ResNeXt-101_32x16d_submission_latest_copy.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torchvision\n",
    "import torchvision.transforms as transforms\n",
    "import torch.optim as optim\n",
    "import torch.nn as nn\n",
    "from torchvision import models\n",
    "from tqdm import tqdm\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import os\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import itertools\n",
    "import numpy as np\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"]=\"4,6,7\"\n",
    "\n",
    "torch.manual_seed(0)\n",
    "np.random.seed(0)\n",
    "\n",
    "torch.backends.cudnn.deterministic = True\n",
    "torch.backends.cudnn.benchmark = False\n",
    "\n",
    "# font = {'weight' : 'bold',\n",
    "#         'size'   : 90}\n",
    "\n",
    "# plt.rc('font', **font)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "class ImageFolderWithPaths(torchvision.datasets.ImageFolder):\n",
    "    \"\"\"Custom dataset that includes image file paths. Extends\n",
    "    torchvision.datasets.ImageFolder\n",
    "    \"\"\"\n",
    "\n",
    "    # override the __getitem__ method. this is the method that dataloader calls\n",
    "    def __getitem__(self, index):\n",
    "        # this is what ImageFolder normally returns \n",
    "        original_tuple = super(ImageFolderWithPaths, self).__getitem__(index)\n",
    "        # the image file path\n",
    "        path = self.imgs[index][0]\n",
    "        # make a new tuple that includes original and the path\n",
    "        tuple_with_path = (original_tuple + (path,))\n",
    "        return tuple_with_path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_size = 16\n",
    "#IMG_SIZE = [224,224]\n",
    "IMG_SIZE = [320,320]  ## Best so far with this resolution and 32x48d\n",
    "#IMG_SIZE = [384,384]\n",
    "num_classes = 42\n",
    "device = \"cuda\"\n",
    "target_names = [str(i) for i in range(42)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "transform = transforms.Compose(\n",
    "    [transforms.Resize(IMG_SIZE),\n",
    "     transforms.ToTensor(),\n",
    "     transforms.Normalize(mean = [0.485, 0.456, 0.406], std = [0.229, 0.224, 0.225])])\n",
    "\n",
    "#trainvalset = torchvision.datasets.ImageFolder(\"train/train/\",transform=transform)\n",
    "\n",
    "trainvalset = ImageFolderWithPaths(\"train/train/\",transform=transform)\n",
    "#print(trainvalset_[1])\n",
    "\n",
    "trainset, valset = torch.utils.data.random_split(trainvalset, [len(trainvalset)-30000, 30000])\n",
    "\n",
    "trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size,\n",
    "                                          shuffle=True, num_workers=1)\n",
    "\n",
    "valloader = torch.utils.data.DataLoader(valset, batch_size=batch_size,\n",
    "                                          shuffle=True, num_workers=1)\n",
    "\n",
    "testset = ImageFolderWithPaths(\"test/\",transform=transform)\n",
    "\n",
    "testloader = torch.utils.data.DataLoader(testset, batch_size=batch_size,\n",
    "                                         shuffle=False, num_workers=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([14, 40,  7, 21, 25, 11, 41, 21,  8,  8, 32, 15, 34,  8,  9, 22])\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def imshow(img):\n",
    "    img = img / 2 + 0.5     # unnormalize\n",
    "    npimg = img.numpy()\n",
    "    plt.imshow(np.transpose(npimg, (1, 2, 0)))\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "# get some random training images\n",
    "dataiter = iter(trainloader)\n",
    "images, labels,path = dataiter.next()\n",
    "print(labels)\n",
    "\n",
    "# show images\n",
    "imshow(torchvision.utils.make_grid(images))\n",
    "# print labels\n",
    "#print(' '.join('%5s' % classes[labels[j]] for j in range(4)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using cache found in /root/.cache/torch/hub/facebookresearch_WSL-Images_master\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ResNet(\n",
      "  (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
      "  (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "  (relu): ReLU(inplace=True)\n",
      "  (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "  (layer1): Sequential(\n",
      "    (0): Bottleneck(\n",
      "      (conv1): Conv2d(64, 1536, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(1536, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(1536, 1536, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(1536, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(1536, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (downsample): Sequential(\n",
      "        (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (1): Bottleneck(\n",
      "      (conv1): Conv2d(256, 1536, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(1536, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(1536, 1536, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(1536, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(1536, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (2): Bottleneck(\n",
      "      (conv1): Conv2d(256, 1536, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(1536, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(1536, 1536, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(1536, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(1536, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (layer2): Sequential(\n",
      "    (0): Bottleneck(\n",
      "      (conv1): Conv2d(256, 3072, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(3072, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(3072, 3072, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(3072, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(3072, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (downsample): Sequential(\n",
      "        (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "        (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (1): Bottleneck(\n",
      "      (conv1): Conv2d(512, 3072, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(3072, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(3072, 3072, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(3072, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(3072, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (2): Bottleneck(\n",
      "      (conv1): Conv2d(512, 3072, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(3072, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(3072, 3072, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(3072, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(3072, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (3): Bottleneck(\n",
      "      (conv1): Conv2d(512, 3072, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(3072, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(3072, 3072, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(3072, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(3072, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (layer3): Sequential(\n",
      "    (0): Bottleneck(\n",
      "      (conv1): Conv2d(512, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (downsample): Sequential(\n",
      "        (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "        (1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (1): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (2): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (3): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (4): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (5): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (6): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (7): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (8): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (9): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (10): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (11): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (12): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (13): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (14): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (15): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (16): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (17): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (18): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (19): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (20): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (21): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (22): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 6144, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(6144, 6144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(6144, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(6144, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (layer4): Sequential(\n",
      "    (0): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 12288, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(12288, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(12288, 12288, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(12288, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(12288, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (downsample): Sequential(\n",
      "        (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "        (1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (1): Bottleneck(\n",
      "      (conv1): Conv2d(2048, 12288, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(12288, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(12288, 12288, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(12288, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(12288, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "    (2): Bottleneck(\n",
      "      (conv1): Conv2d(2048, 12288, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn1): BatchNorm2d(12288, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv2): Conv2d(12288, 12288, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)\n",
      "      (bn2): BatchNorm2d(12288, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(12288, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))\n",
      "  (fc): Linear(in_features=2048, out_features=42, bias=True)\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "\n",
    "#model = models.resnet18(pretrained=True)\n",
    "# model = torchvision.models.wide_resnet101_2(pretrained=True)\n",
    "#model = torch.hub.load('facebookresearch/WSL-Images', 'resnext101_32x32d_wsl')\n",
    "model = torch.hub.load('facebookresearch/WSL-Images', 'resnext101_32x48d_wsl')\n",
    "#print(model)\n",
    "num_ftrs = model.fc.in_features\n",
    "model.fc = nn.Linear(num_ftrs, num_classes)\n",
    "#model = nn.DataParallel(model)\n",
    "model.load_state_dict(torch.load(\"ResNeXt-101_32x48d_320x320.pth\"))\n",
    "model.to(device)\n",
    "\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = optim.SGD(model.parameters(), lr=0.00001, momentum=0.9)\n",
    "print(model)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_confusion_matrix(cm,\n",
    "                          target_names,\n",
    "                          title='Confusion matrix',\n",
    "                          cmap=None,\n",
    "                          normalize=True):\n",
    "    \"\"\"\n",
    "    given a sklearn confusion matrix (cm), make a nice plot\n",
    "\n",
    "    Arguments\n",
    "    ---------\n",
    "    cm:           confusion matrix from sklearn.metrics.confusion_matrix\n",
    "\n",
    "    target_names: given classification classes such as [0, 1, 2]\n",
    "                  the class names, for example: ['high', 'medium', 'low']\n",
    "\n",
    "    title:        the text to display at the top of the matrix\n",
    "\n",
    "    cmap:         the gradient of the values displayed from matplotlib.pyplot.cm\n",
    "                  see http://matplotlib.org/examples/color/colormaps_reference.html\n",
    "                  plt.get_cmap('jet') or plt.cm.Blues\n",
    "\n",
    "    normalize:    If False, plot the raw numbers\n",
    "                  If True, plot the proportions\n",
    "\n",
    "    Usage\n",
    "    -----\n",
    "    plot_confusion_matrix(cm           = cm,                  # confusion matrix created by\n",
    "                                                              # sklearn.metrics.confusion_matrix\n",
    "                          normalize    = True,                # show proportions\n",
    "                          target_names = y_labels_vals,       # list of names of the classes\n",
    "                          title        = best_estimator_name) # title of graph\n",
    "\n",
    "    Citiation\n",
    "    ---------\n",
    "    http://scikit-learn.org/stable/auto_examples/model_selection/plot_confusion_matrix.html\n",
    "\n",
    "    \"\"\"\n",
    "\n",
    "    accuracy = np.trace(cm) / float(np.sum(cm))\n",
    "    misclass = 1 - accuracy\n",
    "\n",
    "    if cmap is None:\n",
    "        cmap = plt.get_cmap('Blues')\n",
    "\n",
    "    plt.figure(figsize=(80, 60))\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    #plt.rc('xtick', labelsize=20) \n",
    "    #plt.rc('ytick', labelsize=20) \n",
    "\n",
    "    if target_names is not None:\n",
    "        tick_marks = np.arange(len(target_names))\n",
    "        plt.xticks(tick_marks, target_names, rotation=45,fontsize=50)\n",
    "        plt.yticks(tick_marks, target_names,fontsize=50)\n",
    "\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "\n",
    "\n",
    "    thresh = cm.max() / 1.5 if normalize else cm.max() / 2\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        if normalize:\n",
    "            plt.text(j, i, \"{:0.4f}\".format(cm[i, j]),\n",
    "                     horizontalalignment=\"center\",\n",
    "                     color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "        else:\n",
    "            plt.text(j, i, \"{:,}\".format(cm[i, j]),\n",
    "                     horizontalalignment=\"center\",\n",
    "                     color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label\\naccuracy={:0.4f}; misclass={:0.4f}'.format(accuracy, misclass))\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluate(net):\n",
    "    net.eval()\n",
    "    correct = 0\n",
    "    total = 0\n",
    "    y_true = []\n",
    "    y_pred = []\n",
    "    with torch.no_grad():\n",
    "        pbar = tqdm(valloader)\n",
    "        i_batch = 0\n",
    "        for data in pbar:\n",
    "            images, labels,path = data\n",
    "            images = images.to(device)\n",
    "            labels = labels.to(device)\n",
    "            outputs = net(images)\n",
    "            _, predicted = torch.max(outputs.data, 1)\n",
    "            \n",
    "            for output,label in zip(predicted,labels):\n",
    "                y_pred.append(output.item())\n",
    "                y_true.append(label.item())\n",
    "            total += labels.size(0)\n",
    "            correct += (predicted == labels).sum().item()\n",
    "            pbar.set_description(\"Val Accuracy %.4f\" % (correct/float(total)))\n",
    "            i_batch+=1\n",
    "#             if(i_batch==10):\n",
    "#                 break\n",
    "\n",
    "    print('Accuracy of the network on the 10000 test images: %d %%' % (\n",
    "        100 * correct / total))\n",
    "    \n",
    "    cm = confusion_matrix(y_true, y_pred)\n",
    "    \n",
    "    plot_confusion_matrix(cm= cm, \n",
    "                      normalize    = True,\n",
    "                      target_names = target_names,\n",
    "                      title        = \"Confusion Matrix\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def submit(net):\n",
    "    net.eval()\n",
    "    correct = 0\n",
    "    total = 0\n",
    "    y_true = []\n",
    "    y_pred = []\n",
    "    file_ = open(\"out.csv\",\"w\")\n",
    "    file_.write(\"filename, category\\n\")\n",
    "    with torch.no_grad():\n",
    "        pbar = tqdm(testloader)\n",
    "        i_batch = 0\n",
    "        for data in pbar:\n",
    "            images, labels,paths = data\n",
    "            images = images.to(device)\n",
    "            labels = labels.to(device)\n",
    "            outputs = net(images)\n",
    "            _, predicted = torch.max(outputs.data, 1)\n",
    "            #print(path[0],path[0][10:])\n",
    "            for path,pred in zip(paths,predicted):\n",
    "                file_.write(path[10:]+\", \"+\"{:02d}\".format(pred.item())+\"\\n\")\n",
    "            #file_.write()\n",
    "            i_batch+=1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 762/762 [10:47<00:00,  1.18it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Finished Training\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# print(\"Started Training\")\n",
    "# for epoch in range(2):  # loop over the dataset multiple times\n",
    "#     running_loss = 0.0\n",
    "#     for i, data in enumerate(trainloader, 0):\n",
    "#         # get the inputs; data is a list of [inputs, labels]\n",
    "#         inputs, labels = data\n",
    "#         inputs = inputs.to(device)\n",
    "#         labels = labels.to(device)\n",
    "\n",
    "#         # zero the parameter gradients\n",
    "#         optimizer.zero_grad()\n",
    "\n",
    "#         # forward + backward + optimize\n",
    "#         outputs = model(inputs)\n",
    "#         loss = criterion(outputs, labels)\n",
    "#         loss.backward()\n",
    "#         optimizer.step()\n",
    "\n",
    "#         # print statistics\n",
    "#         running_loss += loss.item()\n",
    "#         if i % 20 == 19:    # print every 2000 mini-batches\n",
    "#             print('[%d, %5d] loss: %.3f' %\n",
    "#                   (epoch + 1, i + 1, running_loss / 20))\n",
    "#             running_loss = 0.0\n",
    "#         #if i % 200 == 199:\n",
    "#     evaluate(model)\n",
    "#     model.train()\n",
    "#     torch.save(model.state_dict(), \"ResNeXt-101_32x48d.pth\")\n",
    "submit(model)\n",
    "\n",
    "print('Finished Training')"
   ]
  }
 ],
 "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": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}