SegNet.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# SegNet Model Definition"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from __future__ import print_function\n",
    "from collections import OrderedDict\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torchvision.models as models\n",
    "import pprint\n",
    "\n",
    "F = nn.functional\n",
    "DEBUG = False\n",
    "\n",
    "\n",
    "vgg16_dims = [\n",
    "                    (64, 64, 'M'),                                # Stage - 1\n",
    "                    (128, 128, 'M'),                              # Stage - 2\n",
    "                    (256, 256, 256,'M'),                          # Stage - 3\n",
    "                    (512, 512, 512, 'M'),                         # Stage - 4\n",
    "                    (512, 512, 512, 'M')                          # Stage - 5\n",
    "            ]\n",
    "\n",
    "decoder_dims = [\n",
    "                    ('U', 512, 512, 512),                         # Stage - 5\n",
    "                    ('U', 512, 512, 512),                         # Stage - 4\n",
    "                    ('U', 256, 256, 256),                         # Stage - 3\n",
    "                    ('U', 128, 128),                              # Stage - 2\n",
    "                    ('U', 64, 64)                                 # Stage - 1\n",
    "                ]\n",
    "\n",
    "\n",
    "class SegNet(nn.Module):\n",
    "    def __init__(self, input_channels, output_channels):\n",
    "        super(SegNet, self).__init__()\n",
    "\n",
    "        self.input_channels = input_channels\n",
    "        self.output_channels = output_channels\n",
    "\n",
    "        self.num_channels = input_channels\n",
    "\n",
    "        #self.vgg16 = models.vgg16(pretrained=True)\n",
    "\n",
    "\n",
    "        # Encoder layers\n",
    "\n",
    "        self.encoder_conv_00 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=self.input_channels,\n",
    "                                                          out_channels=64,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(64)\n",
    "                                                ])\n",
    "        self.encoder_conv_01 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=64,\n",
    "                                                          out_channels=64,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(64)\n",
    "                                                ])\n",
    "        self.encoder_conv_10 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=64,\n",
    "                                                          out_channels=128,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(128)\n",
    "                                                ])\n",
    "        self.encoder_conv_11 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=128,\n",
    "                                                          out_channels=128,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(128)\n",
    "                                                ])\n",
    "        self.encoder_conv_20 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=128,\n",
    "                                                          out_channels=256,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(256)\n",
    "                                                ])\n",
    "        self.encoder_conv_21 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=256,\n",
    "                                                          out_channels=256,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(256)\n",
    "                                                ])\n",
    "        self.encoder_conv_22 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=256,\n",
    "                                                          out_channels=256,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(256)\n",
    "                                                ])\n",
    "        self.encoder_conv_30 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=256,\n",
    "                                                          out_channels=512,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(512)\n",
    "                                                ])\n",
    "        self.encoder_conv_31 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=512,\n",
    "                                                          out_channels=512,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(512)\n",
    "                                                ])\n",
    "        self.encoder_conv_32 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=512,\n",
    "                                                          out_channels=512,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(512)\n",
    "                                                ])\n",
    "        self.encoder_conv_40 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=512,\n",
    "                                                          out_channels=512,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(512)\n",
    "                                                ])\n",
    "        self.encoder_conv_41 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=512,\n",
    "                                                          out_channels=512,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(512)\n",
    "                                                ])\n",
    "        self.encoder_conv_42 = nn.Sequential(*[\n",
    "                                                nn.Conv2d(in_channels=512,\n",
    "                                                          out_channels=512,\n",
    "                                                          kernel_size=3,\n",
    "                                                          padding=1),\n",
    "                                                nn.BatchNorm2d(512)\n",
    "                                                ])\n",
    "\n",
    "        #self.init_vgg_weigts()\n",
    "\n",
    "        # Decoder layers\n",
    "\n",
    "        self.decoder_convtr_42 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=512,\n",
    "                                                                   out_channels=512,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(512)\n",
    "                                               ])\n",
    "        self.decoder_convtr_41 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=512,\n",
    "                                                                   out_channels=512,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(512)\n",
    "                                               ])\n",
    "        self.decoder_convtr_40 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=512,\n",
    "                                                                   out_channels=512,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(512)\n",
    "                                               ])\n",
    "        self.decoder_convtr_32 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=512,\n",
    "                                                                   out_channels=512,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(512)\n",
    "                                               ])\n",
    "        self.decoder_convtr_31 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=512,\n",
    "                                                                   out_channels=512,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(512)\n",
    "                                               ])\n",
    "        self.decoder_convtr_30 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=512,\n",
    "                                                                   out_channels=256,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(256)\n",
    "                                               ])\n",
    "        self.decoder_convtr_22 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=256,\n",
    "                                                                   out_channels=256,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(256)\n",
    "                                               ])\n",
    "        self.decoder_convtr_21 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=256,\n",
    "                                                                   out_channels=256,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(256)\n",
    "                                               ])\n",
    "        self.decoder_convtr_20 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=256,\n",
    "                                                                   out_channels=128,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(128)\n",
    "                                               ])\n",
    "        self.decoder_convtr_11 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=128,\n",
    "                                                                   out_channels=128,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(128)\n",
    "                                               ])\n",
    "        self.decoder_convtr_10 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=128,\n",
    "                                                                   out_channels=64,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(64)\n",
    "                                               ])\n",
    "        self.decoder_convtr_01 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=64,\n",
    "                                                                   out_channels=64,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1),\n",
    "                                                nn.BatchNorm2d(64)\n",
    "                                               ])\n",
    "        self.decoder_convtr_00 = nn.Sequential(*[\n",
    "                                                nn.ConvTranspose2d(in_channels=64,\n",
    "                                                                   out_channels=self.output_channels,\n",
    "                                                                   kernel_size=3,\n",
    "                                                                   padding=1)\n",
    "                                               ])\n",
    "\n",
    "\n",
    "    def forward(self, input_img):\n",
    "        \"\"\"\n",
    "        Forward pass `input_img` through the network\n",
    "        \"\"\"\n",
    "\n",
    "        # Encoder\n",
    "\n",
    "        # Encoder Stage - 1\n",
    "        dim_0 = input_img.size()\n",
    "        x_00 = F.relu(self.encoder_conv_00(input_img))\n",
    "        x_01 = F.relu(self.encoder_conv_01(x_00))\n",
    "        x_0, indices_0 = F.max_pool2d(x_01, kernel_size=2, stride=2, return_indices=True)\n",
    "\n",
    "        # Encoder Stage - 2\n",
    "        dim_1 = x_0.size()\n",
    "        x_10 = F.relu(self.encoder_conv_10(x_0))\n",
    "        x_11 = F.relu(self.encoder_conv_11(x_10))\n",
    "        x_1, indices_1 = F.max_pool2d(x_11, kernel_size=2, stride=2, return_indices=True)\n",
    "\n",
    "        # Encoder Stage - 3\n",
    "        dim_2 = x_1.size()\n",
    "        x_20 = F.relu(self.encoder_conv_20(x_1))\n",
    "        x_21 = F.relu(self.encoder_conv_21(x_20))\n",
    "        x_22 = F.relu(self.encoder_conv_22(x_21))\n",
    "        x_2, indices_2 = F.max_pool2d(x_22, kernel_size=2, stride=2, return_indices=True)\n",
    "\n",
    "        # Encoder Stage - 4\n",
    "        dim_3 = x_2.size()\n",
    "        x_30 = F.relu(self.encoder_conv_30(x_2))\n",
    "        x_31 = F.relu(self.encoder_conv_31(x_30))\n",
    "        x_32 = F.relu(self.encoder_conv_32(x_31))\n",
    "        x_3, indices_3 = F.max_pool2d(x_32, kernel_size=2, stride=2, return_indices=True)\n",
    "\n",
    "        # Encoder Stage - 5\n",
    "        dim_4 = x_3.size()\n",
    "        x_40 = F.relu(self.encoder_conv_40(x_3))\n",
    "        x_41 = F.relu(self.encoder_conv_41(x_40))\n",
    "        x_42 = F.relu(self.encoder_conv_42(x_41))\n",
    "        x_4, indices_4 = F.max_pool2d(x_42, kernel_size=2, stride=2, return_indices=True)\n",
    "\n",
    "        # Decoder\n",
    "\n",
    "        dim_d = x_4.size()\n",
    "\n",
    "        # Decoder Stage - 5\n",
    "        x_4d = F.max_unpool2d(x_4, indices_4, kernel_size=2, stride=2, output_size=dim_4)\n",
    "        x_42d = F.relu(self.decoder_convtr_42(x_4d))\n",
    "        x_41d = F.relu(self.decoder_convtr_41(x_42d))\n",
    "        x_40d = F.relu(self.decoder_convtr_40(x_41d))\n",
    "        dim_4d = x_40d.size()\n",
    "\n",
    "        # Decoder Stage - 4\n",
    "        x_3d = F.max_unpool2d(x_40d, indices_3, kernel_size=2, stride=2, output_size=dim_3)\n",
    "        x_32d = F.relu(self.decoder_convtr_32(x_3d))\n",
    "        x_31d = F.relu(self.decoder_convtr_31(x_32d))\n",
    "        x_30d = F.relu(self.decoder_convtr_30(x_31d))\n",
    "        dim_3d = x_30d.size()\n",
    "\n",
    "        # Decoder Stage - 3\n",
    "        x_2d = F.max_unpool2d(x_30d, indices_2, kernel_size=2, stride=2, output_size=dim_2)\n",
    "        x_22d = F.relu(self.decoder_convtr_22(x_2d))\n",
    "        x_21d = F.relu(self.decoder_convtr_21(x_22d))\n",
    "        x_20d = F.relu(self.decoder_convtr_20(x_21d))\n",
    "        dim_2d = x_20d.size()\n",
    "\n",
    "        # Decoder Stage - 2\n",
    "        x_1d = F.max_unpool2d(x_20d, indices_1, kernel_size=2, stride=2, output_size=dim_1)\n",
    "        x_11d = F.relu(self.decoder_convtr_11(x_1d))\n",
    "        x_10d = F.relu(self.decoder_convtr_10(x_11d))\n",
    "        dim_1d = x_10d.size()\n",
    "\n",
    "        # Decoder Stage - 1\n",
    "        x_0d = F.max_unpool2d(x_10d, indices_0, kernel_size=2, stride=2, output_size=dim_0)\n",
    "        x_01d = F.relu(self.decoder_convtr_01(x_0d))\n",
    "        x_00d = self.decoder_convtr_00(x_01d)\n",
    "        dim_0d = x_00d.size()\n",
    "\n",
    "        x_softmax = F.softmax(x_00d, dim=1)\n",
    "\n",
    "\n",
    "        if DEBUG:\n",
    "            print(\"dim_0: {}\".format(dim_0))\n",
    "            print(\"dim_1: {}\".format(dim_1))\n",
    "            print(\"dim_2: {}\".format(dim_2))\n",
    "            print(\"dim_3: {}\".format(dim_3))\n",
    "            print(\"dim_4: {}\".format(dim_4))\n",
    "\n",
    "            print(\"dim_d: {}\".format(dim_d))\n",
    "            print(\"dim_4d: {}\".format(dim_4d))\n",
    "            print(\"dim_3d: {}\".format(dim_3d))\n",
    "            print(\"dim_2d: {}\".format(dim_2d))\n",
    "            print(\"dim_1d: {}\".format(dim_1d))\n",
    "            print(\"dim_0d: {}\".format(dim_0d))\n",
    "\n",
    "\n",
    "        return x_00d, x_softmax"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Dataset Class Definition"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import os\n",
    "import torch\n",
    "from torch.utils.data import Dataset\n",
    "from PIL import Image\n",
    "\n",
    "CLASSES = {'background', 'person'}\n",
    "\n",
    "PALETTE = {\n",
    "    (0,   0,   0)   : 0 ,   # background\n",
    "    (255, 255, 255) : 1 ,   # person\n",
    "}\n",
    "\n",
    "class UTPDataset(Dataset):\n",
    "    \"\"\"Unite The People 20017 Dataset\"\"\"\n",
    "    def __init__(self, img_dir, transform=None):\n",
    "        self.transform = transform\n",
    "        self.image_root_dir = img_dir\n",
    "        \n",
    "        self.img_extension = '_full.png'\n",
    "        self.mask_extension = '_segmentation_full.png'\n",
    "        self.start_index = 0\n",
    "        self.end_index = 10000\n",
    "\n",
    "    def __getitem__(self, index):\n",
    "        if not isinstance(index, int):\n",
    "            index = index.item()\n",
    "        \n",
    "        index += self.start_index\n",
    "        image_id = str(index).zfill(5)\n",
    "        image_path = os.path.join(self.image_root_dir, image_id + self.img_extension)\n",
    "        mask_path = os.path.join(self.image_root_dir, image_id + self.mask_extension)\n",
    "\n",
    "        image = self.load_image(path=image_path)\n",
    "        mask = self.load_mask(path=mask_path)\n",
    "        \n",
    "        data = {\n",
    "                    'image': torch.FloatTensor(image),\n",
    "                    'mask' : torch.LongTensor(mask)\n",
    "               }\n",
    "        \n",
    "        return data\n",
    "    \n",
    "    def load_image(self, path=None):\n",
    "        raw_image = Image.open(path)\n",
    "        raw_image = raw_image.convert('RGB')\n",
    "        raw_image = np.transpose(raw_image.resize((224, 224)), (2,1,0))\n",
    "        imx_t = np.array(raw_image, dtype=np.float32)/255.0\n",
    "\n",
    "        return imx_t\n",
    "\n",
    "    def load_mask(self, path=None):\n",
    "        raw_image = Image.open(path)\n",
    "        raw_image = raw_image.convert('RGB')\n",
    "        raw_image = raw_image.resize((224, 224))\n",
    "        imx_t = np.array(raw_image)\n",
    "        label_seg = np.zeros((2,224,224), dtype=np.int)\n",
    "        \n",
    "        for k in PALETTE:\n",
    "          label_seg[PALETTE[k]][(imx_t==k).all(axis=2)] = 1\n",
    "          \n",
    "        return label_seg\n",
    "      \n",
    "    def __len__(self):\n",
    "        return (len(os.listdir(self.image_root_dir)) // 2) - self.start_index - self.end_index\n",
    "    \n",
    "#if __name__ == \"__main__\":\n",
    "#    dataset_test = UTPDataset('dataset/dataset')\n",
    "#    data = dataset_test.__getitem__(4)\n",
    "#    print(data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training Script"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch.optim as optim\n",
    "import time\n",
    "import sys\n",
    "\n",
    "NUM_INPUT_CHANNELS = 3\n",
    "NUM_OUTPUT_CHANNELS = 2\n",
    "\n",
    "NUM_EPOCHS = 100\n",
    "LEARNING_RATE = 1e-5\n",
    "MOMENTUM = 0.8\n",
    "BATCH_SIZE = 8\n",
    "\n",
    "def train():\n",
    "  \n",
    "  print(\"Training:\")\n",
    "  prev_loss = float('inf')\n",
    "\n",
    "  for epoch in range(NUM_EPOCHS):\n",
    "    print(\"Epoch #{}\".format(epoch+1))\n",
    "    \n",
    "    for phase in ['train', 'val']:\n",
    "      if phase == 'train':\n",
    "        model.train()\n",
    "      else:\n",
    "        model.eval()\n",
    "      \n",
    "      t_start = time.time()\n",
    "      running_loss = 0.0\n",
    "\n",
    "      for idx, batch in enumerate(data_loaders[phase]):\n",
    "        input_tensor, target_tensor = batch['image'], batch['mask']\n",
    "        \n",
    "        if phase == 'val':\n",
    "            with torch.no_grad():\n",
    "                input_tensor = input_tensor.cuda()\n",
    "                target_tensor = target_tensor.cuda()\n",
    "\n",
    "                predicted_tensor, softmaxed_tensor = model(input_tensor)\n",
    "                target_tensor = torch.argmax(target_tensor, dim=1)\n",
    "\n",
    "                optimizer.zero_grad()\n",
    "                loss = criterion(softmaxed_tensor, target_tensor)\n",
    "\n",
    "        else:\n",
    "          input_tensor = input_tensor.cuda()\n",
    "          target_tensor = target_tensor.cuda()\n",
    "\n",
    "          predicted_tensor, softmaxed_tensor = model(input_tensor)       \n",
    "          target_tensor = torch.argmax(target_tensor, dim=1)\n",
    "          \n",
    "          optimizer.zero_grad()\n",
    "          loss = criterion(softmaxed_tensor, target_tensor)  \n",
    "          loss.backward()\n",
    "          optimizer.step()\n",
    "\n",
    "        running_loss += loss.float()\n",
    "\n",
    "        if idx % 10 == 0:\n",
    "          print(\"\\r\\tBatch progress: {:.2f}% [{}/{}]\".format((idx/len(data_loaders[phase]))*100, idx, len(data_loaders[phase])), end='')\n",
    "          sys.stdout.flush()\n",
    "    \n",
    "      delta = time.time() - t_start\n",
    "      print(\"\\t{} Loss: {:.8f}\\tTime: {:.8f}\".format(phase, running_loss, delta))\n",
    "      \n",
    "    if running_loss < prev_loss:\n",
    "        torch.save(model.state_dict(), os.path.join('.', 'epoch-{}.pth'.format(epoch+1)))\n",
    "        prev_loss = running_loss"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data and model initializations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/user/miniconda/envs/py36/lib/python3.6/site-packages/torch/cuda/__init__.py:117: UserWarning: \n",
      "    Found GPU2 Quadro 2000 which is of cuda capability 2.1.\n",
      "    PyTorch no longer supports this GPU because it is too old.\n",
      "    \n",
      "  warnings.warn(old_gpu_warn % (d, name, major, capability[1]))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training:\n",
      "Epoch #1\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 122.86852264\tTime: 70.15759134\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 29.64442635\tTime: 9.11308193\n",
      "Epoch #2\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 110.65926361\tTime: 74.57799459\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 27.04174423\tTime: 8.75054598\n",
      "Epoch #3\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 105.27581787\tTime: 74.83750105\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 26.33188629\tTime: 8.80310345\n",
      "Epoch #4\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 103.76653290\tTime: 75.02324247\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 26.08611107\tTime: 8.34460425\n",
      "Epoch #5\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 103.02478790\tTime: 74.89157891\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.94671631\tTime: 8.65996814\n",
      "Epoch #6\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 102.55515289\tTime: 74.74896097\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.97869492\tTime: 8.48062062\n",
      "Epoch #7\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 102.15342712\tTime: 74.95080113\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.83907700\tTime: 8.62104106\n",
      "Epoch #8\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 101.79238892\tTime: 74.94285345\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.72568321\tTime: 8.54877496\n",
      "Epoch #9\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 101.45703888\tTime: 74.94368577\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.66784668\tTime: 8.81207466\n",
      "Epoch #10\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 101.04473877\tTime: 74.85746956\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.62657356\tTime: 8.80781388\n",
      "Epoch #11\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 100.61896515\tTime: 75.02638602\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.59936714\tTime: 8.63105655\n",
      "Epoch #12\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 100.20145416\tTime: 74.87966466\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.46895409\tTime: 8.93152952\n",
      "Epoch #13\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 99.86335754\tTime: 75.21647620\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.52954102\tTime: 8.60657072\n",
      "Epoch #14\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 99.50078583\tTime: 74.99701929\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.45244980\tTime: 8.53091788\n",
      "Epoch #15\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 99.07091522\tTime: 74.84882760\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.52532387\tTime: 8.46690822\n",
      "Epoch #16\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 98.68675995\tTime: 75.04178810\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.47597122\tTime: 8.56723475\n",
      "Epoch #17\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 98.29865265\tTime: 74.88046741\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.38376427\tTime: 8.66352415\n",
      "Epoch #18\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 98.03034973\tTime: 74.97538853\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.61172867\tTime: 8.70047593\n",
      "Epoch #19\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 97.80165863\tTime: 74.80961537\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.36285782\tTime: 8.42213178\n",
      "Epoch #20\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 97.50052643\tTime: 74.99158955\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.47313690\tTime: 8.69329619\n",
      "Epoch #21\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 97.18723297\tTime: 74.92066908\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.35345840\tTime: 8.91002035\n",
      "Epoch #22\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 96.99977112\tTime: 75.07468128\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.37752342\tTime: 8.44697213\n",
      "Epoch #23\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 96.78012848\tTime: 74.96320391\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.37100410\tTime: 8.75618505\n",
      "Epoch #24\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 96.53233337\tTime: 75.04616761\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.28515434\tTime: 8.98310661\n",
      "Epoch #25\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 96.28167725\tTime: 75.03409314\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.30288315\tTime: 8.69751310\n",
      "Epoch #26\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 96.15204620\tTime: 74.95915031\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.44582176\tTime: 8.90349674\n",
      "Epoch #27\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 95.90850830\tTime: 75.07725024\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.28882599\tTime: 8.64256740\n",
      "Epoch #28\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 95.68109131\tTime: 75.08124948\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.32262039\tTime: 9.17336750\n",
      "Epoch #29\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 95.46609497\tTime: 75.06001616\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.31667709\tTime: 8.47263217\n",
      "Epoch #30\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 95.30626678\tTime: 75.08692288\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.27689171\tTime: 8.46496463\n",
      "Epoch #31\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 95.14891052\tTime: 74.91409636\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.23870087\tTime: 8.91815305\n",
      "Epoch #32\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 94.94899750\tTime: 75.11910868\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.39032173\tTime: 8.91267586\n",
      "Epoch #33\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 94.80417633\tTime: 75.07365155\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.28295326\tTime: 8.40678048\n",
      "Epoch #34\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 94.60597229\tTime: 75.04484916\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.19001770\tTime: 8.66696334\n",
      "Epoch #35\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 94.46442413\tTime: 74.96170497\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.30357170\tTime: 8.80276203\n",
      "Epoch #36\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 94.27162170\tTime: 74.94620442\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.22784424\tTime: 8.78959608\n",
      "Epoch #37\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 94.13472748\tTime: 74.97198653\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.29402161\tTime: 8.78479767\n",
      "Epoch #38\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 93.96743011\tTime: 74.94099092\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.25204849\tTime: 8.68992996\n",
      "Epoch #39\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 93.92673492\tTime: 75.02879119\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.27775764\tTime: 8.91946912\n",
      "Epoch #40\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 93.74156189\tTime: 75.02274227\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.20166779\tTime: 9.25506091\n",
      "Epoch #41\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 93.67245483\tTime: 74.95067024\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.21768951\tTime: 8.36166906\n",
      "Epoch #42\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 93.44451141\tTime: 75.03461194\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.23347282\tTime: 8.76227880\n",
      "Epoch #43\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 93.29083252\tTime: 75.10489798\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.24483490\tTime: 9.08918023\n",
      "Epoch #44\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 93.20104218\tTime: 75.04663420\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.22653961\tTime: 8.69341493\n",
      "Epoch #45\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 93.03091431\tTime: 75.03965044\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.24384880\tTime: 8.89833546\n",
      "Epoch #46\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 92.95458221\tTime: 75.21312165\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.18354607\tTime: 8.85687613\n",
      "Epoch #47\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 92.85872650\tTime: 75.05878496\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.17053604\tTime: 8.70204639\n",
      "Epoch #48\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 92.74820709\tTime: 75.27554011\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.18640518\tTime: 8.90832877\n",
      "Epoch #49\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 92.64572144\tTime: 75.06993222\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.20653152\tTime: 8.95797610\n",
      "Epoch #50\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 92.51390839\tTime: 75.12068582\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.15923691\tTime: 9.05918932\n",
      "Epoch #51\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 92.39601898\tTime: 75.07436228\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.20502281\tTime: 8.76398778\n",
      "Epoch #52\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 92.34125519\tTime: 74.88430333\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.18169022\tTime: 8.82854462\n",
      "Epoch #53\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 92.19025421\tTime: 74.98196363\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.42590714\tTime: 8.89983106\n",
      "Epoch #54\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 92.15801239\tTime: 75.14020038\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.16673470\tTime: 8.86770177\n",
      "Epoch #55\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 92.00972748\tTime: 75.05928493\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.20737267\tTime: 9.00426292\n",
      "Epoch #56\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.97502899\tTime: 75.00350165\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.21333885\tTime: 8.78531480\n",
      "Epoch #57\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.91748810\tTime: 74.89858842\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.25909805\tTime: 8.48666787\n",
      "Epoch #58\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.78836060\tTime: 75.16203332\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.20123672\tTime: 8.89362216\n",
      "Epoch #59\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.73310852\tTime: 74.95250773\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.12563133\tTime: 9.04248238\n",
      "Epoch #60\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.66881561\tTime: 75.10593748\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.23870659\tTime: 8.59581423\n",
      "Epoch #61\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.60132599\tTime: 75.01238680\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.20275307\tTime: 9.24475670\n",
      "Epoch #62\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.52295685\tTime: 75.01137471\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.29148483\tTime: 8.84270000\n",
      "Epoch #63\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.47375488\tTime: 74.91146708\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.13651466\tTime: 9.04545045\n",
      "Epoch #64\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.35118103\tTime: 75.18963265\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.15304184\tTime: 8.91311049\n",
      "Epoch #65\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.23444366\tTime: 75.14372468\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.22440720\tTime: 8.69534564\n",
      "Epoch #66\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.20445251\tTime: 74.98751640\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.16621399\tTime: 8.62685633\n",
      "Epoch #67\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.11061096\tTime: 75.08541965\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.15604973\tTime: 8.86729145\n",
      "Epoch #68\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 91.05806732\tTime: 75.07976961\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.22109985\tTime: 9.30330396\n",
      "Epoch #69\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.98606873\tTime: 75.25634289\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.20167542\tTime: 8.94924545\n",
      "Epoch #70\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.87774658\tTime: 74.94410348\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.21664429\tTime: 8.45850492\n",
      "Epoch #71\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.86239624\tTime: 75.15522242\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.14247322\tTime: 8.44854784\n",
      "Epoch #72\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.79315948\tTime: 75.04758668\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.14675713\tTime: 8.82440805\n",
      "Epoch #73\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.74771118\tTime: 75.22270656\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.20351601\tTime: 8.58706474\n",
      "Epoch #74\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.70184326\tTime: 74.95368695\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.15367889\tTime: 9.06387639\n",
      "Epoch #75\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.57585907\tTime: 74.91697025\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.10920334\tTime: 8.76792097\n",
      "Epoch #76\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.56600952\tTime: 75.06653857\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.15858269\tTime: 8.62729287\n",
      "Epoch #77\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.48593140\tTime: 75.18814135\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.17338562\tTime: 8.63412380\n",
      "Epoch #78\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.43587494\tTime: 75.32233453\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.17241287\tTime: 8.58185410\n",
      "Epoch #79\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.38916016\tTime: 74.93961406\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.23786926\tTime: 9.19123793\n",
      "Epoch #80\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.31055450\tTime: 75.08720589\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.13980675\tTime: 8.78649926\n",
      "Epoch #81\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.24742126\tTime: 74.96352053\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.18645668\tTime: 8.93721747\n",
      "Epoch #82\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.17340088\tTime: 75.26943517\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.17569542\tTime: 8.72739530\n",
      "Epoch #83\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.14915466\tTime: 75.05091357\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.22016335\tTime: 8.84048533\n",
      "Epoch #84\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.10826111\tTime: 75.29114819\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.13761902\tTime: 8.73129582\n",
      "Epoch #85\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.05150604\tTime: 74.98047566\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.18156815\tTime: 8.53211927\n",
      "Epoch #86\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 90.01157379\tTime: 75.13831496\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.16696739\tTime: 8.64112067\n",
      "Epoch #87\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.97369385\tTime: 74.92931700\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.14060402\tTime: 9.16441107\n",
      "Epoch #88\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.90209961\tTime: 75.17150927\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.18084335\tTime: 8.71037507\n",
      "Epoch #89\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.87383270\tTime: 75.14703155\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.15387726\tTime: 8.45418501\n",
      "Epoch #90\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.82932281\tTime: 74.95972776\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.10242462\tTime: 9.15575361\n",
      "Epoch #91\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.80775452\tTime: 74.97612214\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.12821960\tTime: 9.14173317\n",
      "Epoch #92\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.75653076\tTime: 75.03814936\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.15579033\tTime: 8.69017529\n",
      "Epoch #93\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.71375275\tTime: 75.08231091\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.20469856\tTime: 8.64661837\n",
      "Epoch #94\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.65663147\tTime: 75.07367778\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.08230019\tTime: 8.85029960\n",
      "Epoch #95\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.62481689\tTime: 75.10796905\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.14805603\tTime: 8.91997886\n",
      "Epoch #96\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.61910248\tTime: 75.17136049\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.16593933\tTime: 8.70549798\n",
      "Epoch #97\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.58168030\tTime: 75.10025620\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.17056274\tTime: 8.34303188\n",
      "Epoch #98\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.49401093\tTime: 75.30418658\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.21178436\tTime: 8.61811399\n",
      "Epoch #99\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.47731018\tTime: 75.17935133\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.08442307\tTime: 8.45308733\n",
      "Epoch #100\n",
      "\tBatch progress: 96.65% [260/269]\ttrain Loss: 89.45884705\tTime: 75.20530176\n",
      "\tBatch progress: 88.24% [60/68]\tval Loss: 25.10339737\tTime: 9.03545403\n"
     ]
    }
   ],
   "source": [
    "from torch.utils.data import DataLoader\n",
    "\n",
    "full_dataset = UTPDataset(img_dir='dataset/dataset')\n",
    "\n",
    "train_size = int(0.8 * len(full_dataset))\n",
    "val_size = len(full_dataset) - train_size\n",
    "\n",
    "train_dataset, val_dataset = torch.utils.data.random_split(full_dataset, [train_size, val_size])\n",
    "\n",
    "train_dataloader = DataLoader(train_dataset,\n",
    "                                  batch_size=BATCH_SIZE,\n",
    "                                  shuffle=True,\n",
    "                                  num_workers=6)\n",
    "\n",
    "val_dataloader = DataLoader(val_dataset,\n",
    "                                  batch_size=BATCH_SIZE,\n",
    "                                  shuffle=True,\n",
    "                                  num_workers=6)\n",
    "\n",
    "data_loaders = {\"train\": train_dataloader, \"val\": val_dataloader}\n",
    "\n",
    "if torch.cuda.is_available():\n",
    "  model = SegNet(input_channels=3, output_channels=2).cuda()\n",
    "  criterion = torch.nn.CrossEntropyLoss().cuda()\n",
    "else:\n",
    "  print('Error: Cuda was not available')\n",
    "  \n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=LEARNING_RATE, amsgrad=True)\n",
    "#optimizer = torch.optim.SGD(model.parameters(), lr=LEARNING_RATE, momentum=MOMENTUM)\n",
    "\n",
    "train()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predict random Image and get scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iou: 80.39857227840571\n",
      "precision: 89.47368421052632\n",
      "recall: 88.7976346911958\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x864 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model.eval()\n",
    "\n",
    "full_dataset = UTPDataset(img_dir='dataset/dataset')\n",
    "test_dataloader = DataLoader(full_dataset,\n",
    "                                  batch_size=BATCH_SIZE,\n",
    "                                  shuffle=True,\n",
    "                                  num_workers=6)\n",
    "\n",
    "dataiter = iter(test_dataloader)\n",
    "test = dataiter.next()\n",
    "\n",
    "image_id = 6\n",
    "\n",
    "image, mask = test['image'], test['mask']\n",
    "rgb = image[image_id]\n",
    "rgb.transpose_(0, 2)\n",
    "\n",
    "fig = plt.figure(figsize=(12,12))\n",
    "\n",
    "fig.add_subplot(1,4,1).title.set_text('RGB image')\n",
    "plt.imshow(rgb)\n",
    "\n",
    "image = image.cuda()\n",
    "\n",
    "output = model(image)\n",
    "\n",
    "prediction = output[1].cpu()\n",
    "\n",
    "prediction = prediction.data.numpy()[image_id][0]\n",
    "mask = mask.data.numpy()[image_id][0]\n",
    "\n",
    "fig.add_subplot(1,4,2).title.set_text('Softmaxed output')\n",
    "plt.imshow(prediction)\n",
    "\n",
    "prediction = (prediction > 0.8).astype(int)\n",
    "\n",
    "fig.add_subplot(1,4,3).title.set_text('Discretized prediction')\n",
    "plt.imshow(prediction)\n",
    "\n",
    "fig.add_subplot(1,4,4).title.set_text('Ground Truth')\n",
    "plt.imshow(mask)\n",
    "\n",
    "fig.tight_layout()\n",
    "\n",
    "#assert(False)\n",
    "prediction = 1 - prediction\n",
    "mask = 1 - mask\n",
    "\n",
    "intersection = np.logical_and(prediction, mask)\n",
    "union = np.logical_or(prediction, mask)\n",
    "iou_score = np.sum(intersection) / np.sum(union)\n",
    "\n",
    "print('iou: {}'.format(100*iou_score))\n",
    "\n",
    "precision = np.sum(intersection) / np.sum(prediction)\n",
    "recall = np.sum(intersection) / np.sum(mask)\n",
    "\n",
    "print('precision: {}'.format(100*precision))\n",
    "print('recall: {}'.format(100*recall))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evaluation loop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [],
   "source": [
    "#model = SegNet(input_channels=3, output_channels=2).cuda()\n",
    "model.load_state_dict(torch.load('test-small-50.pth'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7.25\n",
      "26.35\n",
      "9.2\n"
     ]
    }
   ],
   "source": [
    "from torch.utils.data import DataLoader\n",
    "\n",
    "def compute_iou(prediction, mask):\n",
    "    intersection = np.logical_and(prediction, mask)\n",
    "    union = np.logical_or(prediction, mask)\n",
    "    \n",
    "    iou = np.sum(intersection) / np.sum(union)\n",
    "    precision = (np.sum(intersection) / np.sum(prediction)) if np.sum(prediction) else 0\n",
    "    recall = np.sum(intersection) / np.sum(mask)\n",
    "    \n",
    "    return iou, precision, recall\n",
    "\n",
    "\n",
    "model.eval()\n",
    "\n",
    "test_dataset = UTPDataset(img_dir='dataset/dataset')\n",
    "\n",
    "test_dataloader = DataLoader(test_dataset,\n",
    "                                  batch_size=BATCH_SIZE,\n",
    "                                  shuffle=True,\n",
    "                                  num_workers=6)\n",
    "\n",
    "accumulated_iou = 0\n",
    "accumulated_precision = 0\n",
    "accumulated_recall = 0\n",
    "\n",
    "for batch_idx, batch in enumerate(test_dataloader):\n",
    "    image, mask = batch['image'].cuda(), batch['mask']\n",
    "\n",
    "    output = model(image)\n",
    "\n",
    "    prediction_batch = output[1].cpu()\n",
    "    \n",
    "    for img_idx, img in enumerate(prediction_batch):\n",
    "        current_prediction = prediction_batch.data.numpy()[img_idx][0]\n",
    "        current_prediction = 1 - (current_prediction > 0.8).astype(int)\n",
    "        current_mask = 1 - mask.data.numpy()[img_idx][0]\n",
    "        \n",
    "        score = compute_iou(current_prediction, current_mask)\n",
    "        \n",
    "        accumulated_iou += score[0]\n",
    "        accumulated_precision += score[1]\n",
    "        accumulated_recall += score[2] \n",
    "\n",
    "total_iou = 100 * (accumulated_iou / test_dataset.__len__())\n",
    "total_precission = 100 * (accumulated_precision / test_dataset.__len__())\n",
    "total_recall = 100 * (accumulated_recall / test_dataset.__len__())\n",
    "\n",
    "print(round(total_iou,2))\n",
    "print(round(total_precission,2))\n",
    "print(round(total_recall,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Inference Avg Time Script"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Avg execution time (s): 0.173\n"
     ]
    }
   ],
   "source": [
    "inputs = torch.randn(1, 3, 1024, 2048).cuda()\n",
    "\n",
    "model.eval()\n",
    "\n",
    "i = 0\n",
    "time_spent = []\n",
    "while i < 100:\n",
    "    start_time = time.time()\n",
    "    with torch.no_grad():\n",
    "        _ = model(inputs)\n",
    "\n",
    "    torch.cuda.synchronize()\n",
    "    if i != 0:\n",
    "        time_spent.append(time.time() - start_time)\n",
    "    i += 1\n",
    "print('Avg execution time (s): {:.3f}'.format(np.mean(time_spent)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predict single image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "ename": "SyntaxError",
     "evalue": "invalid syntax (<ipython-input-1-dda5dce7b063>, line 14)",
     "output_type": "error",
     "traceback": [
      "\u001b[0;36m  File \u001b[0;32m\"<ipython-input-1-dda5dce7b063>\"\u001b[0;36m, line \u001b[0;32m14\u001b[0m\n\u001b[0;31m    img = img.view(1, 3, 224, 224)doc\u001b[0m\n\u001b[0m                                    ^\u001b[0m\n\u001b[0;31mSyntaxError\u001b[0m\u001b[0;31m:\u001b[0m invalid syntax\n"
     ]
    }
   ],
   "source": [
    "image_path = 'dataset/dataset/10181_full.png'\n",
    "raw_image = Image.open('dataset/dataset/10339_full.png')\n",
    "raw_image = raw_image.convert('RGB')\n",
    "\n",
    "print(raw_image)\n",
    "\n",
    "fig = plt.figure()\n",
    "fig.add_subplot(1,2,1)\n",
    "plt.imshow(raw_image.resize((224, 224)))\n",
    "raw_image = np.transpose(raw_image.resize((224, 224)), (2,1,0))\n",
    "imx_t = np.array(raw_image, dtype=np.float32)/255.0\n",
    "\n",
    "img = torch.FloatTensor(imx_t).cuda()\n",
    "img = img.view(1, 3, 224, 224)doc\n",
    "output = model(img)\n",
    "\n",
    "foo = output[1]\n",
    "foo = foo.cpu()\n",
    "fig.add_subplot(1,2,2)\n",
    "plt.imshow(foo.data.numpy()[0][0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'Image' 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<ipython-input-2-4eb99a9fb211>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;31m#rgb = Image.open('GeneratedSequences/scene1/mask/MainCamera/000005.png')\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 3\u001b[0;31m \u001b[0mraw_image\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mImage\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mopen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'dataset/dataset/10286_segmentation_full.png'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      4\u001b[0m \u001b[0mraw_image\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mraw_image\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mresize\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m224\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m224\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mimx_t\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mraw_image\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'Image' is not defined"
     ]
    }
   ],
   "source": [
    "#rgb = Image.open('GeneratedSequences/scene1/mask/MainCamera/000005.png')\n",
    "\n",
    "raw_image = Image.open('dataset/dataset/10286_segmentation_full.png')\n",
    "raw_image = raw_image.resize((224, 224))\n",
    "imx_t = np.array(raw_image)\n",
    "imx_t = imx_t[:,:,:3]\n",
    "label_seg = np.zeros((2,224,224), dtype=np.int)\n",
    "\n",
    "label_seg[0][(imx_t==[255,0,0]).all(axis=2)] = 1\n",
    "label_seg[1][np.where(label_seg[0] == 0)] = 1\n",
    "\n",
    "fig = plt.figure()\n",
    "fig.add_subplot(1,2,1)\n",
    "plt.imshow(label_seg[0])\n",
    "fig.add_subplot(1,2,2)\n",
    "plt.imshow(label_seg[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# UnrealROX to UTP Script"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "directory = 'scene2/'\n",
    "dataset_path = 'dataset/dataset/'\n",
    "\n",
    "new_id = len(os.listdir(dataset_path)) // 2\n",
    "\n",
    "for camera in ['CornerCamera0/', 'CornerCamera1/', 'MainCamera/']:\n",
    "    for idx in range(1,len(os.listdir(os.path.join(directory, 'rgb', camera)))):\n",
    "        if idx % 10 == 0:\n",
    "            filename = str(idx).zfill(6) + '.png'\n",
    "            \n",
    "            raw_image = Image.open(directory + 'mask/' + camera + filename)\n",
    "            raw_image = raw_image.convert('RGB')\n",
    "\n",
    "            has_human = False\n",
    "            pixels = raw_image.load()\n",
    "            for i in range(raw_image.size[0]):\n",
    "                for j in range(raw_image.size[1]):\n",
    "                        if pixels[i,j] == (255, 0, 0):\n",
    "                            raw_image.putpixel((i,j), (255, 255, 255))\n",
    "                            has_human = True\n",
    "                        else:\n",
    "                            raw_image.putpixel((i,j), (0, 0, 0))\n",
    "            \n",
    "            if not has_human:\n",
    "                print('no_human')\n",
    "                continue\n",
    "                \n",
    "            os.rename(directory + 'rgb/' + camera + filename, dataset_path + str(new_id).zfill(5) + '_full.png')\n",
    "            #Image.open(directory + 'rgb/' + camera + filename + '.jpg').save(dataset_path + str(new_id).zfill(5) + '_full.png')\n",
    "            \n",
    "            raw_image.save(dataset_path + str(new_id).zfill(5) + '_segmentation_full.png')\n",
    "            print(dataset_path + str(new_id).zfill(5) + '_segmentation_full.png')\n",
    "            new_id += 1"
   ]
  }
 ],
 "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
}