did not improve from 66.02563\n",
      "Epoch 23/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 54.3580 - mae: 54.3580 - val_loss: 66.8413 - val_mae: 66.8413\n",
      "\n",
      "Epoch 00023: val_mae did not improve from 66.02563\n",
      "Epoch 24/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 55.9451 - mae: 55.9451 - val_loss: 66.8258 - val_mae: 66.8258\n",
      "\n",
      "Epoch 00024: val_mae did not improve from 66.02563\n",
      "Epoch 25/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 53.0686 - mae: 53.0686 - val_loss: 66.5152 - val_mae: 66.5152\n",
      "\n",
      "Epoch 00025: val_mae did not improve from 66.02563\n",
      "Epoch 26/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 51.3486 - mae: 51.3486 - val_loss: 67.0256 - val_mae: 67.0256\n",
      "\n",
      "Epoch 00026: val_mae did not improve from 66.02563\n",
      "Epoch 27/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 56.6761 - mae: 56.6761 - val_loss: 67.2402 - val_mae: 67.2402\n",
      "\n",
      "Epoch 00027: val_mae did not improve from 66.02563\n",
      "Epoch 28/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 53.0920 - mae: 53.0920 - val_loss: 67.3981 - val_mae: 67.3981\n",
      "\n",
      "Epoch 00028: val_mae did not improve from 66.02563\n",
      "Epoch 29/30\n",
      "159/159 [==============================] - 11s 69ms/step - loss: 53.1715 - mae: 53.1715 - val_loss: 67.2077 - val_mae: 67.2077\n",
      "\n",
      "Epoch 00029: val_mae did not improve from 66.02563\n",
      "Epoch 30/30\n",
      "159/159 [==============================] - 11s 68ms/step - loss: 53.2444 - mae: 53.2444 - val_loss: 67.0961 - val_mae: 67.0961\n",
      "\n",
      "Epoch 00030: val_mae did not improve from 66.02563\n"
     ]
    }
   ],
   "source": [
    "history3 = model3.fit(train_dataset, epochs=EPOCHS,\n",
    "                    steps_per_epoch=STEPS_PER_EPOCH,\n",
    "                    validation_data=valid_dataset,\n",
    "                    validation_steps = VAL_STEPS,\n",
    "                    callbacks=[checkpointer],\n",
    "                    verbose=1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "80385112",
   "metadata": {
    "id": "wD99_MNv61Rt"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Training and Validation MAE')"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 8))\n",
    "plt.plot(history3.history['mae'], label='Training MAE')\n",
    "plt.plot(history3.history['val_mae'], label='Validation MAE')\n",
    "plt.legend(loc='lower right')\n",
    "plt.title('Training and Validation MAE')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3b162f74",
   "metadata": {
    "id": "EBgpEiFv64x5"
   },
   "source": [
    "## finetune"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "3c88a22c",
   "metadata": {
    "id": "SbDG_59c63_v"
   },
   "outputs": [],
   "source": [
    "base_model.trainable=True\n",
    "\n",
    "fine_tune_at = int(len(base_model.layers) * 0.9)\n",
    "for layer in base_model.layers[:fine_tune_at] :\n",
    "    layer.trainable = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "cb44e05d",
   "metadata": {
    "id": "RXfeTGdW660X",
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "159/159 [==============================] - 10s 59ms/step - loss: 50.4791 - mae: 50.4791 - val_loss: 67.9750 - val_mae: 67.9750\n",
      "\n",
      "Epoch 00001: val_mae improved from inf to 67.97500, saving model to ./no_regul_pretrain_best_model_lr.h5\n",
      "Epoch 2/30\n",
      "159/159 [==============================] - 9s 58ms/step - loss: 49.8406 - mae: 49.8406 - val_loss: 67.8893 - val_mae: 67.8893\n",
      "\n",
      "Epoch 00002: val_mae improved from 67.97500 to 67.88928, saving model to ./no_regul_pretrain_best_model_lr.h5\n",
      "Epoch 3/30\n",
      "159/159 [==============================] - 9s 58ms/step - loss: 51.5339 - mae: 51.5339 - val_loss: 67.9384 - val_mae: 67.9384\n",
      "\n",
      "Epoch 00003: val_mae did not improve from 67.88928\n",
      "Epoch 4/30\n",
      "159/159 [==============================] - 9s 58ms/step - loss: 52.5414 - mae: 52.5414 - val_loss: 68.0061 - val_mae: 68.0061\n",
      "\n",
      "Epoch 00004: val_mae did not improve from 67.88928\n",
      "Epoch 5/30\n",
      "159/159 [==============================] - 9s 58ms/step - loss: 51.6303 - mae: 51.6303 - val_loss: 68.0880 - val_mae: 68.0880\n",
      "\n",
      "Epoch 00005: val_mae did not improve from 67.88928\n",
      "Epoch 6/30\n",
      "159/159 [==============================] - 9s 58ms/step - loss: 50.1697 - mae: 50.1697 - val_loss: 68.0943 - val_mae: 68.0943\n",
      "\n",
      "Epoch 00006: val_mae did not improve from 67.88928\n",
      "Epoch 7/30\n",
      "159/159 [==============================] - 9s 58ms/step - loss: 49.5657 - mae: 49.5657 - val_loss: 68.2062 - val_mae: 68.2062\n",
      "\n",
      "Epoch 00007: val_mae did not improve from 67.88928\n",
      "Epoch 8/30\n",
      "159/159 [==============================] - 9s 58ms/step - loss: 51.4606 - mae: 51.4606 - val_loss: 68.2261 - val_mae: 68.2261\n",
      "\n",
      "Epoch 00008: val_mae did not improve from 67.88928\n",
      "Epoch 9/30\n",
      "159/159 [==============================] - 9s 58ms/step - loss: 51.2561 - mae: 51.2561 - val_loss: 68.2093 - val_mae: 68.2093\n",
      "\n",
      "Epoch 00009: val_mae did not improve from 67.88928\n",
      "Epoch 10/30\n",
      "159/159 [==============================] - 9s 57ms/step - loss: 52.1225 - mae: 52.1225 - val_loss: 68.1897 - val_mae: 68.1897\n",
      "\n",
      "Epoch 00010: val_mae did not improve from 67.88928\n",
      "Epoch 11/30\n",
      "159/159 [==============================] - 9s 56ms/step - loss: 50.3511 - mae: 50.3511 - val_loss: 68.2543 - val_mae: 68.2543\n",
      "\n",
      "Epoch 00011: val_mae did not improve from 67.88928\n",
      "Epoch 12/30\n",
      "159/159 [==============================] - 9s 55ms/step - loss: 50.9508 - mae: 50.9508 - val_loss: 68.3283 - val_mae: 68.3283\n",
      "\n",
      "Epoch 00012: val_mae did not improve from 67.88928\n",
      "Epoch 13/30\n",
      "159/159 [==============================] - 9s 55ms/step - loss: 50.4271 - mae: 50.4271 - val_loss: 68.3181 - val_mae: 68.3181\n",
      "\n",
      "Epoch 00013: val_mae did not improve from 67.88928\n",
      "Epoch 14/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 50.9970 - mae: 50.9970 - val_loss: 68.2771 - val_mae: 68.2771\n",
      "\n",
      "Epoch 00014: val_mae did not improve from 67.88928\n",
      "Epoch 15/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 50.0092 - mae: 50.0092 - val_loss: 68.2880 - val_mae: 68.2880\n",
      "\n",
      "Epoch 00015: val_mae did not improve from 67.88928\n",
      "Epoch 16/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 51.4989 - mae: 51.4989 - val_loss: 68.2703 - val_mae: 68.2703\n",
      "\n",
      "Epoch 00016: val_mae did not improve from 67.88928\n",
      "Epoch 17/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 51.7361 - mae: 51.7361 - val_loss: 68.1981 - val_mae: 68.1981\n",
      "\n",
      "Epoch 00017: val_mae did not improve from 67.88928\n",
      "Epoch 18/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 50.2432 - mae: 50.2432 - val_loss: 68.3534 - val_mae: 68.3534\n",
      "\n",
      "Epoch 00018: val_mae did not improve from 67.88928\n",
      "Epoch 19/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 50.3892 - mae: 50.3892 - val_loss: 68.4295 - val_mae: 68.4295\n",
      "\n",
      "Epoch 00019: val_mae did not improve from 67.88928\n",
      "Epoch 20/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 49.6221 - mae: 49.6221 - val_loss: 68.3590 - val_mae: 68.3590\n",
      "\n",
      "Epoch 00020: val_mae did not improve from 67.88928\n",
      "Epoch 21/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 51.4010 - mae: 51.4010 - val_loss: 68.3429 - val_mae: 68.3429\n",
      "\n",
      "Epoch 00021: val_mae did not improve from 67.88928\n",
      "Epoch 22/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 50.9087 - mae: 50.9087 - val_loss: 68.4314 - val_mae: 68.4314\n",
      "\n",
      "Epoch 00022: val_mae did not improve from 67.88928\n",
      "Epoch 23/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 49.7808 - mae: 49.7808 - val_loss: 68.3358 - val_mae: 68.3358\n",
      "\n",
      "Epoch 00023: val_mae did not improve from 67.88928\n",
      "Epoch 24/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 50.1011 - mae: 50.1011 - val_loss: 68.3940 - val_mae: 68.3940\n",
      "\n",
      "Epoch 00024: val_mae did not improve from 67.88928\n",
      "Epoch 25/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 50.6430 - mae: 50.6430 - val_loss: 68.3791 - val_mae: 68.3791\n",
      "\n",
      "Epoch 00025: val_mae did not improve from 67.88928\n",
      "Epoch 26/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 50.1072 - mae: 50.1072 - val_loss: 68.3486 - val_mae: 68.3486\n",
      "\n",
      "Epoch 00026: val_mae did not improve from 67.88928\n",
      "Epoch 27/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 50.9211 - mae: 50.9211 - val_loss: 68.2861 - val_mae: 68.2861\n",
      "\n",
      "Epoch 00027: val_mae did not improve from 67.88928\n",
      "Epoch 28/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 49.4521 - mae: 49.4521 - val_loss: 68.2705 - val_mae: 68.2705\n",
      "\n",
      "Epoch 00028: val_mae did not improve from 67.88928\n",
      "Epoch 29/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 50.1627 - mae: 50.1627 - val_loss: 68.2837 - val_mae: 68.2837\n",
      "\n",
      "Epoch 00029: val_mae did not improve from 67.88928\n",
      "Epoch 30/30\n",
      "159/159 [==============================] - 9s 54ms/step - loss: 48.2689 - mae: 48.2689 - val_loss: 68.3904 - val_mae: 68.3904\n",
      "\n",
      "Epoch 00030: val_mae did not improve from 67.88928\n"
     ]
    }
   ],
   "source": [
    "checkpoint_filepath = './no_regul_pretrain_best_model_lr.h5'\n",
    "checkpointer = ModelCheckpoint(\n",
    "    filepath=checkpoint_filepath,\n",
    "    monitor='val_mae',\n",
    "    save_best_only=True,\n",
    "    verbose=1\n",
    ")\n",
    "\n",
    "model3.compile(optimizer=Adam(learning_rate=LEARNING_RATE/10),\n",
    "              loss='mae',\n",
    "              metrics=['mae'])\n",
    "\n",
    "history3 = model3.fit(train_dataset, epochs=EPOCHS,\n",
    "                    steps_per_epoch=STEPS_PER_EPOCH,\n",
    "                    validation_data=valid_dataset,\n",
    "                    validation_steps = VAL_STEPS,\n",
    "                    callbacks=[checkpointer],\n",
    "                    verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "40de0362",
   "metadata": {
    "id": "NiX_z_z47GNa"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Training and Validation MAE')"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 8))\n",
    "plt.plot(history3.history['mae'], label='Training MAE')\n",
    "plt.plot(history3.history['val_mae'], label='Validation MAE')\n",
    "plt.legend(loc='lower right')\n",
    "plt.title('Training and Validation MAE')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "20efeb36",
   "metadata": {
    "id": "TiwRqdlj7Hb9"
   },
   "source": [
    "## Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "d4bce8f8",
   "metadata": {
    "id": "-4-CouK_7JGQ"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "460"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(test_img_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "286c6624",
   "metadata": {
    "id": "qCeTjfpN7LI6"
   },
   "outputs": [],
   "source": [
    "PRED_BATCH_SIZE = 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "289784d0",
   "metadata": {
    "id": "oEVEmeMt7Mmp"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0d5d147c0299424fbd06d05a25cebcdd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/460 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_dataset, file_names = create_test_dataset(test_img_path, (IMAGE_SIZE, IMAGE_SIZE), PRED_BATCH_SIZE )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "f03763ed",
   "metadata": {
    "id": "DbhuyKjM7OsU"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Error in loading the saved optimizer state. As a result, your model is starting with a freshly initialized optimizer.\n"
     ]
    }
   ],
   "source": [
    "eff_model = tf.keras.models.load_model('./no_regul_pretrain_best_model_lr.h5')\n",
    "\n",
    "STEPS = len(test_img_path) / PRED_BATCH_SIZE\n",
    "prediction = eff_model.predict(test_dataset,\n",
    "                            batch_size = PRED_BATCH_SIZE,\n",
    "                            steps=STEPS)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ab2a50e3",
   "metadata": {
    "id": "eb00479c-a650-4b38-bd21-d4c1018ba57e"
   },
   "source": [
    "## Submission"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "911e0367",
   "metadata": {
    "id": "d169965e-8066-464e-b755-9a2db0d018b6"
   },
   "outputs": [],
   "source": [
    "submission = pd.read_csv('/home/lab05/AI_proj_Team2/AI_proj_Team2/data/sample_submission.csv')\n",
    "submission['leaf_weight'] = prediction.squeeze()\n",
    "submission.to_csv('./submit_VGG16_CNN.csv', index=False)\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "222e750d",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "[Baseline]_Simple CNN.ipynb",
   "version": ""
  },
  "kernelspec": {
   "display_name": "Python [conda env:machine] *",
   "language": "python",
   "name": "conda-env-machine-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}