train_with_existing-data+KHG-data.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import matplotlib as plt\n",
    "import h5py\n",
    "from PIL import Image\n",
    "import cv2\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'/home/vividlee/Desktop/YeloCar'"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "os.getcwd()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['dataset_hsv_deep4pc_test4_all.hdf5', 'dataset_hsv_01.hdf5']\n"
     ]
    }
   ],
   "source": [
    "os.getcwd()\n",
    "os.chdir('/home/vividlee/Desktop/YeloCar/HDF5/deep4pc_test5/')\n",
    "\n",
    "hdf5_files = os.listdir()\n",
    "print(hdf5_files)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'/home/vividlee/Desktop/YeloCar/HDF5/deep4pc_test5'"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "os.getcwd()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dataset_hsv_deep4pc_test4_all.hdf5\n",
      "dataset_hsv_01.hdf5\n",
      "write in /Model/Model_test5_existing-data+KHG-data/dataset_hsv_test5_existing-data+KHG-data.hdf5\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# downloading KHG0data\n",
    "\n",
    "# existing-data : dataset_hsv_deep4pc_test4_all.hdf5\n",
    "# KHG-data : dataset_hsv_01.hdf5\n",
    "# merged data including 'existing-data+KHG-data'\n",
    "\n",
    "def createFolder(directory):\n",
    "    try:\n",
    "        if not os.path.exists(directory):\n",
    "            os.makedirs(directory)\n",
    "    except OSError:\n",
    "        print ('Error: Creating directory. ' +  directory)\n",
    "        \n",
    "test_name = 'test5_'\n",
    "merged_data = 'existing-data+KHG-data'\n",
    "\n",
    "image_train = None\n",
    "action_train = None\n",
    "image_val = None\n",
    "action_val = None\n",
    "    \n",
    "for hdf5_file in hdf5_files :\n",
    "    print(hdf5_file)\n",
    "    \n",
    "    with h5py.File('./' + hdf5_file, 'r') as hf :\n",
    "        if image_train is None:\n",
    "            image_train = np.array(hf['x_train'][:, :, :, :])\n",
    "            action_train = np.array(hf['y_train'][:, :])\n",
    "            image_val = np.array(hf['x_val'][:, :, :, :])\n",
    "            action_val = np.array(hf['y_val'][:, :])\n",
    "        else:\n",
    "            image_train = np.concatenate((image_train, hf['x_train']), axis=0)\n",
    "            action_train = np.concatenate((action_train, hf['y_train']), axis=0)\n",
    "            image_val = np.concatenate((image_val, hf['x_val']), axis=0)\n",
    "            action_val = np.concatenate((action_val, hf['y_val']), axis=0)\n",
    "\n",
    "train_len = len(image_train)\n",
    "val_len = len(image_val)\n",
    "\n",
    "os.chdir('/home/vividlee/Desktop/YeloCar')\n",
    "os.getcwd()\n",
    "\n",
    "model_directory = 'Model_' + test_name + merged_data\n",
    "createFolder('./Model/' + model_directory)\n",
    " \n",
    "with h5py.File('./Model/' + model_directory + '/dataset_hsv_' + test_name + merged_data + '.hdf5', 'w') as hf:\n",
    "    print(\"write in /Model/\"  + model_directory + \"/dataset_hsv_\" + test_name +  merged_data + \".hdf5\\n\")\n",
    "    dset_x_train = hf.create_dataset('x_train', (train_len, 48, 64, 3), chunks=True, compression=\"gzip\")\n",
    "    dset_y_train = hf.create_dataset('y_train', (train_len, 2), chunks=True, compression=\"gzip\")\n",
    "    dset_x_val = hf.create_dataset('x_val', (val_len, 48, 64, 3), chunks=True, compression=\"gzip\")\n",
    "    dset_y_val = hf.create_dataset('y_val', (val_len, 2), chunks=True, compression=\"gzip\")\n",
    "\n",
    "    dset_x_train[:, :, :, :] = image_train\n",
    "    dset_y_train[:, :] = action_train\n",
    "    dset_x_val[:, :, :, :] = image_val\n",
    "    dset_y_val[:, :] = action_val"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-02-21 16:22:06.419153: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /home/vividlee/anaconda3/envs/py38/lib/python3.8/site-packages/cv2/../../lib64:\n",
      "2022-02-21 16:22:06.419249: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcublas.so.11'; dlerror: libcublas.so.11: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /home/vividlee/anaconda3/envs/py38/lib/python3.8/site-packages/cv2/../../lib64:\n",
      "2022-02-21 16:22:06.419324: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcublasLt.so.11'; dlerror: libcublasLt.so.11: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /home/vividlee/anaconda3/envs/py38/lib/python3.8/site-packages/cv2/../../lib64:\n",
      "2022-02-21 16:22:06.447392: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcusolver.so.11'; dlerror: libcusolver.so.11: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /home/vividlee/anaconda3/envs/py38/lib/python3.8/site-packages/cv2/../../lib64:\n",
      "2022-02-21 16:22:06.447599: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcusparse.so.11'; dlerror: libcusparse.so.11: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /home/vividlee/anaconda3/envs/py38/lib/python3.8/site-packages/cv2/../../lib64:\n",
      "2022-02-21 16:22:06.448968: W tensorflow/core/common_runtime/gpu/gpu_device.cc:1850] Cannot dlopen some GPU libraries. Please make sure the missing libraries mentioned above are installed properly if you would like to use GPU. Follow the guide at https://www.tensorflow.org/install/gpu for how to download and setup the required libraries for your platform.\n",
      "Skipping registering GPU devices...\n",
      "2022-02-21 16:22:06.449994: I tensorflow/core/platform/cpu_feature_guard.cc:151] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2022-02-21 16:22:06.529184: W tensorflow/core/framework/cpu_allocator_impl.cc:82] Allocation of 9471909888 exceeds 10% of free system memory.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " conv2d (Conv2D)             (None, 46, 62, 32)        896       \n",
      "                                                                 \n",
      " max_pooling2d (MaxPooling2D  (None, 23, 31, 32)       0         \n",
      " )                                                               \n",
      "                                                                 \n",
      " conv2d_1 (Conv2D)           (None, 21, 29, 32)        9248      \n",
      "                                                                 \n",
      " max_pooling2d_1 (MaxPooling  (None, 10, 14, 32)       0         \n",
      " 2D)                                                             \n",
      "                                                                 \n",
      " conv2d_2 (Conv2D)           (None, 8, 12, 32)         9248      \n",
      "                                                                 \n",
      " max_pooling2d_2 (MaxPooling  (None, 4, 6, 32)         0         \n",
      " 2D)                                                             \n",
      "                                                                 \n",
      " flatten (Flatten)           (None, 768)               0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 128)               98432     \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 64)                8256      \n",
      "                                                                 \n",
      " dense_2 (Dense)             (None, 2)                 130       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 126,210\n",
      "Trainable params: 126,210\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0599 - mae: 0.0599 - mse: 0.0096\n",
      "Epoch 00001: val_mae improved from inf to 0.06644, saving model to ./Model/Model_test5_existing-data+KHG-data/model_1_p5_01_0.0599_0.0664\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-02-21 16:24:58.868980: W tensorflow/python/util/util.cc:368] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: ./Model/Model_test5_existing-data+KHG-data/model_1_p5_01_0.0599_0.0664/assets\n",
      "1004/1004 [==============================] - 169s 167ms/step - loss: 0.0599 - mae: 0.0599 - mse: 0.0096 - val_loss: 0.0664 - val_mae: 0.0664 - val_mse: 0.0107\n",
      "Epoch 2/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0509 - mae: 0.0509 - mse: 0.0078\n",
      "Epoch 00002: val_mae improved from 0.06644 to 0.06590, saving model to ./Model/Model_test5_existing-data+KHG-data/model_1_p5_02_0.0509_0.0659\n",
      "INFO:tensorflow:Assets written to: ./Model/Model_test5_existing-data+KHG-data/model_1_p5_02_0.0509_0.0659/assets\n",
      "1004/1004 [==============================] - 167s 166ms/step - loss: 0.0509 - mae: 0.0509 - mse: 0.0078 - val_loss: 0.0659 - val_mae: 0.0659 - val_mse: 0.0106\n",
      "Epoch 3/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0489 - mae: 0.0489 - mse: 0.0074\n",
      "Epoch 00003: val_mae improved from 0.06590 to 0.06547, saving model to ./Model/Model_test5_existing-data+KHG-data/model_1_p5_03_0.0489_0.0655\n",
      "INFO:tensorflow:Assets written to: ./Model/Model_test5_existing-data+KHG-data/model_1_p5_03_0.0489_0.0655/assets\n",
      "1004/1004 [==============================] - 169s 168ms/step - loss: 0.0489 - mae: 0.0489 - mse: 0.0074 - val_loss: 0.0655 - val_mae: 0.0655 - val_mse: 0.0101\n",
      "Epoch 4/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0476 - mae: 0.0476 - mse: 0.0071\n",
      "Epoch 00004: val_mae improved from 0.06547 to 0.06431, saving model to ./Model/Model_test5_existing-data+KHG-data/model_1_p5_04_0.0476_0.0643\n",
      "INFO:tensorflow:Assets written to: ./Model/Model_test5_existing-data+KHG-data/model_1_p5_04_0.0476_0.0643/assets\n",
      "1004/1004 [==============================] - 168s 168ms/step - loss: 0.0476 - mae: 0.0476 - mse: 0.0071 - val_loss: 0.0643 - val_mae: 0.0643 - val_mse: 0.0100\n",
      "Epoch 5/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0465 - mae: 0.0465 - mse: 0.0068\n",
      "Epoch 00005: val_mae did not improve from 0.06431\n",
      "1004/1004 [==============================] - 167s 166ms/step - loss: 0.0465 - mae: 0.0465 - mse: 0.0068 - val_loss: 0.0652 - val_mae: 0.0652 - val_mse: 0.0104\n",
      "Epoch 6/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0456 - mae: 0.0456 - mse: 0.0066\n",
      "Epoch 00006: val_mae did not improve from 0.06431\n",
      "1004/1004 [==============================] - 167s 167ms/step - loss: 0.0456 - mae: 0.0456 - mse: 0.0066 - val_loss: 0.0647 - val_mae: 0.0647 - val_mse: 0.0098\n",
      "Epoch 7/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0448 - mae: 0.0448 - mse: 0.0065\n",
      "Epoch 00007: val_mae did not improve from 0.06431\n",
      "1004/1004 [==============================] - 168s 167ms/step - loss: 0.0448 - mae: 0.0448 - mse: 0.0065 - val_loss: 0.0653 - val_mae: 0.0653 - val_mse: 0.0106\n",
      "Epoch 8/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0441 - mae: 0.0441 - mse: 0.0063\n",
      "Epoch 00008: val_mae did not improve from 0.06431\n",
      "1004/1004 [==============================] - 168s 167ms/step - loss: 0.0441 - mae: 0.0441 - mse: 0.0063 - val_loss: 0.0653 - val_mae: 0.0653 - val_mse: 0.0106\n",
      "Epoch 9/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0435 - mae: 0.0435 - mse: 0.0062\n",
      "Epoch 00009: val_mae did not improve from 0.06431\n",
      "1004/1004 [==============================] - 167s 166ms/step - loss: 0.0435 - mae: 0.0435 - mse: 0.0062 - val_loss: 0.0655 - val_mae: 0.0655 - val_mse: 0.0104\n",
      "Epoch 10/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0429 - mae: 0.0429 - mse: 0.0061\n",
      "Epoch 00010: val_mae did not improve from 0.06431\n",
      "1004/1004 [==============================] - 168s 167ms/step - loss: 0.0429 - mae: 0.0429 - mse: 0.0061 - val_loss: 0.0652 - val_mae: 0.0652 - val_mse: 0.0104\n",
      "Epoch 11/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0425 - mae: 0.0425 - mse: 0.0060\n",
      "Epoch 00011: val_mae did not improve from 0.06431\n",
      "1004/1004 [==============================] - 168s 167ms/step - loss: 0.0425 - mae: 0.0425 - mse: 0.0060 - val_loss: 0.0644 - val_mae: 0.0644 - val_mse: 0.0100\n",
      "Epoch 12/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0420 - mae: 0.0420 - mse: 0.0059\n",
      "Epoch 00012: val_mae did not improve from 0.06431\n",
      "1004/1004 [==============================] - 168s 168ms/step - loss: 0.0420 - mae: 0.0420 - mse: 0.0059 - val_loss: 0.0651 - val_mae: 0.0651 - val_mse: 0.0101\n",
      "Epoch 13/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0416 - mae: 0.0416 - mse: 0.0058\n",
      "Epoch 00013: val_mae did not improve from 0.06431\n",
      "1004/1004 [==============================] - 168s 167ms/step - loss: 0.0416 - mae: 0.0416 - mse: 0.0058 - val_loss: 0.0667 - val_mae: 0.0667 - val_mse: 0.0108\n",
      "Epoch 14/100\n",
      "1004/1004 [==============================] - ETA: 0s - loss: 0.0412 - mae: 0.0412 - mse: 0.0057\n",
      "Epoch 00014: val_mae did not improve from 0.06431\n",
      "1004/1004 [==============================] - 169s 169ms/step - loss: 0.0412 - mae: 0.0412 - mse: 0.0057 - val_loss: 0.0649 - val_mae: 0.0649 - val_mse: 0.0102\n",
      "\n",
      "\n",
      "2014/2014 [==============================] - 12s 6ms/step - loss: 0.0649 - mae: 0.0649 - mse: 0.0102\n",
      "[0.06490463763475418, 0.06490463763475418, 0.010246477089822292]\n",
      "[0.06644020974636078, 0.06590058654546738, 0.06546829640865326, 0.06431181728839874, 0.06523092091083527, 0.06465049088001251, 0.06532075256109238, 0.06532999128103256, 0.06547509133815765, 0.06517685949802399, 0.06435021013021469, 0.06510477513074875, 0.06665048748254776, 0.06490464508533478]\n",
      "\n",
      "\n",
      "[0.05993957072496414, 0.05088336393237114, 0.04892105236649513, 0.04760449752211571, 0.046500492841005325, 0.045623794198036194, 0.04482612758874893, 0.044140271842479706, 0.0434737503528595, 0.042932868003845215, 0.042483504861593246, 0.042019087821245193, 0.04163806140422821, 0.04120723158121109]\n",
      "\n",
      "\n",
      "[0.06644020974636078, 0.06590058654546738, 0.06546829640865326, 0.06431181728839874, 0.06523092091083527, 0.06465049088001251, 0.06532075256109238, 0.06532999128103256, 0.06547509133815765, 0.06517685949802399, 0.06435021013021469, 0.06510477513074875, 0.06665048748254776, 0.06490464508533478]\n",
      "\n",
      "\n",
      "[0.05993957072496414, 0.05088336393237114, 0.04892105236649513, 0.04760449752211571, 0.046500492841005325, 0.045623794198036194, 0.04482612758874893, 0.044140271842479706, 0.0434737503528595, 0.042932868003845215, 0.042483504861593246, 0.042019087821245193, 0.04163806140422821, 0.04120723158121109]\n",
      "\n",
      "\n"
     ]
    },
    {
     "ename": "NameError",
     "evalue": "name 'numpy' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m/tmp/ipykernel_30004/34237705.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     83\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_mae\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     84\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'\\n'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 85\u001b[0;31m \u001b[0mx_len\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnumpy\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_loss\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     86\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_loss\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     87\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'\\n'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'numpy' is not defined"
     ]
    }
   ],
   "source": [
    "os.chdir('/home/vividlee/Desktop/YeloCar')\n",
    "os.getcwd()\n",
    "\n",
    "def createFolder(directory):\n",
    "    try:\n",
    "        if not os.path.exists(directory):\n",
    "            os.makedirs(directory)\n",
    "    except OSError:\n",
    "        print ('Error: Creating directory. ' +  directory)\n",
    "\n",
    "dataset_version = 'deep4pc_test5'\n",
    "model_version = 'model_' + '1_p5'\n",
    "\n",
    "# createFolder('./Test/' + model_version)\n",
    "\n",
    "with h5py.File('./Model/Model_test5_existing-data+KHG-data/dataset_hsv_test5_existing-data+KHG-data.hdf5', 'r') as hf:\n",
    "    image_train = np.array(hf['x_train'][:, :, :, :])\n",
    "    action_train = np.array(hf['y_train'][:, :])\n",
    "    image_val = np.array(hf['x_val'][:, :, :, :])\n",
    "    action_val = np.array(hf['y_val'][:, :])\n",
    "    \n",
    "    \n",
    "(x_train, y_train), (x_test, y_test) = (image_train, action_train), (image_val, action_val)\n",
    "\n",
    "x_train /= 255.\n",
    "x_test /= 255.\n",
    "\n",
    "model = tf.keras.models.Sequential([\n",
    "    tf.keras.layers.Conv2D(32, 3, activation='relu', input_shape=(48, 64, 3)),\n",
    "    tf.keras.layers.MaxPooling2D(),\n",
    "    tf.keras.layers.Conv2D(32, 3, activation='relu'),\n",
    "    tf.keras.layers.MaxPooling2D(),\n",
    "    tf.keras.layers.Conv2D(32, 3, activation='relu'),\n",
    "    tf.keras.layers.MaxPooling2D(),\n",
    "    tf.keras.layers.Flatten(),\n",
    "    tf.keras.layers.Dense(128, activation='relu'),\n",
    "    tf.keras.layers.Dense(64, activation='relu'),\n",
    "    tf.keras.layers.Dense(2)\n",
    "])\n",
    "\n",
    "optimizer = tf.keras.optimizers.RMSprop(0.001)\n",
    "\n",
    "model.compile(loss='mae', optimizer=optimizer, metrics=['mae', 'mse'])\n",
    "\n",
    "model.summary()\n",
    "\n",
    "model_path = './Model/Model_test5_existing-data+KHG-data' + '/' + model_version + '_{epoch:02d}_{loss:.4f}_{val_mae:.4f}'\n",
    "\n",
    "checkpointer = tf.keras.callbacks.ModelCheckpoint(filepath=model_path, monitor='val_mae', verbose=1, save_best_only=True)\n",
    "early_stopping_callback = tf.keras.callbacks.EarlyStopping(monitor='val_mae', patience=10)\n",
    "\n",
    "history = model.fit(x=x_train, y=y_train, batch_size=256, epochs=100, validation_data=(x_test, y_test), callbacks=[early_stopping_callback, checkpointer])\n",
    "\n",
    "print(\"\\n\")\n",
    "print(model.evaluate(x_test, y_test))\n",
    "\n",
    "# outputs = model(x_test[0:20])\n",
    "\n",
    "# print(outputs)\n",
    "\n",
    "# f = open('./Test/' + model_version + '/' + model_version + '_test.txt', 'a')\n",
    "# f.write(\"{}\".format(outputs))\n",
    "\n",
    "# x_test *= 255\n",
    "    \n",
    "# for i in range (0, 20) :\n",
    "#     cv2.imwrite('./Test/' + model_version + '/'+ model_path + '_' + str(i) + '.jpeg', x_test[i])\n",
    "\n",
    "# plotting\n",
    "# trainset : blue, line\n",
    "# testset : red, dot_line\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.06644020974636078, 0.06590058654546738, 0.06546829640865326, 0.06431181728839874, 0.06523092091083527, 0.06465049088001251, 0.06532075256109238, 0.06532999128103256, 0.06547509133815765, 0.06517685949802399, 0.06435021013021469, 0.06510477513074875, 0.06665048748254776, 0.06490464508533478]\n",
      "\n",
      "\n",
      "[0.05993957072496414, 0.05088336393237114, 0.04892105236649513, 0.04760449752211571, 0.046500492841005325, 0.045623794198036194, 0.04482612758874893, 0.044140271842479706, 0.0434737503528595, 0.042932868003845215, 0.042483504861593246, 0.042019087821245193, 0.04163806140422821, 0.04120723158121109]\n",
      "\n",
      "\n",
      "[0.06644020974636078, 0.06590058654546738, 0.06546829640865326, 0.06431181728839874, 0.06523092091083527, 0.06465049088001251, 0.06532075256109238, 0.06532999128103256, 0.06547509133815765, 0.06517685949802399, 0.06435021013021469, 0.06510477513074875, 0.06665048748254776, 0.06490464508533478]\n",
      "\n",
      "\n",
      "[0.05993957072496414, 0.05088336393237114, 0.04892105236649513, 0.04760449752211571, 0.046500492841005325, 0.045623794198036194, 0.04482612758874893, 0.044140271842479706, 0.0434737503528595, 0.042932868003845215, 0.042483504861593246, 0.042019087821245193, 0.04163806140422821, 0.04120723158121109]\n",
      "\n",
      "\n",
      "14\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<function matplotlib.pyplot.show(close=None, block=None)>"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_vloss = history.history['val_loss']   # testset\n",
    "y_loss = history.history['loss']        # trainset\n",
    "print(y_vloss)\n",
    "print('\\n')\n",
    "print(y_loss)\n",
    "print('\\n')\n",
    "y_vmae = history.history['val_mae']\n",
    "y_mae = history.history['mae']\n",
    "print(y_vmae)\n",
    "print('\\n')\n",
    "print(y_mae)\n",
    "print('\\n')\n",
    "x_len = np.arange(len(y_loss))\n",
    "print(len(y_loss))\n",
    "print('\\n')\n",
    "\n",
    "plt.figure(figsize=(10, 8))\n",
    "plt.plot(x_len+1, y_vloss, 'ro--', label='val_loss')\n",
    "plt.plot(x_len+1, y_loss, 'bo-', label='loss')\n",
    "plt.xlabel('Epoch', fontsize=14)\n",
    "plt.ylabel('Loss', fontsize=14)\n",
    "plt.legend(loc='upper right', fontsize=14)\n",
    "plt.xlim([0.5, len(y_loss)+0.5])\n",
    "plt.ylim([0.03, 0.08])\n",
    "plt.xticks([1, 3, 5, 7, 9, 11, 13, 15],fontsize=13)\n",
    "plt.yticks(fontsize=13)\n",
    "plt.grid()\n",
    "plt.show()\n",
    "\n",
    "plt.figure(figsize=(10, 8))\n",
    "plt.plot(x_len+1, y_vmae, 'ro--', label='val_mae')\n",
    "plt.plot(x_len+1, y_mae, 'bo-', label='mae')\n",
    "plt.xlabel('Epoch', fontsize=14)\n",
    "plt.ylabel('MAE', fontsize=14)\n",
    "plt.legend(loc='upper right', fontsize=14)\n",
    "plt.xlim([0.5, len(y_loss)+0.5])\n",
    "plt.ylim([0.03, 0.08])\n",
    "plt.xticks([1, 3, 5, 7, 9, 11, 13, 15], fontsize=13)\n",
    "plt.yticks(fontsize=13)\n",
    "plt.grid()\n",
    "plt.show"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}