Performance_test.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\allace\\Anaconda3\\lib\\site-packages\\h5py\\__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import tensorflow as tf\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import json\n",
    "from tensorflow.python.keras.models import Model, load_model\n",
    "from PIL import Image\n",
    "PATH = \"../tubs/tub6_more_data\" #Shadow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.8.0\n"
     ]
    }
   ],
   "source": [
    "#make sure tensorflow version is the same as one installed on donkeycar (1.8.0)\n",
    "print(tf.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from tensorflow.python.client import device_lib\n",
    "import numpy as np\n",
    "\n",
    "def get_available_gpus():\n",
    "    local_device_protos = device_lib.list_local_devices()\n",
    "    return [x.name for x in local_device_protos if x.device_type == 'GPU']\n",
    "get_available_gpus()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#create df from path\n",
    "def df_from_jsons(PATH, max_items=None):\n",
    "    \n",
    "    #Get json files from path\n",
    "    json_files = np.array([file for file in os.listdir(PATH) if file.endswith('json') ])\n",
    "    json_files = json_files[1:]\n",
    "    np.random.shuffle(json_files)\n",
    "    \n",
    "    df = pd.DataFrame()\n",
    "    for i in json_files[1:max_items+1]:\n",
    "        with open(f'{PATH}/{i}') as f:\n",
    "            data = json.load(f)\n",
    "\n",
    "            userThrottle = data[\"user/throttle\"]\n",
    "            userAngle = data[\"user/angle\"]\n",
    "            camImage_path = data[\"cam/image_array\"]\n",
    "            index = int(camImage_path[:-21])\n",
    "            camImage_path =  PATH + \"/\" + camImage_path\n",
    "            df_row = pd.DataFrame([[userThrottle, userAngle, camImage_path]], index = [index], \n",
    "                                  columns = [\"user/throttle\", \"user/angle\", \"cam/image_array\"])\n",
    "\n",
    "            df = df.append(df_row)\n",
    "    df = df.sort_index()\n",
    "    print(\"Loaded {} json files\".format(len(json_files[1:max_items])))\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#df_test = df_from_jsons(PATH, max_items=8000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#df_test.to_feather('tmp/df_test')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\allace\\Anaconda3\\lib\\site-packages\\pandas\\io\\feather_format.py:112: FutureWarning: `nthreads` argument is deprecated, pass `use_threads` instead\n",
      "  return feather.read_dataframe(path, nthreads=nthreads)\n"
     ]
    }
   ],
   "source": [
    "df_test = pd.read_feather('tmp/df_test')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(7999, 4)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>index</th>\n",
       "      <th>user/throttle</th>\n",
       "      <th>user/angle</th>\n",
       "      <th>cam/image_array</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>9</td>\n",
       "      <td>0.44</td>\n",
       "      <td>0.01</td>\n",
       "      <td>../tubs/tub6_more_data/9_cam-image_array_.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>20</td>\n",
       "      <td>0.34</td>\n",
       "      <td>0.01</td>\n",
       "      <td>../tubs/tub6_more_data/20_cam-image_array_.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>26</td>\n",
       "      <td>0.06</td>\n",
       "      <td>-0.17</td>\n",
       "      <td>../tubs/tub6_more_data/26_cam-image_array_.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>28</td>\n",
       "      <td>0.08</td>\n",
       "      <td>-0.20</td>\n",
       "      <td>../tubs/tub6_more_data/28_cam-image_array_.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>31</td>\n",
       "      <td>0.12</td>\n",
       "      <td>-0.25</td>\n",
       "      <td>../tubs/tub6_more_data/31_cam-image_array_.jpg</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   index  user/throttle  user/angle  \\\n",
       "0      9           0.44        0.01   \n",
       "1     20           0.34        0.01   \n",
       "2     26           0.06       -0.17   \n",
       "3     28           0.08       -0.20   \n",
       "4     31           0.12       -0.25   \n",
       "\n",
       "                                  cam/image_array  \n",
       "0   ../tubs/tub6_more_data/9_cam-image_array_.jpg  \n",
       "1  ../tubs/tub6_more_data/20_cam-image_array_.jpg  \n",
       "2  ../tubs/tub6_more_data/26_cam-image_array_.jpg  \n",
       "3  ../tubs/tub6_more_data/28_cam-image_array_.jpg  \n",
       "4  ../tubs/tub6_more_data/31_cam-image_array_.jpg  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model = load_model(\"./models/speedup3dcnn/pilot_3Dspeedup_2Dense\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "img_in (InputLayer)             (None, 3, 120, 160,  0                                            \n",
      "__________________________________________________________________________________________________\n",
      "cropping3d_6 (Cropping3D)       (None, 3, 60, 160, 3 0           img_in[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv3d_6 (Conv3D)               (None, 1, 14, 39, 4) 904         cropping3d_6[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling3d_6 (MaxPooling3D)  (None, 1, 7, 19, 4)  0           conv3d_6[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_6 (BatchNor (None, 1, 7, 19, 4)  16          max_pooling3d_6[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "dropout_13 (Dropout)            (None, 1, 7, 19, 4)  0           batch_normalization_6[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "flattened (Flatten)             (None, 532)          0           dropout_13[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_8 (Dense)                 (None, 100)          53300       flattened[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_14 (Dropout)            (None, 100)          0           dense_8[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_9 (Dense)                 (None, 50)           5050        dropout_14[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dropout_15 (Dropout)            (None, 50)           0           dense_9[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "angle_out (Dense)               (None, 15)           765         dropout_15[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "throttle_out (Dense)            (None, 1)            51          dropout_15[0][0]                 \n",
      "==================================================================================================\n",
      "Total params: 60,086\n",
      "Trainable params: 60,078\n",
      "Non-trainable params: 8\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_prev_img_names(img_name, quantity, skipFrame = 1):\n",
    "    img_names = []\n",
    "    last_name = img_name\n",
    "    for i in range(0, quantity):\n",
    "        path = img_name.split(\"/\")\n",
    "        index = int(path[-1][:-21])-i*skipFrame\n",
    "        if index < 0:\n",
    "            name = last_name\n",
    "        else:\n",
    "            path[-1] = str(index) + \"_cam-image_array_.jpg\"\n",
    "            name = \"/\".join(path)\n",
    "        \n",
    "        img_names.append(name)\n",
    "        last_name = name\n",
    "    img_names.reverse()\n",
    "    return img_names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['1_cam-image_array_.jpg', '2_cam-image_array_.jpg', '3_cam-image_array_.jpg']"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_prev_img_names(\"3_cam-image_array_.jpg\", 3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_record(record_dict, path = PATH, stacked = True):\n",
    "    data = {}\n",
    "    for key, val in record_dict.items():\n",
    "        if key == 'cam/image_array':\n",
    "            if stacked == True:\n",
    "                img_names = get_prev_img_names(val, 3)\n",
    "                val_array = []\n",
    "                for name in img_names:\n",
    "                    img_array = np.array(Image.open(name))\n",
    "                    \n",
    "                    val_array.append(img_array)\n",
    "                val = np.array(val_array)\n",
    "            else:\n",
    "                img = Image.open(val)\n",
    "                val = np.array(img)\n",
    "        data[key] = val\n",
    "    return data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "record_dict = df_test.sample(n=1).to_dict(orient='record')[0]\n",
    "record = read_record(record_dict)\n",
    "img_arrays = record['cam/image_array']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3, 120, 160, 3)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img_arrays.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=RGB size=160x120 at 0x2118F369198>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image.fromarray(img_arrays[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_record_gen(record_transform=None, shuffle=True, df=None):\n",
    "    while True:\n",
    "        for _ in df.iterrows():\n",
    "            if shuffle:\n",
    "                record_dict = df.sample(n=1).to_dict(orient='record')[0]\n",
    "            \n",
    "            record_dict = read_record(record_dict)\n",
    "            \n",
    "            if record_transform:\n",
    "                record_dict = record_transform(record_dict)\n",
    "            \n",
    "            yield record_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_batch_gen(keys=None, batch_size=128, record_transform=None, shuffle=True, df=None):\n",
    "    \n",
    "    record_gen = get_record_gen(record_transform=record_transform, shuffle=shuffle, df=df)\n",
    "    \n",
    "    while True:\n",
    "        record_list = [next(record_gen) for _ in range(batch_size)]\n",
    "        \n",
    "        batch_arrays = {}\n",
    "        for i, k in enumerate(keys):\n",
    "            arr = np.array([r[k] for r in record_list])\n",
    "            batch_arrays[k] = arr\n",
    "        yield batch_arrays"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_train_gen(X_keys, Y_keys, batch_size=128, record_transform=None, df=None): #\n",
    "    batch_gen = get_batch_gen(X_keys + Y_keys, batch_size=batch_size,\n",
    "                                  record_transform=record_transform, df=df)\n",
    "    while True:\n",
    "        batch = next(batch_gen)\n",
    "        X = [batch[k] for k in X_keys]\n",
    "        Y = [batch[k] for k in Y_keys] \n",
    "        yield X, Y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_train_val_gen(df, X_keys, Y_keys, batch_size = 128, train_frac=0.8, #\n",
    "                      train_record_transform=None, val_record_transform=None): #\n",
    "    train_df = df.sample(frac=train_frac, random_state=200) #\n",
    "    val_df = df.drop(train_df.index) #\n",
    "    \n",
    "    train_gen = get_train_gen(X_keys=X_keys, Y_keys=Y_keys, batch_size=batch_size,\n",
    "                             record_transform=train_record_transform, df=train_df)\n",
    "    val_gen = get_train_gen(X_keys=X_keys, Y_keys=Y_keys, batch_size=batch_size,\n",
    "                           record_transform=val_record_transform, df=val_df)\n",
    "    return train_gen, val_gen"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "def record_transform(record):\n",
    "    record['user/angle'] = to_bin(record['user/angle'])\n",
    "    return record"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_keys = ['cam/image_array']\n",
    "y_keys = ['user/angle', 'user/throttle']\n",
    "bs = 128 # or 128/3\n",
    "train_frac = 1\n",
    "new_model_path = 'models/pilot_jp_2'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "def to_bin(a):\n",
    "    arr = np.zeros(15)\n",
    "    a = a + 1\n",
    "    b = round(a/(2/14))\n",
    "    arr[int(b)] = 1\n",
    "    return arr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_gen, _ = get_train_val_gen(df_test, X_keys, y_keys,\n",
    "                                               train_record_transform=record_transform, #\n",
    "                                               val_record_transform=record_transform, #\n",
    "                                               batch_size=bs, #\n",
    "                                               train_frac=train_frac) #"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['loss', 'angle_out_loss', 'throttle_out_loss']"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.metrics_names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[2.1641774678230288, 2.4039510011672975, 0.062162109911441804]"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate_generator(test_gen, steps=50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}