Vicious_Visionaries.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Vicious_Visionaries.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/tessytom/content-creation/blob/main/Vicious_Visionaries.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {},
        "id": "_qv6qtU2wIxx"
      },
      "source": [
        "# @title Install dependencies\n",
        "!pip install torch_intermediate_layer_getter --quiet\n",
        "!pip install Pillow --quiet\n",
        "!pip install pytorch_lightning --quiet"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HHnLPR4GXMpj",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "cellView": "form",
        "outputId": "3e6cb52e-d469-4c12-9636-fe911e96ef1f"
      },
      "source": [
        "#@title Connect Drive\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DXSxr8iAwUEi",
        "cellView": "form"
      },
      "source": [
        "#@title load CIFAR-10-C Data\n",
        "# !wget 'https://zenodo.org/record/2535967/files/CIFAR-10-C.tar'\n",
        "# !tar -xvf \"CIFAR-10-C.tar\" -C '/content/drive/MyDrive/NMA_DL_project/CIFAR-10-C'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {},
        "id": "xanbaEUhwIxy",
        "cellView": "form"
      },
      "source": [
        "#@title Imports\n",
        "import random\n",
        "import time\n",
        "\n",
        "import numpy as np\n",
        "import tqdm.notebook as tqdm\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "from PIL import Image, ImageFilter\n",
        "from shutil import copyfile, rmtree\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "from torch import optim\n",
        "import torch.nn.functional as F\n",
        "# For getting output of intermediate layers in Pytorch\n",
        "from torch_intermediate_layer_getter import IntermediateLayerGetter as LayerGetter\n",
        "# For interactive visualization\n",
        "from torch.utils.tensorboard import SummaryWriter\n",
        "\n",
        "import torchvision\n",
        "from torchvision import transforms\n",
        "import torchvision.models as models\n",
        "\n",
        "import pytorch_lightning\n",
        "\n",
        "from sklearn.model_selection import KFold\n",
        "\n",
        "import copy \n",
        "import pickle as pkl \n",
        "\n",
        "from functools import partial\n",
        "import collections\n",
        "\n",
        "import numpy as np\n",
        "from scipy.stats import spearmanr, pearsonr, rankdata\n",
        "from functools import partial\n",
        "from scipy.spatial.distance import squareform\n",
        "\n",
        "%load_ext tensorboard"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {},
        "id": "3WbKWLkKwIxy"
      },
      "source": [
        "#@title Helper functions\n",
        "def set_seed(seed=None, seed_torch=True):\n",
        "  if seed is None:\n",
        "    seed = np.random.choice(2 ** 32)\n",
        "  random.seed(seed)\n",
        "  np.random.seed(seed)\n",
        "  if seed_torch:\n",
        "    torch.manual_seed(seed)\n",
        "    torch.cuda.manual_seed_all(seed)\n",
        "    torch.cuda.manual_seed(seed)\n",
        "    torch.backends.cudnn.benchmark = False\n",
        "    torch.backends.cudnn.deterministic = True\n",
        "\n",
        "  print(f'Random seed {seed} has been set.')\n",
        "\n",
        "# In case that `DataLoader` is used\n",
        "def seed_worker(worker_id):\n",
        "  worker_seed = torch.initial_seed() % 2**32\n",
        "  np.random.seed(worker_seed)\n",
        "  random.seed(worker_seed)\n",
        "\n",
        "def set_device():\n",
        "  device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
        "  if device != \"cuda\":\n",
        "    print(\"WARNING: For this notebook to perform best, \"\n",
        "        \"if possible, in the menu under `Runtime` -> \"\n",
        "        \"`Change runtime type.`  select `GPU` \")\n",
        "  else:\n",
        "    print(\"GPU is enabled in this notebook.\")\n",
        "\n",
        "  return device\n",
        "\n",
        "# function to compute accuracy (could be probably implemented in a better way)\n",
        "def accuracy(outputs, labels):\n",
        "  correct_pred=0\n",
        "  total_pred=0\n",
        "  predictions = torch.argmax(outputs,dim=1)\n",
        "  for label, prediction in zip(labels, predictions):\n",
        "    if label == prediction:\n",
        "      correct_pred += 1\n",
        "    total_pred += 1\n",
        "  accuracy = correct_pred/total_pred\n",
        "  return accuracy\n",
        "\n",
        "# function to plot the trianing loss and accuracy and the test accuracy\n",
        "def plot_train_loss_acc(losses, accuracies_train, accuracies_test, fext):\n",
        "    plt.figure()\n",
        "    fig, axs = plt.subplots(1,2)\n",
        "    axs[0].plot( losses, label='Training Loss' )\n",
        "    axs[0].set_xlabel('epochs') ; axs[0].set_ylabel('Loss')\n",
        "    axs[0].legend()\n",
        "    axs[1].plot( accuracies_train, label='Training Accuracy' )\n",
        "    axs[1].plot( accuracies_test, label='Test Accuracy' )\n",
        "    axs[1].set_xlabel('epochs') ; axs[1].set_ylabel('Accuracy')\n",
        "    axs[1].set_ylim([0.0,1.05])\n",
        "    axs[1].legend()\n",
        "    fig.tight_layout(pad=3.0)\n",
        "    plt.savefig(fext, bbox_inches='tight')\n",
        "\n",
        "# function to show image \n",
        "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",
        "# function to show some of the images \n",
        "def show_sample_images(dataloader): \n",
        "  # get some random training images\n",
        "  dataiter = iter(dataloader)\n",
        "  images, labels = dataiter.next()\n",
        "  print(images.size())\n",
        "  # show images\n",
        "  imshow(torchvision.utils.make_grid(images))\n",
        "  # setup labels \n",
        "  classes = ('plane', 'car', 'bird', 'cat',\n",
        "           'deer', 'dog', 'frog', 'horse', 'ship', 'truck')\n",
        "  # print labels\n",
        "  print(' '.join('%5s' % classes[labels[j]] for j in range(batch_size)))\n",
        "\n",
        "def test_function(model,fig_savepath,res_savepath, **kwargs): \n",
        "\n",
        "  # set model in eval mode \n",
        "  model.eval()\n",
        "  # initialize accuracy dict \n",
        "  all_accuracies = dict()\n",
        "  \n",
        "  #loop through test sets\n",
        "  for test_name, test_set in kwargs.items(): \n",
        "\n",
        "    # first create data loader \n",
        "    test_loader = torch.utils.data.DataLoader(test_set, batch_size=100,\n",
        "                                         shuffle=False, num_workers=2)\n",
        "\n",
        "    # start testing here\n",
        "    test_acc = []#0.0\n",
        "    for i, data in enumerate(test_loader):\n",
        "      # get the inputs; data is a list of [inputs, labels]\n",
        "      inputs, labels = data[0].to(DEVICE), data[1].to(DEVICE)\n",
        "\n",
        "      with torch.no_grad():\n",
        "        outputs = model(inputs)\n",
        "        test_acc.append(accuracy(outputs, labels))\n",
        "\n",
        "    all_accuracies[test_name] = np.array(test_acc).mean()\n",
        "\n",
        "  #basic bar plot \n",
        "  plt.bar(all_accuracies.keys(), all_accuracies.values())\n",
        "  plt.tick_params(axis='x', which='major', labelsize=10)\n",
        "  plt.xticks(rotation=90)\n",
        "  plt.savefig(fig_savepath, bbox_inches='tight')\n",
        "\n",
        "  #save accuracies \n",
        "  with open(res_savepath+'.pickle', 'wb') as handle:\n",
        "    pkl.dump(all_accuracies, handle, protocol=pkl.HIGHEST_PROTOCOL)\n",
        "\n",
        "  return all_accuracies \n",
        "\n",
        "def get_activation(name):\n",
        "  def hook(model, input, output):\n",
        "      this_activation = output.detach()\n",
        "      this_activation = this_activation.flatten().cpu().numpy()\n",
        "      activations[name] = this_activation\n",
        "  return hook\n",
        "\n",
        "def make_RDM(activity_pattern_matrix, distance='pearson'):\n",
        "    #TODO: implement for multiple matrices.\n",
        "    #TODO: implement other distance norms \n",
        "    '''Compute dissimilarity matrix.\n",
        "    \n",
        "    Parameters\n",
        "    ----------\n",
        "    activity_pattern_matrix : ndarray\n",
        "        Matrix which holds activity patterns for several conditions. Each\n",
        "        column is one condition, each row one measurement channel.\n",
        "    distance : str\n",
        "        The desired distance measure.\n",
        "        \n",
        "    Returns\n",
        "    -------\n",
        "    rdm : ndarray\n",
        "        Dissimilarity matrix indicating dissimilarity for all conditions in\n",
        "        `activity_pattern_matrix`.\n",
        "    '''\n",
        "    if distance=='pearson':\n",
        "        rdm = 1 - np.corrcoef(activity_pattern_matrix, rowvar=False)\n",
        "    # elif distance=='Euclidian':\n",
        "    #     rdm = 1 - np.corrcoef(activity_pattern_matrix, rowvar=False)\n",
        "    return rdm\n",
        "\n",
        "def flatten_RDM(rdm: np.ndarray) -> np.ndarray:\n",
        "    '''Flatten the upper half of a dissimilarity matrix to a vector.\n",
        "    \n",
        "     Multiple dissimilarity matrices can be fed at once, requiring the shape \n",
        "     (n,n,m), where m denotes different matrices of shape (n,n).\n",
        "    \n",
        "    Parameters\n",
        "    ----------\n",
        "    rdm : ndarray\n",
        "        The dissimilarity matrix which shall be flattened.\n",
        "    \n",
        "    Returns\n",
        "    -------\n",
        "    rdv : ndarray\n",
        "        The unique upper half of `rdm`.\n",
        "    \n",
        "    '''\n",
        "    if rdm.ndim==3:\n",
        "        mapfunc = partial(squareform, checks=False)\n",
        "        rdv = np.array(list(map(mapfunc, np.moveaxis(rdm, -1, 0)))).T\n",
        "    elif rdm.ndim==2:\n",
        "        rdv = rdm[np.triu_indices(rdm.shape[0], k=1)].reshape(-1,1)\n",
        "    return rdv\n",
        "\n",
        "def correlate_RDMs(rdv1, rdv2, score_type='pearson'):\n",
        "    '''Relate two flattened dissimilarity matrices to each other\n",
        "    \n",
        "    Parameters\n",
        "    ----------\n",
        "    rdv1 : array_like\n",
        "        First flattened dissimilarity matrix. \n",
        "    rdv2 : array_like\n",
        "        Second flattened dissimilarity matrix.\n",
        "    score_type : {'pearson', 'spearman'}, optional\n",
        "        Type of association measure to compute (defaults to `pearson`).\n",
        "        \n",
        "    Return\n",
        "    ------\n",
        "    corr : float\n",
        "        Correlation coefficient.\n",
        "    p_value : float\n",
        "        Two-tailed p-value.\n",
        "    \n",
        "    '''\n",
        "    if score_type == 'pearson':\n",
        "        corr, p_value = pearsonr(rdv1, rdv2)\n",
        "    elif score_type == 'spearman':\n",
        "        corr, p_value = spearmanr(rdv1, rdv2)\n",
        "    return corr, p_value\n",
        "\n",
        "def get_activations_and_RDMs(model,  plotting_on=False,**kwargs,):\n",
        "\n",
        "  # set model in eval mode \n",
        "  model.eval()\n",
        "  # initialize accuracy dict \n",
        "  all_RDMs = dict()\n",
        "  all_activations = dict()\n",
        "\n",
        "  # register forward hooks \n",
        "  activations = collections.defaultdict(list)\n",
        "\n",
        "  # define function for saving activations \n",
        "  def save_activation(name, model, inp, out):\n",
        "\t  activations[name].append(out.to('cpu'))\n",
        "  \n",
        "  #check if model is VGG or simple model \n",
        "  if hasattr(model, 'features'): \n",
        "    pool_ct = 1\n",
        "    fc_ct = 1\n",
        "    for idx,layer in enumerate(model.features):\n",
        "        if isinstance(layer,nn.MaxPool2d):\n",
        "          layer.register_forward_hook(partial(save_activation,'Pool_'+ str(pool_ct))) \n",
        "          pool_ct +=1  \n",
        "    for idx,layer in enumerate(model.classifier):\n",
        "        if isinstance(layer, nn.ReLU):\n",
        "          layer.register_forward_hook(partial(save_activation,'Fc_'+ str(fc_ct))) \n",
        "          fc_ct +=1\n",
        "  else: \n",
        "    for name, m in model.named_children():\n",
        "      if type(m)!=nn.Dropout:\n",
        "        # partial to assign the layer name to each hook\n",
        "        m.register_forward_hook(partial(save_activation, name))\n",
        "    \n",
        "  # loop through all test loaders that are given as keyword arguments \n",
        "  for test_name, test_set in kwargs.items(): \n",
        "\n",
        "    # first create data loader \n",
        "    test_loader = torch.utils.data.DataLoader(test_set, batch_size=100, shuffle=False, num_workers=2)\n",
        "    # a dictionary that keeps saving the activations as they come\n",
        "    activations = collections.defaultdict(list)\n",
        "    all_labels = [] \n",
        "    for i, data in enumerate(test_loader):\n",
        "    # get the inputs; data is a list of [inputs, labels]\n",
        "      inp, labels = data[0].to(DEVICE), data[1].to(DEVICE)\n",
        "\n",
        "      with torch.no_grad():\n",
        "        out = model(inp)\n",
        "        all_labels.append(labels.to('cpu'))\n",
        "\n",
        "    # concatenate all the outputs we saved to get the the activations for each layer for the whole dataset\n",
        "    activations = {name: torch.cat(outputs, 0) for name, outputs in activations.items()}\n",
        "    all_labels = torch.cat(all_labels, 0)\n",
        "\n",
        "    # now rearrange and average the activations \n",
        "    category_activations = dict()\n",
        "\n",
        "    for layer, activation in activations.items(): \n",
        "\n",
        "      activation = torch.flatten(activation, start_dim=1, end_dim=-1)\n",
        "\n",
        "      layer_size = activation.size()[1]\n",
        "      act_mat = np.zeros([10,layer_size])\n",
        "\n",
        "      for cat in np.unique(all_labels): \n",
        "\n",
        "        this_activation = activation[all_labels == cat] \n",
        "        this_activation = torch.mean(this_activation,axis=0)\n",
        "        act_mat[cat,:] = this_activation\n",
        "\n",
        "      category_activations[layer] = act_mat\n",
        "\n",
        "    all_activations[test_name] = category_activations\n",
        "    #now compute RDMs \n",
        "    RDMs = dict()\n",
        "\n",
        "    for layer, cat_activation in category_activations.items(): \n",
        "\n",
        "      this_RDM = make_RDM(cat_activation.transpose(), distance='pearson')\n",
        "\n",
        "      if plotting_on:\n",
        "        plt.figure()\n",
        "        plt.imshow(this_RDM)\n",
        "\n",
        "      RDMs[layer] = this_RDM \n",
        "    \n",
        "    all_RDMs[test_name] = RDMs\n",
        "\n",
        "  return all_activations, all_RDMs \n",
        "\n",
        "# define class for corrupted data \n",
        "class CorruptedDataset(torch.utils.data.Dataset):\n",
        "    def __init__(self, path,file_name, severity_level, transform = None):\n",
        "        self.imgs_path = path \n",
        "        file_name = self.imgs_path+file_name\n",
        "        data = np.load(file_name)\n",
        "        labels = np.load(self.imgs_path + 'labels.npy')\n",
        "\n",
        "        if severity_level == 'high':\n",
        "            self.data = data[-10000:]\n",
        "            self.labels = labels[-10000:]\n",
        "        elif severity_level == 'low':\n",
        "            self.data = data[:10000]\n",
        "            self.labels = labels[:10000]\n",
        "        print(self.data.shape)\n",
        "        self.class_map = {\"airplane\" : 0, \"automobile\" : 1, \"bird\" : 2, \"cat\" : 3,\n",
        "                          \"deer\" : 4, \"dog\" : 5, \"frog\" : 6, \"horse\" : 7, \"ship\" : 8,\n",
        "                          \"truck\" : 9}\n",
        "        self.img_dim = (32,32)\n",
        "        self.transform = transform\n",
        "    def __len__(self):\n",
        "        return len(self.data)\n",
        "    def __getitem__(self, idx):\n",
        "        img, class_id = self.data[idx], self.labels[idx]\n",
        "        class_id = torch.tensor([class_id], dtype=torch.long)\n",
        "        if self.transform:\n",
        "            img_tensor = self.transform(img)\n",
        "        return img_tensor, class_id.squeeze()\n",
        "\n",
        "def train_loop_noval(model, optimizer, criterion, trainloader, num_epochs =10): \n",
        "\n",
        "  # Run the training loop for defined number of epochs\n",
        "  training_loss = []\n",
        "  training_accuracy = []\n",
        "  \n",
        "  print(num_epochs)\n",
        "\n",
        "  for epoch in range(0, num_epochs):\n",
        "\n",
        "    # Print epoch\n",
        "    # print(f'Starting epoch {epoch+1}')\n",
        "\n",
        "    model.train()\n",
        "    running_loss = 0.0\n",
        "    acc = 0.0\n",
        "\n",
        "    for i, data in enumerate(trainloader):\n",
        "        # get the inputs; data is a list of [inputs, labels]\n",
        "      inputs, labels = data[0].to(DEVICE), data[1].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",
        "      acc = accuracy(outputs, labels)\n",
        "      loss.backward()\n",
        "      optimizer.step()\n",
        "\n",
        "        # print statistics\n",
        "      running_loss += loss.item()\n",
        "    \n",
        "    if (epoch + 1) % 1 == 0:\n",
        "      print('[%d] epoch. Loss: %.3f. Training accuracy: %.3f'%\n",
        "           (epoch + 1, running_loss, acc))\n",
        "    training_loss.append(running_loss)\n",
        "    training_accuracy.append(acc)\n",
        "    \n",
        "  return model, training_accuracy, training_loss"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "execution": {},
        "id": "2JMFSBDWwIxz",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "cellView": "form",
        "outputId": "3fcd82e4-5877-4bc9-b6a0-0f7ab977ecb8"
      },
      "source": [
        "#@title Set random seed \n",
        "SEED = 42\n",
        "set_seed(seed=SEED)\n",
        "DEVICE = set_device()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Random seed 42 has been set.\n",
            "GPU is enabled in this notebook.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YURpgjzI66P8",
        "cellView": "form"
      },
      "source": [
        "#@title Finetuning function - old \n",
        "\n",
        "def finetuning (model, optimizer, criterion, dataset_list, model_save_name, batch_size=100, num_epochs =10):\n",
        "\n",
        "  # checking if GPU is available\n",
        "  if torch.cuda.is_available():\n",
        "    model = model.to(DEVICE)\n",
        "    criterion = criterion.to(DEVICE)\n",
        "\n",
        "  #concatenating the corrupted data for finetuning based on list\n",
        "  dataset_corrup_concat = torch.utils.data.ConcatDataset(dataset_list)\n",
        "\n",
        "  #load training data\n",
        "  trainloader = torch.utils.data.DataLoader(dataset_corrup_concat, batch_size=batch_size, shuffle=True)\n",
        "\n",
        "  #start train_loop_noval\n",
        "  model, training_accuracy, training_loss = train_loop_noval(model, optimizer, criterion, trainloader, num_epochs)\n",
        "\n",
        "  # Saving the model\n",
        "  path = F\"drive/MyDrive/{model_save_name}\" \n",
        "  torch.save(model.state_dict(), path)\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "HDm5w2gD12P4",
        "cellView": "form",
        "outputId": "0eb96008-a968-417d-cabf-2438c95375e4"
      },
      "source": [
        "#@title Model specification \n",
        "# @title Load a much simpler model (and overwrite)\n",
        "class CNOTP(nn.Module):\n",
        "    def __init__(self):\n",
        "        super().__init__()\n",
        "        self.conv1 = nn.Conv2d(3, 18, 5)\n",
        "        self.pool1 = nn.MaxPool2d(2, 2)\n",
        "        self.pool2 = nn.MaxPool2d(2, 2)\n",
        "        self.conv2 = nn.Conv2d(18, 36, 5)\n",
        "        self.fc1 = nn.Linear(36 * 5 * 5, 120)\n",
        "        self.drop = nn.Dropout(p=0.5)\n",
        "        self.fc2 = nn.Linear(120, 84)\n",
        "        self.fc3 = nn.Linear(84, 10)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.pool1(F.relu(self.conv1(x)))\n",
        "        x = self.pool2(F.relu(self.conv2(x)))\n",
        "        x = torch.flatten(x, 1) # flatten all dimensions except batch\n",
        "        x = self.drop(F.relu(self.fc1(x)))\n",
        "        x = self.drop(F.relu(self.fc2(x)))\n",
        "        x = self.fc3(x)\n",
        "        # x = F.softmax(x)\n",
        "        return x\n",
        "\n",
        "model = CNOTP()\n",
        "print(model)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "CNOTP(\n",
            "  (conv1): Conv2d(3, 18, kernel_size=(5, 5), stride=(1, 1))\n",
            "  (pool1): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  (pool2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  (conv2): Conv2d(18, 36, kernel_size=(5, 5), stride=(1, 1))\n",
            "  (fc1): Linear(in_features=900, out_features=120, bias=True)\n",
            "  (drop): Dropout(p=0.5, inplace=False)\n",
            "  (fc2): Linear(in_features=120, out_features=84, bias=True)\n",
            "  (fc3): Linear(in_features=84, out_features=10, bias=True)\n",
            ")\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pBAkxmIZsl1L",
        "cellView": "form"
      },
      "source": [
        "#@title Define training loop \n",
        "def train_loop(model, optimizer, criterion, trainloader, valloader, num_epochs =10, early_stop_param=10): \n",
        "\n",
        "  # Run the training loop for defined number of epochs\n",
        "  training_loss = []\n",
        "  training_accuracy = []\n",
        "  val_accuracy = []\n",
        "  wait = 0 \n",
        "  best_acc = 0\n",
        "  \n",
        "  print(num_epochs)\n",
        "\n",
        "  for epoch in range(0, num_epochs):\n",
        "\n",
        "    # Print epoch\n",
        "    # print(f'Starting epoch {epoch+1}')\n",
        "\n",
        "    model.train()\n",
        "    running_loss = 0.0\n",
        "    acc = 0.0\n",
        "\n",
        "    for i, data in enumerate(trainloader):\n",
        "        # get the inputs; data is a list of [inputs, labels]\n",
        "      inputs, labels = data[0].to(DEVICE), data[1].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",
        "      acc = accuracy(outputs, labels)\n",
        "      loss.backward()\n",
        "      optimizer.step()\n",
        "\n",
        "        # print statistics\n",
        "      running_loss += loss.item()\n",
        "\n",
        "    # start testing here\n",
        "    val_acc = []#0.0\n",
        "    model.eval()\n",
        "    for i, data in enumerate(valloader):\n",
        "      # get the inputs; data is a list of [inputs, labels]\n",
        "      inputs, labels = data[0].to(DEVICE), data[1].to(DEVICE)\n",
        "\n",
        "      with torch.no_grad():\n",
        "        # forward + backward + optimize\n",
        "        outputs = model(inputs)\n",
        "        val_acc.append(accuracy(outputs, labels))\n",
        "    \n",
        "    val_accuracy.append(np.array(val_acc).mean())\n",
        "    \n",
        "    if (np.array(val_acc).mean() > best_acc):\n",
        "      best_acc = np.array(val_acc).mean()\n",
        "      best_epoch = epoch\n",
        "      best_model = copy.deepcopy(model)\n",
        "      wait = 0\n",
        "    else:\n",
        "      wait += 1\n",
        "\n",
        "    if wait == early_stop_param:\n",
        "       print('Early stepped at [%d] epoch. Loss: %.3f. Training accuracy: %.3f . Test accuracy: %.3f' %\n",
        "           (epoch + 1, running_loss, acc, np.array(val_acc).mean()))\n",
        "       training_loss.append(running_loss)\n",
        "       training_accuracy.append(acc) \n",
        "       break \n",
        "    \n",
        "    if (epoch + 1) % 1 == 0:\n",
        "      print('[%d] epoch. Loss: %.3f. Training accuracy: %.3f . Test accuracy: %.3f' %\n",
        "           (epoch + 1, running_loss, acc, np.array(val_acc).mean()))\n",
        "    training_loss.append(running_loss)\n",
        "    training_accuracy.append(acc)\n",
        "    \n",
        "  return best_model, best_acc, training_accuracy, training_loss, val_accuracy "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8mf6bYmHO1CW",
        "cellView": "form",
        "outputId": "3b74167c-37d6-4a7c-f2e8-e6eb8eda8bfc"
      },
      "source": [
        "#@title Load pretrained VGG19 model\n",
        "#! git clone 'https://github.com/huyvnphan/PyTorch_CIFAR10.git' \n",
        "\n",
        "#! python ./PyTorch_CIFAR10/train.py --download_weights 1\n",
        "\n",
        "from PyTorch_CIFAR10.cifar10_models.vgg import vgg11_bn, vgg13_bn, vgg16_bn, vgg19_bn\n",
        "\n",
        "# Untrained model\n",
        "model = vgg19_bn()\n",
        "\n",
        "# Pretrained model\n",
        "model = vgg19_bn(pretrained=True)\n",
        "model.to(DEVICE)\n",
        "print(model)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "VGG(\n",
            "  (features): Sequential(\n",
            "    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (2): ReLU(inplace=True)\n",
            "    (3): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (4): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (7): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (8): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (9): ReLU(inplace=True)\n",
            "    (10): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (11): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (12): ReLU(inplace=True)\n",
            "    (13): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (14): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (15): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (16): ReLU(inplace=True)\n",
            "    (17): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (18): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (19): ReLU(inplace=True)\n",
            "    (20): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (21): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (22): ReLU(inplace=True)\n",
            "    (23): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (24): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (25): ReLU(inplace=True)\n",
            "    (26): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (27): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (28): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (29): ReLU(inplace=True)\n",
            "    (30): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (31): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (32): ReLU(inplace=True)\n",
            "    (33): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (34): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (35): ReLU(inplace=True)\n",
            "    (36): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (37): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (38): ReLU(inplace=True)\n",
            "    (39): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (40): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (41): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (42): ReLU(inplace=True)\n",
            "    (43): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (44): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (45): ReLU(inplace=True)\n",
            "    (46): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (47): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (48): ReLU(inplace=True)\n",
            "    (49): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (50): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
            "    (51): ReLU(inplace=True)\n",
            "    (52): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  )\n",
            "  (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))\n",
            "  (classifier): Sequential(\n",
            "    (0): Linear(in_features=512, out_features=4096, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Dropout(p=0.5, inplace=False)\n",
            "    (3): Linear(in_features=4096, out_features=4096, bias=True)\n",
            "    (4): ReLU(inplace=True)\n",
            "    (5): Dropout(p=0.5, inplace=False)\n",
            "    (6): Linear(in_features=4096, out_features=10, bias=True)\n",
            "  )\n",
            ")\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lGSqApt1w2JZ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "cellView": "form",
        "outputId": "d554a7c8-de21-43be-93c2-1c4baf621852"
      },
      "source": [
        "#@title VGG19 stuff\n",
        "model = models.vgg19(pretrained = False, progress=True)\n",
        "input_lastLayer = model.classifier[6].in_features\n",
        "model.classifier[6] = nn.Linear(input_lastLayer,10)\n",
        "#print(model)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "VGG(\n",
            "  (features): Sequential(\n",
            "    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (3): ReLU(inplace=True)\n",
            "    (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (6): ReLU(inplace=True)\n",
            "    (7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (8): ReLU(inplace=True)\n",
            "    (9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (11): ReLU(inplace=True)\n",
            "    (12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (13): ReLU(inplace=True)\n",
            "    (14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (15): ReLU(inplace=True)\n",
            "    (16): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (17): ReLU(inplace=True)\n",
            "    (18): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (19): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (20): ReLU(inplace=True)\n",
            "    (21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (22): ReLU(inplace=True)\n",
            "    (23): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (24): ReLU(inplace=True)\n",
            "    (25): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (26): ReLU(inplace=True)\n",
            "    (27): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (29): ReLU(inplace=True)\n",
            "    (30): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (31): ReLU(inplace=True)\n",
            "    (32): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (33): ReLU(inplace=True)\n",
            "    (34): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (35): ReLU(inplace=True)\n",
            "    (36): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  )\n",
            "  (avgpool): AdaptiveAvgPool2d(output_size=(7, 7))\n",
            "  (classifier): Sequential(\n",
            "    (0): Linear(in_features=25088, out_features=4096, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Dropout(p=0.5, inplace=False)\n",
            "    (3): Linear(in_features=4096, out_features=4096, bias=True)\n",
            "    (4): ReLU(inplace=True)\n",
            "    (5): Dropout(p=0.5, inplace=False)\n",
            "    (6): Linear(in_features=4096, out_features=10, bias=True)\n",
            "  )\n",
            ")\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JQsT5yFVybs6",
        "cellView": "form"
      },
      "source": [
        "#@title Training part \n",
        "\n",
        "transform_training = transforms.Compose(\n",
        "    [transforms.ToTensor(),\n",
        "     #transforms.Resize((224,224)),#\n",
        "     transforms.RandomHorizontalFlip(p=0.2),\n",
        "     transforms.RandomVerticalFlip(p=0.2),\n",
        "     transforms.Normalize((.5, .5, .5), (.5, .5, .5))]) #we set back this Normalization values because we could not train the model otherwise\n",
        "\n",
        "transform_test = transforms.Compose(\n",
        "    [transforms.ToTensor(),\n",
        "     transforms.Normalize((.5, .5, .5), (.5, .5, .5))]) #we set back this Normalization values because we could not train the model otherwise\n",
        "\n",
        "batch_size = 100 # that might be adjusted later on \n",
        "\n",
        "# trainset = torchvision.datasets.CIFAR10(root='./data', train=True,\n",
        "#                                         download=True, transform=transform_training)\n",
        "\n",
        "# k_folds = 5\n",
        "# num_epochs = 200\n",
        "\n",
        "# best_accs = []\n",
        "\n",
        "# # Define the K-fold Cross Validator\n",
        "# kfold = KFold(n_splits=k_folds, shuffle=True)\n",
        "    \n",
        "#   # Start print\n",
        "# print('--------------------------------')\n",
        "\n",
        "#   # K-fold Cross Validation model evaluation\n",
        "# for fold, (train_ids, val_ids) in enumerate(kfold.split(trainset)):\n",
        "    \n",
        "#     # Print\n",
        "#   print(f'FOLD {fold}')\n",
        "#   print('--------------------------------')\n",
        "    \n",
        "#     # Sample elements randomly from a given list of ids, no replacement.\n",
        "#   train_subsampler = torch.utils.data.SubsetRandomSampler(train_ids)\n",
        "\n",
        "#   val_subsampler = torch.utils.data.SubsetRandomSampler(val_ids)\n",
        "\n",
        "#   trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size, sampler=train_subsampler)\n",
        "\n",
        "#   valloader = torch.utils.data.DataLoader(trainset,batch_size=batch_size, sampler=val_subsampler)\n",
        "\n",
        "#   # Init the neural network\n",
        "#   model = CNOTP()\n",
        "    \n",
        "#   # Initialize optimizer\n",
        "#   optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)\n",
        "\n",
        "#   criterion = nn.CrossEntropyLoss()\n",
        "#   # checking if GPU is available\n",
        "#   if torch.cuda.is_available():\n",
        "#     model = model.to(DEVICE)\n",
        "#     criterion = criterion.to(DEVICE)\n",
        "    \n",
        "#   best_model, best_acc, training_accuracy, training_loss, val_accuracy = train_loop(model, optimizer, criterion, trainloader, valloader, num_epochs = num_epochs, early_stop_param=10)\n",
        "\n",
        "#   # Saving the model\n",
        "#   model_save_name = f'./model-fold-{fold}.pth'\n",
        "#   path = F\"drive/MyDrive/{model_save_name}\" \n",
        "#   torch.save(best_model.state_dict(), path)\n",
        "\n",
        "#   # plot the training process \n",
        "#   plot_train_loss_acc(training_loss, training_accuracy, val_accuracy, f'model-fold-{fold}-training.png')\n",
        "\n",
        "#   # keep track of the best accs\n",
        "#   best_accs.append(best_acc)\n",
        "\n",
        "# # print and save best accs \n",
        "# print(best_accs)\n",
        "# savepath = '/content/drive/MyDrive/NMA_DL_project/'\n",
        "# with open(savepath+'best_accs.pickle', 'wb') as handle:\n",
        "#   pkl.dump(best_accs, handle, protocol=pkl.HIGHEST_PROTOCOL)\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 480
        },
        "id": "Yn12Pb4-gb7f",
        "outputId": "b2c68db1-da11-4ecb-db6b-2e9afb8b2b83"
      },
      "source": [
        "#@title Testing part - naive model \n",
        "\n",
        "# # load the model \n",
        "# path = 'drive/MyDrive/NMA_DL_project/model-fold-2.pth'\n",
        "# print(path)\n",
        "# model = CNOTP()\n",
        "# model.load_state_dict(torch.load(path))\n",
        "# print(model)\n",
        "\n",
        "# setup test loaders \n",
        "transform_test = transforms.Compose(\n",
        "    [transforms.ToTensor(),\n",
        "     transforms.Normalize((.5, .5, .5), (.5, .5, .5))]) #we set back this Normalization values because we could not train the model otherwise\n",
        "\n",
        "batch_size = 100 # that might be adjusted later on \n",
        "\n",
        "corruption_type_list = ['motion_blur.npy', 'defocus_blur.npy', 'gaussian_noise.npy', 'shot_noise.npy']\n",
        "severity_level = ['low', 'high']\n",
        "\n",
        "img_path = '/content/drive/MyDrive/NMA_DL_project/CIFAR-10-C/CIFAR-10-C/'\n",
        "\n",
        "motion_low = CorruptedDataset(img_path ,corruption_type_list[0], severity_level[0], transform=transform_test)\n",
        "motion_high = CorruptedDataset(img_path , corruption_type_list[0], severity_level[1], transform=transform_test)\n",
        "defocus_low = CorruptedDataset(img_path ,corruption_type_list[1], severity_level[0], transform=transform_test)\n",
        "defocus_high = CorruptedDataset(img_path ,corruption_type_list[1], severity_level[1], transform=transform_test)\n",
        "\n",
        "gaussian_low = CorruptedDataset(img_path, corruption_type_list[2], severity_level[0], transform=transform_test)\n",
        "gaussian_high = CorruptedDataset(img_path, corruption_type_list[2], severity_level[1], transform=transform_test)\n",
        "shot_low = CorruptedDataset(img_path, corruption_type_list[3], severity_level[0], transform=transform_test)\n",
        "shot_high = CorruptedDataset(img_path, corruption_type_list[3], severity_level[1], transform=transform_test)\n",
        "\n",
        "testset = torchvision.datasets.CIFAR10(root='./data', train=False,\n",
        "                                       download=True, transform=transform_test)\n",
        "\n",
        "# test and plot \n",
        "fig_savepath = '/content/drive/MyDrive/NMA_DL_project/naive_test_acc_VGG19.png'\n",
        "res_savepath = '/content/drive/MyDrive/NMA_DL_project/test_results_VGG19'\n",
        "\n",
        "test_accuracies = test_function(model,fig_savepath,res_savepath,\n",
        "                               clean = testset, \n",
        "                               motion_low = motion_low, \n",
        "                               motion_high = motion_high, \n",
        "                               defocus_low = defocus_low,\n",
        "                               defocus_high = defocus_high, \n",
        "                               gaussian_low = gaussian_low,\n",
        "                               gaussian_high = gaussian_high,\n",
        "                               shot_low = shot_low,\n",
        "                               shot_high = shot_high) \n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(10000, 32, 32, 3)\n",
            "(10000, 32, 32, 3)\n",
            "(10000, 32, 32, 3)\n",
            "(10000, 32, 32, 3)\n",
            "(10000, 32, 32, 3)\n",
            "(10000, 32, 32, 3)\n",
            "(10000, 32, 32, 3)\n",
            "(10000, 32, 32, 3)\n",
            "Files already downloaded and verified\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAE2CAYAAABx82k0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAdvklEQVR4nO3debhlVX3m8e8LiIoCYiiNUkghQU0FGcsJHFBBUQQyKIFufByIPMYmkohEVEQhtopTmjbEgDggkUZQ1EJQTCsyqAhVTDJIiwUoOBXKlICBim//sfetOvfcU3VO4a21N+u+n+e5T929z761fpxa/O4+67fW2rJNREQ89K3XdQARETE7ktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISG3TV8Oabb+4FCxZ01XxExEPS0qVLb7c9b9RrnSX0BQsWsGTJkq6aj4h4SJJ0y+pey5BLREQlktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqERnC4t+HwuOPKdoezd/YO+i7UVEPBi5Q4+IqEQSekREJZLQIyIqkYQeEVGJh2RRNPotReuIbuQOPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohITJXRJe0m6QdKNko4c8fqTJJ0v6QpJV0t6+eyHGhERazI2oUtaHzgBeBmwEDhQ0sKhy44CzrC9E3AA8M+zHWhERKzZJHfozwRutL3M9v3A6cB+Q9cY2KT9flPgZ7MXYkRETGKSzbm2AH46cHwr8Kyha94DfEPS3wCPAvaYlegiImJis1UUPRD4jO35wMuBUyXN+LslHSJpiaQly5cvn6WmIyICJkvotwFbDhzPb88NOhg4A8D294BHAJsP/0W2T7K9yPaiefPmPbiIIyJipEkS+mXAtpK2lrQhTdFz8dA1PwFeDCDpj2kSem7BIyIKGpvQba8ADgXOA66nmc1yraRjJe3bXnY48AZJVwH/B3itba+roCMiYqaJnlhk+1zg3KFzRw98fx2w2+yGFhERayMrRSMiKpFnikZEJ/Ls2dmXO/SIiEokoUdEVCIJPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlUhCj4ioRBJ6REQlktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqMQGXQfwUJcnl8ck0k+ihNyhR0RUIgk9IqISSegREZVIQo+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEokoUdEVCIJPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiYkSuqS9JN0g6UZJR67mmv0lXSfpWkmnzW6YERExztgHXEhaHzgB2BO4FbhM0mLb1w1csy3wdmA323dIety6CjgiIkab5A79mcCNtpfZvh84Hdhv6Jo3ACfYvgPA9q9mN8yIiBhnkoS+BfDTgeNb23ODngI8RdJ3JF0iaa9Rf5GkQyQtkbRk+fLlDy7iiIgYabaKohsA2wK7AwcCn5D0mOGLbJ9ke5HtRfPmzZulpiMiAiZL6LcBWw4cz2/PDboVWGz7Ads3Af+PJsFHREQhkyT0y4BtJW0taUPgAGDx0DVfprk7R9LmNEMwy2YxzoiIGGNsQre9AjgUOA+4HjjD9rWSjpW0b3vZecCvJV0HnA8cYfvX6yroiIiYaey0RQDb5wLnDp07euB7A29pvyIiogNZKRoRUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEokoUdEVCIJPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlUhCj4ioRBJ6REQlktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEokoUdEVCIJPSKiEknoERGVSEKPiKjERAld0l6SbpB0o6Qj13DdX0iypEWzF2JERExibEKXtD5wAvAyYCFwoKSFI67bGDgM+P5sBxkREeNNcof+TOBG28ts3w+cDuw34rp/AI4DfjuL8UVExIQmSehbAD8dOL61PbeSpJ2BLW2fs6a/SNIhkpZIWrJ8+fK1DjYiIlbv9y6KSloP+Chw+LhrbZ9ke5HtRfPmzft9m46IiAGTJPTbgC0Hjue356ZsDGwHfFvSzcCzgcUpjEZElDVJQr8M2FbS1pI2BA4AFk+9aPsu25vbXmB7AXAJsK/tJesk4oiIGGlsQre9AjgUOA+4HjjD9rWSjpW077oOMCIiJrPBJBfZPhc4d+jc0au5dvffP6yIiFhbWSkaEVGJJPSIiEokoUdEVCIJPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlUhCj4ioRBJ6REQlktAjIiqRhB4RUYkk9IiISkz0CLrovwVHnlO0vZs/sHfR9iJivNyhR0RUInfoETHn1fIJN3foERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlchK0Yg5puSqyOz5U1bu0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlUhCj4ioxETz0CXtBRwPrA+cbPsDQ6+/BfgrYAWwHHi97VtmOdaItZY51zGXjL1Dl7Q+cALwMmAhcKCkhUOXXQEssr098AXgg7MdaERErNkkQy7PBG60vcz2/cDpwH6DF9g+3/a97eElwPzZDTMiIsaZJKFvAfx04PjW9tzqHAx8bdQLkg6RtETSkuXLl08eZUREjDWrRVFJBwGLgA+Net32SbYX2V40b9682Ww6ImLOm6Qoehuw5cDx/PbcNJL2AN4JvMD2f85OeBERMalJ7tAvA7aVtLWkDYEDgMWDF0jaCTgR2Nf2r2Y/zIiIGGdsQre9AjgUOA+4HjjD9rWSjpW0b3vZh4BHA2dKulLS4tX8dRERsY5MNA/d9rnAuUPnjh74fo9ZjisiItZSVopGRFQiCT0iohJJ6BERlUhCj4ioRBJ6REQlktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEokoUdEVCIJPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlUhCj4ioRBJ6REQlktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqEQSekREJZLQIyIqMVFCl7SXpBsk3SjpyBGvP1zS59vXvy9pwWwHGhERazY2oUtaHzgBeBmwEDhQ0sKhyw4G7rD9R8A/AsfNdqAREbFmk9yhPxO40fYy2/cDpwP7DV2zH3BK+/0XgBdL0uyFGRER48j2mi+QXgnsZfuv2uNXA8+yfejANde019zaHv+4veb2ob/rEOCQ9vCpwA2z9R8yoc2B28deVUZfYulLHNCfWPoSBySWUfoSB3QTy1a25416YYOSUdg+CTipZJuDJC2xvair9gf1JZa+xAH9iaUvcUBi6XMc0K9YYLIhl9uALQeO57fnRl4jaQNgU+DXsxFgRERMZpKEfhmwraStJW0IHAAsHrpmMfCa9vtXAt/yuLGciIiYVWOHXGyvkHQocB6wPvAp29dKOhZYYnsx8EngVEk3Ar+hSfp91Nlwzwh9iaUvcUB/YulLHJBYRulLHNCvWMYXRSMi4qEhK0UjIiqRhB4RUYkk9IiISiShFyTpxZIe2XUc0W/pJzNJekTXMTwUzImiaLsfzeMZmNVj+ycdxHEK8ByamUAXARcCF9u+o4NYdgPeA2xF874IsO0nF47jYuACmvfjO7bvKdn+UCy7AguY3k8+20Ec6Scz47gR+CXN+3ERzftxV8kYBmKZB7yBmX3l9V3EM6j6hC7pb4B303SG37WnbXv7DmN6Is18/bcCT7RddMVuG8MPgb8DlgL/NXXedtEFYZK2Bp7Xfj0b+E/gItt/VziOU4FtgCtZ9X7Y9ptLxjEUU/rJ9FieRNNPdgNeDtxpe8cO4vguzS+V4ffki6VjGVa8g3TgMOCpXXTAYZIOoumQT6fZ/+GfaDpGF+6y/bWO2l7J9k2Sfgvc3369EPjjDkJZBCzsw4K49JOZJM2nSeTPA3YArgUu7iicjWy/raO212gu3KGfD+xpe0UPYrkd+DHwL8D5tm/uIIad22/3p1kodhbNXTEAti8vHM+PaZLWaTRJ60rbv1vzT62TOM4E3mz756XbHhFL+snMeH5Hs2r9fba/UrLtEbG8F/iu7XO7jGOUuZDQP0mzs+M5TO+QH+0onj8Bng88F9gWuMH2qwu2f/4aXrbtF5WKBUDSYTTvxZbAD2nG0y+0/eNC7Z8NGNgY2BG4lOn9ZN8ScYyIK/1kgKQdaN6L5wNPAn4EXGD7kwVjuIemrwh4FE0/eYBVdYVNSsWyOnNhyOUn7deG7VdnJG1C0xm3oimobMqqcf0ibL+wZHvj2D4eOF7So4HX0RTg5tPcFZbw4ULtTCz9ZCbbV7Wf5n5MM+xyEPACmm1HSsWwcam2Hqzq79D7RNLVNON+F9Pchd7aYSxvGXH6LmCp7SsLxvERmjuvRwPfo53FYHtZqRj6Jv1kZBxLgIcDUwXJi2zfUqr9oVh2HnH6LuCWrod2q0/o7RSjvwf+BFg5l7X0R8ahmB7dxvDvHcZwGk0h8Oz21CuAq2nuCM+0/cFCcbyS5n/OX5Zobw1xTH2cHnQXsAQ4vItfMOkn0+KYZ3t5ibbGkXQJsDPwg/bU04FraD5J/bXtb3QV21xYWPQ5mrHZrYFjgJtpiivFSdpO0hU0FfrrJC2VtF0XsdAMa+xs+3DbhwO7AI+jGaN8bakgbH8BeJakD7df+5Rqe8j/Ao4AtqB5b95KU6g9HfhUyUDST0a6X9JHJS1pvz4iadOC7Q/6GbCT7V1s70JTe1kG7AkU+QW3OnMhof9BWzh5wPYF7eT/ru7OTwLeYnsr208CDqe77Tcfx0Dxj6a483jb9w2dX6ckvZ9maul17debJb2vVPsD9rV9ou17bN/dPl3rpbY/D2xWOJb0k5k+BdxDM+tmf+Bu4NMF2x/0FNvXTh3Yvg54Wh+GCedCUfSB9s+fS9qb5rfrYzuK5VG2V84esP1tSY/qKJbPAd+XNDUFbB/gtDae6wrGsTew49RUxXaV5BXAOwrGAHCvpP1pHnIOzYKe37bflx6XTD+ZaRvbfzFwfIykYmP4Q66V9HGaT28Af0nzSerhrMo3nZgLY+ivoCmibAl8DNgEOKZ9MEfpWL4EXA6c2p46CNjF9p+VjqWNZxHNYg1olt0v6SCGq4Hdbf+mPX4s8O3SK3klPRk4nmbJvYFLaFZI3kbzb1RsEUv6ycgYvgccMfXv0G5J8GHbz+kglkcCb6Ip5gN8B/hnmhuAjTqtedSe0PtE0mY04/hTHeEi4D0l9+iQtIntu9vEOcNUYi0Yz4HAB4DzaebzPh84sh3qmJPST0bGsyNwCk3hUTT73LzW9lUl4+i76hO6pKcAH6cZ99tO0vY046Xv7Ti0Tkj6qu1XSLqJ6UMJnWy61Mb0BOAZ7eGltn9RsO2/t/1BSR9jxNBKl3u5dKmP/aSNaxOaAO7uoO0zbO8v6QeM7iud7Q81ZS4k9AtoZi+caHun9tw1tovNGhhYjThSV6sRu7SaubwrlVpaLmkf22dLes2o122fUiKONpb0kyGrmQe/UskV35KeYPvnkrZaTSydzIsfNBeKohvZvlTS4LnSk/97txoRQNIWrNoWFQDbFxZq/iNreM0Umolk++z2z2KJew3ST2bqzerMqX1++pC4V2cuJPTbJW1De+fTLmQpugGT7QsmuU7SF4cq+euMpONoq/MMbBdLs/f2Ojfp0nJJe9r+t3UdTzs091Zm7nFdbIpr+slMto+Z5DpJb7f9/nUdT9vWnwPH0UzpFD3ay2UuDLk8mWYO767AHcBNwEHuYAe7cSRdMTUsVKCtG4DtbZecS7zWJF1ue43DM7PUzlU0uxsO73G9dF23vbbST2Yq1U/atm4E9rF9fYn21kb1d+jtZP892nmz67nDJ+JMoORv12XAwyi7OOTB0PhLZsUK2x8v1NbvK/1kplL9BOCXfUzmUHFCX10xZWosvWQxpU8GZnPcC1wp6ZtM3y62b7M61mnyGpiWd7akNwFfYvr7UXR6Xl+kn8zUDrUALJH0eeDLTH9PzlrXMYxTbUJnVTFlav/iQX0dZypxlzG1KGQpUHxxVQ8tZXofOWLgNQOdTM8bI/1kphLvyeA+Q/cCLxk4Ns1DQLplu+ovmsUIjxk43gz4VNdxrSbWl3Qdw0AsX+w6hjaOs7qOoY1jz65jGIhlzvUTYLc1nQPe0fV7MRDL27tqey4URWcUkEoWlYba7cUT1CdR6j2S9Crg67bvkXQUzbak73XhR5yNU7joln4ys50Z73/Jf5O10WVcNQ+5TFlP0mZul023Y6Zd/Xd/khFPUO+pUr/p32X7TEnPBfYAPkSzsvdZhdqfVMmiW/pJS9JzaGaozRuqi21Cuadara2SfWWauZDQPwJ8T81DgAFeBfzPjmLpxRPUe2YqYe0NnGT7HDUP4e2bkh9l009W2ZDmaVYbMH2R0d00O2L2UWfDHtUPuQBIWsiqlYffcrN/cRdxfIAePEF9EgU/Sn+VZkfDPWmGW+6j2c9lh3Xd9tooPOSSfjKzna1s36IePMVpnK6GdGFu3KHTJvBOkviQqWGERQPnii1zX0tvK9TO/sBeNFuh3tlu1HXEmJ/pws0F20o/mWnj9ilOjwWQdDvwGtvXFGp/JUm72f7OGs6dOeLHipgTd+gxU18Kb5KeNOq87Z8UjuMhUZwtrUf95LvAO90++EPS7sD7bO9aMo627d4WaOfEHXpfqHkG4rtp9vwGuAA41vZdHYTTl8LbOayaB/4Imme/3kDzUO+SelWcVfN0reEHmx/bQSh96SedP8XpoVCgTUIv61M0Twffvz1+Nc1zEf98tT+x7vSi8Gb76YPH7ba6b+oglN4UZyX9C7AR8ELgZJri36VdxEJP+gmwTNK7mP4Up9LP8Ox9gTZDLgVJutL2juPOFYqlt4U3ST8YTvQF2uxNcVbS1ba3H/jz0cDXbD+vg1h60U/68BSngVh6W6DNHXpZ90l6rqc/F/G+jmLpReFt6KPrejTJ9GclY2j1qTg71SfulfRE4NfAEzqKpRf9pE3cfdk/pjcF2mFJ6GX9NXBKO5a+8rmIXQTiCfcjL2Dwo+sKmjH1L3YQx+a0+5cMFGp/2EEcAF+V9BiacfzLaRLoyV0E0pd+0of96gecBLxlqEA7tUV3pzLk0oEun4s4EEOfCrSdG3hO5LTirO3SxdnhuB4OPKLLf5c+FGj7tF+9pKuGh+JGnetC7tALkHSQ7X8d3tK34618e1GglfRvwKts39kebwacbvulJePoQ3FW0otsf2tgm9bB13AH27P2qEDbp/3q+1CgHSkJvYyp6VWjno/Y1UekbTz9MWbHSLqygzjmTSVzaMZKJT2ugzimsX25pNJTFl8AfIvp27SuDIlutmfddaBAe4ykjwDFZr30dL/619MUaKf+PS5qz3UuCb0A2ye23/7fUSvMOggJ+lOg/S9JT5paSKTmierFf8n1oThr+93tn68r2e4YXRdoR+1XP9g/iu9A2bMC7TRJ6GV9jCZRjDtXQl8KtO8ELpZ0QRvH84BDOoijL8VZJB1GM/x1D/AJmv5xpO1vdBBOpwVa21sDSNqfZiXv3e1wx87AP5SKY1DPCrTTpChawMAKs78F/nHgpU2AP+uymNKTAu3mwLPbw0ts395VLH0wVWCT9FLgjcBRwKldLy3vskA7MCf/uTSJ/MPA0baLr+TtU4F2WO7Qy+jNCrOeFmh3ZdVsG4Cvlg6gL8XZqXDaP18OfNb2tZr6ByoVQP8KtIMreT/R8TbLfSrQTpOEXoDtC4ALJH2mByvMelWgbVciPgP4XHvqMEm72n5H4VD6VJxdKukbNFMn3y5pY+B3hWPoW4H2Nkkn0qzkPa79tLBeyQB6WqCdJkMuBUnajmaq01TH6PMWoKXiuBrY0fbv2uP1gStsb184jqU0w1+DxdkvdTHMIWk9YEdgWbtq9bHAfNtXl46lLyRtRLOS9we2f9Su5H16ybqCpJuY+dD5lQm09A6UoxT9DRcrV5htZXsr4PD2XBc+NuG5Eh4z8P2mHcUwVZw9VdK/AhcCb+8olufQLGq6U9JBNGPonSwsknSYpE3UOFnS5ZJeMv4nZ5fte22fZftH7fHPSxeJbW/dJu23ATu0BdtPA1fRk825MuRSVrYAnen9wBWSzqe583k+cGTpIGx/vV1MNFWc/dsOi7MfB3aQtAPNL/2Tgc/SDIOU9nrbx7cF2j+gWYB2KtDFjJu+OMr2GW2B9kU0BdpePAc3d+hlLZP0LkkL2q+j6H4L0KmvogXagfn3Z9Ek0bNopgk+x/bnS8UxZFdg9/br2Wu8ct1a4WYsdD/gn2yfwOiaRwkzCrTQ3UOQe2JGgZbm/6vOZQy9IK3aAnQqmU1tAXrn6n9qncXS6Ragkpba3kU9edLLiOLsgcBlHRRnaefkfx14Hc0nll8BV5XeUriN5dPAFjQF2h1oPsV92/YupWPpC/Voq+VhSegFSVpEM1a7gFXDXS5dAGxj6bRAK+kS4GrgT4HTh1+3XXQlXl+Ks23bfwj8N5pfKBep2f1xd9uf7SCWFGiH9KFAuzpJ6AVJuoFmhdk1DExDs31LB7F0+ozGdjHRHsBxwNHDr9s+pUQcA/FcTZM0f9MeP5bmTrR4Qu+TdmjsStv/0RZodwaO76LPxngpipa13PbZXQfR6rRA2xYcT5d0ve2rSrW7Br0ozgJIuodV0+E2BB4G/LvtLmYA9alAG2MkoZf1bkknA99k+oKELnbR68sWoPdJ+ibweNvbSdoe2Nd2kVWAA3PvzwK+TTOODvA2278oEcMw2ysLoO0K0f3orki7wrYlTRVoPynp4I5iiTEy5FJQO7/5acC1rBpyse3iW2/2pUDbFgCPAE60vVN77hrb2xVqv1fF2dWRdMXU+1O43d4UaGO83KGX9QzbT+06iNY2wJY0U1c3AF5MM6e29JjxRrYvHdqqZEXB9h+QdBIwX9L/Hn6xdHEWYGj/lPVonuf529JxtP6SpkB7sO1ftAXaD3UUS4yRhF7WdyUttH1d14HQTM+bUaDtwO2StqEdM5b0SuDnBdt/BU1x9qU0u+f1weD+KSuAm2mGXYprh50+OnD8E5ox9OihDLkUJOl6mjvjm2jG0EV30xYvtv3c0u2OiOPJrHrA7h00781/Lz2LQtIOPSnO9krPCrQxRhJ6Qe2GTzN0NG3xxTSLZzop0A5v3ws8kmZ44T/aOIpu46vmoQUfp6Pi7FAsjwAOZuaDmTt9zNlggdZ2JzOAYs2y9L8g27eM+uoonNfRLBjZi+Yj/j40ww+lTG05sIjm6Umb0WzS9Ua6eYLTJ2g243oAoF04c0AHcUAz8+gPaYaBLgDm0zy9qFNufJkmruihjKHPXZ0WaG0fAyDpQmBn2/e0x++hefxbaV0XZwf9ke1XSdrP9imSTqOZhVRczwq0MUYS+tzVlwLt44H7B47vb8+V1nVxdtAD7Z93tls0/ALo6mEbvSnQxnhJ6HPXs4Er2037uyzQfha4VNKX2uM/BT5TOAaA/0FTnH2apNtoi7MdxAFwUrtO4ChgMc3umO/qIhDbr+ui3XhwUhSdo3pWoN0ZeF57eKHtKwq23avibBvT4ayaWTI1BnQnsNT2lYVj6WWBNkbLHfoc1afNlWxfDlzeUfNTy+yfSrPs/ys0SfTVwKUdxbQLzVj11L4/r6DZmfKNks60/cGCsZwK/JCmEHoszaeW6wu2H2shd+gRrCzO7j1QnN0YOMf28zuK5eVTe9S3e9afQzMjaanthQVjucL2TpKutr29pIcBF9nu8gEgsRqZthjR6EtxFpoC6H8OHD9AMz/+vqHzJQwXaDeluwJtjJEhl4hGX4qz0GzL8H1JX2mP9wFOa7c3Lj0rqTcF2hgvQy4RrS6LsyNiWcSqnTC/Y3tJR3H0pkAb4yWhR8RqtYuaRhVoFwClC7QxRhJ6RKxWnwq0MV6KohGxJn0q0MYYKYpGxJr0qUAbY2TIJSLWqC8F2hgvCT0iohIZQ4+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEr8fwSVWSss0TV7AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "si36RlhxAvJM"
      },
      "source": [
        "#@title RSA part - only naive model  \n",
        "#extract activations and compute RDMs \n",
        "all_activations, all_RDMs = get_activations_and_RDMs(model,\n",
        "                               clean = testset, \n",
        "                               motion_low = motion_low, \n",
        "                               motion_high = motion_high, \n",
        "                               defocus_low = defocus_low,\n",
        "                               defocus_high = defocus_high, \n",
        "                               gaussian_low = gaussian_low,\n",
        "                               gaussian_high = gaussian_high,\n",
        "                               shot_low = shot_low,\n",
        "                               shot_high = shot_high) \n",
        "\n",
        "# Visualizing RSA results \n",
        "all_RDM_correlations = dict()\n",
        "\n",
        "for test_idx, (test_name, RDMs) in enumerate(all_RDMs.items()): \n",
        "\n",
        "  if test_name == 'clean': \n",
        "    clean_RDMs = RDMs\n",
        "    continue\n",
        "\n",
        "  RDM_corr = dict()\n",
        "\n",
        "  for layer_idx, (layer, RDM) in enumerate(RDMs.items()): \n",
        "\n",
        "    rdv1 = flatten_RDM(clean_RDMs[layer])\n",
        "    rdv2 = flatten_RDM(RDM)\n",
        "    corr, _  =  pearsonr(rdv1.squeeze(), rdv2.squeeze())\n",
        "    \n",
        "    RDM_corr[layer] =corr\n",
        "\n",
        "  all_RDM_correlations[test_name] = RDM_corr"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 919
        },
        "id": "lKRZ4xiIA6qQ",
        "outputId": "25f70586-88e7-481a-913f-fab2f5512ba2"
      },
      "source": [
        "#@title plot RDM correlations \n",
        "\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "fig = plt.figure(figsize=(20,16))\n",
        "ax = fig.add_subplot(1, 1, 1)\n",
        "test_labels = []\n",
        "layer_names = []\n",
        "\n",
        "for test_name, RDM_corr in all_RDM_correlations.items(): \n",
        "  test_labels.append(test_name)\n",
        "  if test_name == 'motion_low':\n",
        "    for layer,_ in RDM_corr.items(): \n",
        "      layer_names.append(layer)\n",
        "\n",
        "  df = pd.DataFrame(RDM_corr.items(), columns = ['Layer', 'Correlation'])\n",
        "\n",
        "  df.plot(ax = ax)\n",
        "\n",
        "plt.legend(test_labels)\n",
        "plt.xticks([0,1,2,3,4,5,6,7], layer_names)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1440x1152 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "ho6KkIuPHElC",
        "outputId": "a5cab0b3-4c0c-4ab0-9356-5fa66a7de857"
      },
      "source": [
        "#@title RDM plotting \n",
        "f = plt.figure(figsize=(20, 20))\n",
        "ct = 1\n",
        "\n",
        "for test_idx, (test_name, RDMs) in enumerate(all_RDMs.items()): \n",
        "  df = pd.DataFrame(RDMs.items())\n",
        "\n",
        "  for layer_idx in range(len(RDMs.keys())):\n",
        "    ax = f.add_subplot(7,len(RDMs.keys()),ct)\n",
        "    ct += 1\n",
        "    plt.imshow(df[1][layer_idx], vmin = 0, vmax = 1)\n",
        "    ax.get_xaxis().set_visible(False)\n",
        "    ax.get_yaxis().set_visible(False)\n",
        "    if layer_idx == 0: \n",
        "      ax.set_ylabel(test_name)\n",
        "      ax.get_yaxis().set_visible(True)\n",
        "      ax.set_yticks([])\n",
        "    if test_idx ==0:\n",
        "      plt.title(df[0][layer_idx])\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "error",
          "ename": "ValueError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-30-c87d335fccf9>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m   \u001b[0;32mfor\u001b[0m \u001b[0mlayer_idx\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mRDMs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkeys\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 9\u001b[0;31m     \u001b[0max\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd_subplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m7\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mRDMs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkeys\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mct\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     10\u001b[0m     \u001b[0mct\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     11\u001b[0m     \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mimshow\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdf\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mlayer_idx\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvmin\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvmax\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/matplotlib/figure.py\u001b[0m in \u001b[0;36madd_subplot\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1417\u001b[0m                     \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_axstack\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mremove\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0max\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1418\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1419\u001b[0;31m             \u001b[0ma\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msubplot_class_factory\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mprojection_class\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1420\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1421\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_add_axes_internal\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ma\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/matplotlib/axes/_subplots.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, fig, *args, **kwargs)\u001b[0m\n\u001b[1;32m     64\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0mnum\u001b[0m \u001b[0;34m<\u001b[0m \u001b[0;36m1\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mnum\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0mrows\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mcols\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     65\u001b[0m                     raise ValueError(\n\u001b[0;32m---> 66\u001b[0;31m                         f\"num must be 1 <= num <= {rows*cols}, not {num}\")\n\u001b[0m\u001b[1;32m     67\u001b[0m                 self._subplotspec = GridSpec(\n\u001b[1;32m     68\u001b[0m                         rows, cols, figure=self.figure)[int(num) - 1]\n",
            "\u001b[0;31mValueError\u001b[0m: num must be 1 <= num <= 49, not 50"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1440x1440 with 49 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "cellView": "form",
        "id": "olyaErxsjb7-"
      },
      "source": [
        "#@title Load fresh VGG19 model to avoid problems with hooks \n",
        "#! git clone 'https://github.com/huyvnphan/PyTorch_CIFAR10.git' \n",
        "\n",
        "#! python ./PyTorch_CIFAR10/train.py --download_weights 1\n",
        "\n",
        "from PyTorch_CIFAR10.cifar10_models.vgg import vgg11_bn, vgg13_bn, vgg16_bn, vgg19_bn\n",
        "\n",
        "# Untrained model\n",
        "model = vgg19_bn()\n",
        "\n",
        "# Pretrained model\n",
        "model = vgg19_bn(pretrained=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Zty28BYp0R3B"
      },
      "source": [
        "#@title Finetuning function\n",
        "\n",
        "def finetuning (model, optimizer, criterion, dataset_list_train,dataset_list_test, model_save_name, batch_size=100, num_epochs =10):\n",
        "\n",
        "  # checking if GPU is available\n",
        "  if torch.cuda.is_available():\n",
        "    model = model.to(DEVICE)\n",
        "    criterion = criterion.to(DEVICE)\n",
        "\n",
        "  #concatenating the corrupted data for finetuning based on list\n",
        "  dataset_corrup_concat_train = torch.utils.data.ConcatDataset(dataset_list_train)\n",
        "  dataset_corrup_concat_test = torch.utils.data.ConcatDataset(dataset_list_test)\n",
        "\n",
        "\n",
        "  #load training data\n",
        "  trainloader = torch.utils.data.DataLoader(dataset_corrup_concat_train, batch_size=batch_size, shuffle=True)\n",
        "\n",
        "  #load test data\n",
        "  testloader = torch.utils.data.DataLoader(dataset_corrup_concat_test, batch_size=batch_size, shuffle=True)\n",
        "\n",
        "  #start train_loop_noval\n",
        "  model, best_acc, training_accuracy, training_loss, test_accuracy = train_loop(model, optimizer, criterion, trainloader, testloader, num_epochs = num_epochs, early_stop_param=10)\n",
        "\n",
        "  # Saving the model\n",
        "  path = F\"drive/MyDrive/{model_save_name}\" \n",
        "  torch.save(model.state_dict(), path)\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6jm3NEznBxfL",
        "cellView": "form"
      },
      "source": [
        "#@title Split corrupted data into training and test datasets\n",
        "motion_low_train, motion_low_test = torch.utils.data.random_split(motion_low, [8000, 2000])\n",
        "motion_high_train, motion_high_test = torch.utils.data.random_split(motion_high, [8000, 2000])\n",
        "gaussian_low_train, gaussian_low_test = torch.utils.data.random_split(gaussian_low, [8000, 2000])\n",
        "gaussian_high_train, gaussian_high_test = torch.utils.data.random_split(gaussian_high, [8000, 2000])\n",
        "shot_low_train, shot_low_test = torch.utils.data.random_split(shot_low, [8000, 2000])\n",
        "shot_high_train, shot_high_test = torch.utils.data.random_split(shot_high, [8000, 2000])\n",
        "defocus_low_train, defocus_low_test = torch.utils.data.random_split(defocus_low, [8000, 2000])\n",
        "defocus_high_train, defocus_high_test = torch.utils.data.random_split(defocus_high, [8000, 2000])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 684
        },
        "id": "DHnn4KmbIrh1",
        "cellView": "form",
        "outputId": "8822614d-9f32-4f75-db0f-d12baf5699cd"
      },
      "source": [
        "#@title Run first finetuning and testing - only low noise model \n",
        "\n",
        "# path = 'drive/MyDrive/NMA_DL_project/model-fold-2.pth'\n",
        "# model = CNOTP()\n",
        "# model.load_state_dict(torch.load(path))\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "dataset_list_train = [motion_low_train,defocus_low_train,gaussian_low_train,shot_low_train]\n",
        "dataset_list_test = [motion_low_test,defocus_low_test,gaussian_low_test,shot_low_test]\n",
        "\n",
        "model_save_name = 'alltypes_low_model_finetune_test_incl'\n",
        "\n",
        "finetuning (model, optimizer, criterion, dataset_list_train, dataset_list_test,model_save_name, batch_size=100, num_epochs =50)\n",
        "\n",
        "fig_savepath = '/content/drive/MyDrive/NMA_DL_project/low_noise_model_test_acc_VGG19.png'\n",
        "\n",
        "test_accuracies = test_function(model,fig_savepath,res_savepath,\n",
        "                               clean = testset, \n",
        "                               motion_low = motion_low_test, \n",
        "                               motion_high = motion_high_test, \n",
        "                               defocus_low = defocus_low_test,\n",
        "                               defocus_high = defocus_high_test, \n",
        "                               gaussian_low = gaussian_low_test,\n",
        "                               gaussian_high = gaussian_high_test,\n",
        "                               shot_low = shot_low_test,\n",
        "                               shot_high = shot_high_test)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "20\n",
            "[1] epoch. Loss: 121.415. Training accuracy: 0.840 . Test accuracy: 0.902\n",
            "[2] epoch. Loss: 55.447. Training accuracy: 0.950 . Test accuracy: 0.936\n",
            "[3] epoch. Loss: 35.269. Training accuracy: 0.990 . Test accuracy: 0.969\n",
            "[4] epoch. Loss: 26.441. Training accuracy: 0.980 . Test accuracy: 0.957\n",
            "[5] epoch. Loss: 22.873. Training accuracy: 0.990 . Test accuracy: 0.980\n",
            "[6] epoch. Loss: 17.791. Training accuracy: 1.000 . Test accuracy: 0.952\n",
            "[7] epoch. Loss: 18.160. Training accuracy: 0.990 . Test accuracy: 0.975\n",
            "[8] epoch. Loss: 15.832. Training accuracy: 1.000 . Test accuracy: 0.972\n",
            "[9] epoch. Loss: 13.905. Training accuracy: 1.000 . Test accuracy: 0.983\n",
            "[10] epoch. Loss: 15.366. Training accuracy: 1.000 . Test accuracy: 0.974\n",
            "[11] epoch. Loss: 12.134. Training accuracy: 0.990 . Test accuracy: 0.980\n",
            "[12] epoch. Loss: 11.319. Training accuracy: 1.000 . Test accuracy: 0.981\n",
            "[13] epoch. Loss: 10.936. Training accuracy: 1.000 . Test accuracy: 0.978\n",
            "[14] epoch. Loss: 12.777. Training accuracy: 0.990 . Test accuracy: 0.979\n",
            "[15] epoch. Loss: 9.621. Training accuracy: 1.000 . Test accuracy: 0.985\n",
            "[16] epoch. Loss: 11.508. Training accuracy: 0.990 . Test accuracy: 0.981\n",
            "[17] epoch. Loss: 8.799. Training accuracy: 1.000 . Test accuracy: 0.987\n",
            "[18] epoch. Loss: 6.244. Training accuracy: 1.000 . Test accuracy: 0.986\n",
            "[19] epoch. Loss: 7.394. Training accuracy: 0.980 . Test accuracy: 0.986\n",
            "[20] epoch. Loss: 9.529. Training accuracy: 1.000 . Test accuracy: 0.977\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAE2CAYAAABx82k0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAeaElEQVR4nO3de7hdVX3u8e8LiIoCYolWCRCkqKUICPEGXvCCIgj0ohTOwUeRI4+1VFqRiooXqFXx1mMtB4mKApUiKGoQFHsUuagICTe5yDFy0SBqUG4tWKC+5485d7Ky9krWCuyMORn7/TzPfrLnXHNn/FgZ/PZc4zfGmLJNREQ8/K3TdQARETEzktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqIS63XV8Kabbup58+Z11XxExMPS4sWLb7M9Z9RrnSX0efPmsWjRoq6aj4h4WJJ086pey5BLREQlktAjIiqRhB4RUYmxCV3SiZJ+LenqVbwuSf8saYmkqyTtNPNhRkTEOJPcoX8e2GM1r78S2Kb9OgQ4/qGHFRERa2psQrd9AfDb1VyyL3CyGxcDj5P0pJkKMCIiJjMTY+ibAT8fOF7anouIiIKKFkUlHSJpkaRFy5YtK9l0RET1ZmJh0S3A5gPHc9tz09heACwAmD9//oN+ssa8I89+sD/6oNz0ob2KthcR8WDMREJfCBwq6TTgOcCdtm+dgb83HqbyCzeiG2MTuqR/A3YDNpW0FHgv8AgA258CzgH2BJYA9wAHra1gIyJi1cYmdNsHjHndwF/PWEQRMSvkk9zMy0rRiIhKdLbbYi1ylxGTSD+JEnKHHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlUhCj4ioRBJ6REQlktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEokoUdEVCIJPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohLrdR1AxNo078izi7V104f2KtZWxCi5Q4+IqETu0CNi1iv5SQ7W3qe5ie7QJe0h6XpJSyQdOeL1LSSdJ+lySVdJ2nPmQ42IiNUZe4cuaV3gOGB3YClwqaSFtq8duOwo4HTbx0vaFjgHmLcW4o2Ihyh1hXpNMuTybGCJ7RsAJJ0G7AsMJnQDG7Xfbwz8YiaDjPFq+cgYEQ/eJEMumwE/Hzhe2p4b9D7gQElLae7O/2bUXyTpEEmLJC1atmzZgwg3IiJWZaZmuRwAfN72XGBP4BRJ0/5u2wtsz7c9f86cOTPUdEREwGQJ/RZg84Hjue25QQcDpwPY/gHwKGDTmQgwIiImM0lCvxTYRtJWktYH9gcWDl3zM+ClAJL+mCahZ0wlIqKgsQnd9gPAocC5wHU0s1mukXSMpH3ayw4H3ijpSuDfgNfb9toKOiIipptoYZHtc2iKnYPn3jPw/bXArjMbWkRErIks/Y+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEokoUdEVCIJPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlUhCj4ioRBJ6REQlktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEokoUdEVCIJPSKiEknoERGVSEKPiKjERAld0h6Srpe0RNKRq7hmP0nXSrpG0qkzG2ZERIyz3rgLJK0LHAfsDiwFLpW00Pa1A9dsA7wD2NX27ZKesLYCjoiI0Sa5Q382sMT2DbbvA04D9h265o3AcbZvB7D965kNMyIixpkkoW8G/HzgeGl7btBTgadK+p6kiyXtMeovknSIpEWSFi1btuzBRRwRESPNVFF0PWAbYDfgAODTkh43fJHtBbbn254/Z86cGWo6IiJgsoR+C7D5wPHc9tygpcBC2/fbvhH4fzQJPiIiCpkkoV8KbCNpK0nrA/sDC4eu+SrN3TmSNqUZgrlhBuOMiIgxxiZ02w8AhwLnAtcBp9u+RtIxkvZpLzsX+I2ka4HzgCNs/2ZtBR0REdONnbYIYPsc4Jyhc+8Z+N7AW9uviIjoQFaKRkRUIgk9IqISSegREZVIQo+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEokoUdEVCIJPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlUhCj4ioRBJ6REQlktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEpMlNAl7SHpeklLJB25muv+QpIlzZ+5ECMiYhJjE7qkdYHjgFcC2wIHSNp2xHUbAocBP5zpICMiYrxJ7tCfDSyxfYPt+4DTgH1HXPcPwLHA72YwvoiImNAkCX0z4OcDx0vbc8tJ2gnY3PbZMxhbRESsgYdcFJW0DvBx4PAJrj1E0iJJi5YtW/ZQm46IiAGTJPRbgM0Hjue256ZsCGwHfFfSTcBzgYWjCqO2F9ieb3v+nDlzHnzUERExzSQJ/VJgG0lbSVof2B9YOPWi7Tttb2p7nu15wMXAPrYXrZWIIyJipLEJ3fYDwKHAucB1wOm2r5F0jKR91naAERExmfUmucj2OcA5Q+fes4prd3voYUVExJrKStGIiEokoUdEVCIJPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlUhCj4ioRBJ6REQlktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEokoUdEVCIJPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKTJTQJe0h6XpJSyQdOeL1t0q6VtJVkr4tacuZDzUiIlZnbEKXtC5wHPBKYFvgAEnbDl12OTDf9vbAl4APz3SgERGxepPcoT8bWGL7Btv3AacB+w5eYPs82/e0hxcDc2c2zIiIGGeShL4Z8POB46XtuVU5GPjGqBckHSJpkaRFy5YtmzzKiIgYa0aLopIOBOYDHxn1uu0Ftufbnj9nzpyZbDoiYtZbb4JrbgE2Hzie255biaSXAe8CXmT7v2YmvIiImNQkd+iXAttI2krS+sD+wMLBCyQ9EzgB2Mf2r2c+zIiIGGdsQrf9AHAocC5wHXC67WskHSNpn/ayjwCPBc6QdIWkhav46yIiYi2ZZMgF2+cA5wyde8/A9y+b4bgiImINZaVoREQlktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqEQSekREJZLQIyIqkYQeEVGJJPSIiEokoUdEVCIJPSKiEknoERGVSEKPiKhEEnpERCWS0CMiKpGEHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlUhCj4ioRBJ6REQlktAjIiqRhB4RUYkk9IiISiShR0RUIgk9IqISSegREZVIQo+IqEQSekREJSZK6JL2kHS9pCWSjhzx+iMlfbF9/YeS5s10oBERsXpjE7qkdYHjgFcC2wIHSNp26LKDgdtt/xHwT8CxMx1oRESs3iR36M8Glti+wfZ9wGnAvkPX7Auc1H7/JeClkjRzYUZExDiyvfoLpFcDe9j+X+3xa4Hn2D504Jqr22uWtsc/ba+5bejvOgQ4pD18GnD9TP2HTGhT4LaxV5XRl1j6Egf0J5a+xAGJZZS+xAHdxLKl7TmjXlivZBS2FwALSrY5SNIi2/O7an9QX2LpSxzQn1j6Egcklj7HAf2KBSYbcrkF2HzgeG57buQ1ktYDNgZ+MxMBRkTEZCZJ6JcC20jaStL6wP7AwqFrFgKva79/NfAdjxvLiYiIGTV2yMX2A5IOBc4F1gVOtH2NpGOARbYXAp8FTpG0BPgtTdLvo86Ge0boSyx9iQP6E0tf4oDEMkpf4oB+xTK+KBoREQ8PWSkaEVGJJPSIiEokoUdEVCIJvSBJL5X06K7jiH5LP5lO0qO6juHhYFYURdv9aJ7IwKwe2z/rII6TgOfRzAS6ELgAuMj27R3EsivwPmBLmvdFgG0/pXAcFwHn07wf37N9d8n2h2LZBZjHyv3k5A7iSD+ZHscS4Fc078eFNO/HnSVjGIhlDvBGpveVN3QRz6DqE7qkvwHeS9MZft+etu3tO4zpyTTz9d8GPNl20RW7bQw/Bv4OWAz899R520UXhEnaCnhB+/Vc4L+AC23/XeE4TgG2Bq5gxfth228pGcdQTOknK8eyBU0/2RXYE7jD9o4dxPF9ml8qw+/Jl0vHMqx4B+nAYcDTuuiAwyQdSNMhn0Gz/8O/0HSMLtxp+xsdtb2c7Rsl/Q64r/16MfDHHYQyH9i2Dwvi0k+mkzSXJpG/ANgBuAa4qKNwNrD99o7aXq3ZcId+HrC77Qd6EMttwE+BTwHn2b6pgxh2ar/dj2ah2Jk0d8UA2L6scDw/pUlap9IkrSts/371P7VW4jgDeIvtW0u3PSKW9JPp8fyeZtX6B2x/rWTbI2J5P/B92+d0GccosyGhf5ZmZ8ezWblDfryjeP4EeCHwfGAb4Hrbry3Y/nmredm2X1IqFgBJh9G8F5sDP6YZT7/A9k8LtX8WYGBDYEfgElbuJ/uUiGNEXOknAyTtQPNevBDYAvgJcL7tzxaM4W6aviLgMTT95H5W1BU2KhXLqsyGIZeftV/rt1+dkbQRTWfckqagsjErxvWLsP3iku2NY/sTwCckPRY4iKYAN5fmrrCEjxZqZ2LpJ9PZvrL9NPdTmmGXA4EX0Ww7UiqGDUu19WBVf4feJ5Kuohn3u4jmLnRph7G8dcTpO4HFtq8oGMfHaO68Hgv8gHYWg+0bSsXQN+knI+NYBDwSmCpIXmj75lLtD8Wy04jTdwI3dz20W31Cb6cY/T3wJ8DyuaylPzIOxfTYNob/6DCGU2kKgWe1p14FXEVzR3iG7Q8XiuPVNP9z/qpEe6uJY+rj9KA7gUXA4V38gkk/WSmOObaXlWhrHEkXAzsBP2pPPQO4muaT1F/Z/lZXsc2GhUVfoBmb3Qo4GriJprhSnKTtJF1OU6G/VtJiSdt1EQvNsMZOtg+3fTiwM/AEmjHK15cKwvaXgOdI+mj7tXeptof8b+AIYDOa9+ZtNIXa04ATSwaSfjLSfZI+LmlR+/UxSRsXbH/QL4Bn2t7Z9s40tZcbgN2BIr/gVmU2JPQ/aAsn99s+v53839Xd+QLgrba3tL0FcDjdbb/5BAaKfzTFnSfavnfo/Fol6YM0U0uvbb/eIukDpdofsI/tE2zfbfuu9ular7D9RWCTwrGkn0x3InA3zayb/YC7gM8VbH/QU21fM3Vg+1rg6X0YJpwNRdH72z9vlbQXzW/Xx3cUy2NsL589YPu7kh7TUSxfAH4oaWoK2N7AqW081xaMYy9gx6mpiu0qycuBdxaMAeAeSfvRPOQcmgU9v2u/Lz0umX4y3da2/2Lg+GhJxcbwh1wj6XiaT28Af0nzSeqRrMg3nZgNY+ivoimibA58EtgIOLp9MEfpWL4CXAac0p46ENjZ9p+VjqWNZz7NYg1olt0v6iCGq4DdbP+2PX488N3SK3klPQX4BM2SewMX06yQvIXm36jYIpb0k5Ex/AA4Yurfod2S4KO2n9dBLI8G3kxTzAf4HvB/aG4ANui05lF7Qu8TSZvQjONPdYQLgfeV3KND0ka272oT5zRTibVgPAcAHwLOo5nP+0LgyHaoY1ZKPxkZz47ASTSFR9Hsc/N621eWjKPvqk/okp4KHE8z7redpO1pxkvf33FonZD0dduvknQjKw8ldLLpUhvTk4BntYeX2P5lwbb/3vaHJX2SEUMrXe7l0qU+9pM2ro1oArirg7ZPt72fpB8xuq90tj/UlNmQ0M+nmb1wgu1ntueutl1s1sDAasSRulqN2KVVzOVdrtTSckl72z5L0utGvW77pBJxtLGknwxZxTz45Uqu+Jb0JNu3StpyFbF0Mi9+0Gwoim5g+xJJg+dKT/7v3WpEAEmbsWJbVABsX1Co+Y+t5jVTaCaS7bPaP4sl7tVIP5muN6szp/b56UPiXpXZkNBvk7Q17Z1Pu5Cl6AZMts+f5DpJXx6q5K81ko6lrc4zsF0szd7ba92kS8sl7W7739d2PO3Q3NuYvsd1sSmu6SfT2T56kuskvcP2B9d2PG1bfw4cSzOlU/RoL5fZMOTyFJo5vLsAtwM3Age6gx3sxpF0+dSwUIG2rge2t11yLvEak3SZ7dUOz8xQO1fS7G44vMf14rXd9ppKP5muVD9p21oC7G37uhLtrYnq79Dbyf4va+fNruMOn4gzgZK/XW8AHkHZxSEPhsZfMiMesH18obYeqvST6Ur1E4Bf9TGZQ8UJfVXFlKmx9JLFlD4ZmM1xD3CFpG+z8naxfZvVsVaT18C0vLMkvRn4Ciu/H0Wn5/VF+sl07VALwCJJXwS+ysrvyZlrO4Zxqk3orCimTO1fPKiv40wl7jKmFoUsBoovruqhxazcR44YeM1AJ9Pzxkg/ma7EezK4z9A9wMsHjk3zEJBu2a76i2YxwuMGjjcBTuw6rlXE+vKuYxiI5ctdx9DGcWbXMbRx7N51DAOxzLp+Auy6unPAO7t+LwZieUdXbc+Goui0AlLJotJQu714gvokSr1Hkl4DfNP23ZKOotmW9P0u/IizcQoX3dJPprcz7f0v+W+yJrqMq+YhlynrSNrE7bLpdsy0q//uzzLiCeo9Veo3/bttnyHp+cDLgI/QrOx9TqH2J1Wy6JZ+0pL0PJoZanOG6mIbUe6pVmuqZF9ZyWxI6B8DfqDmIcAArwH+saNYevEE9Z6ZSlh7AQtsn63mIbx9U/KjbPrJCuvTPM1qPVZeZHQXzY6YfdTZsEf1Qy4AkrZlxcrD77jZv7iLOD5ED56gPomCH6W/TrOj4e40wy330uznssPabntNFB5yST+Z3s6Wtm9WD57iNE5XQ7owO+7QaRN4J0l8yNQwwvyBc8WWua+htxdqZz9gD5qtUO9oN+o6YszPdOGmgm2ln0y3YfsUp8cDSLoNeJ3tqwu1v5ykXW1/bzXnzhjxY0XMijv0mK4vhTdJW4w6b/tnheN4WBRnS+tRP/k+8C63D/6QtBvwAdu7lIyjbbu3BdpZcYfeF2qegfhemj2/Ac4HjrF9Zwfh9KXwdjYr5oE/iubZr9fTPNS7pF4VZ9U8XWv4webHdBBKX/pJ509xejgUaJPQyzqR5ung+7XHr6V5LuKfr/In1p5eFN5sP2PwuN1W980dhNKb4qykTwEbAC8GPkNT/Luki1joST8BbpD0blZ+ilPpZ3j2vkCbIZeCJF1he8dx5wrF0tvCm6QfDSf6Am32pjgr6Srb2w/8+VjgG7Zf0EEsvegnfXiK00AsvS3Q5g69rHslPd8rPxfx3o5i6UXhbeij6zo0yfQXJWNo9ak4O9Un7pH0ZOA3wJM6iqUX/aRN3H3ZP6Y3BdphSehl/RVwUjuWvvy5iF0E4gn3Iy9g8KPrAzRj6l/uII5NafcvGSjU/riDOAC+LulxNOP4l9Ek0M90EUhf+kkf9qsfsAB461CBdmqL7k5lyKUDXT4XcSCGPhVoOzfwnMiVirO2Sxdnh+N6JPCoLv9d+lCg7dN+9ZKuHB6KG3WuC7lDL0DSgbb/dXhL34638u1FgVbSvwOvsX1He7wJcJrtV5SMow/FWUkvsf2dgW1aB1/DHWzP2qMCbZ/2q+9DgXakJPQypqZXjXo+Ylcfkbb2yo8xO1rSFR3EMWcqmUMzVirpCR3EsRLbl0kqPWXxRcB3WHmb1uUh0c32rLsMFGiPlvQxoNisl57uV/8GmgLt1L/Hhe25ziWhF2D7hPbb/ztqhVkHIUF/CrT/LWmLqYVEap6oXvyXXB+Ks7bf2/55UMl2x+i6QDtqv/rB/lF8B8qeFWhXkoRe1idpEsW4cyX0pUD7LuAiSee3cbwAOKSDOPpSnEXSYTTDX3cDn6bpH0fa/lYH4XRaoLW9FYCk/WhW8t7VDnfsBPxDqTgG9axAu5IURQsYWGH2t8A/Dby0EfBnXRZTelKg3RR4bnt4se3buoqlD6YKbJJeAbwJOAo4peul5V0WaAfm5D+fJpF/FHiP7eIreftUoB2WO/QyerPCrKcF2l1YMdsG4OulA+hLcXYqnPbPPYGTbV+jqX+gUgH0r0A7uJL30x1vs9ynAu1KktALsH0+cL6kz/dghVmvCrTtSsRnAV9oTx0maRfb7ywcSp+Ks4slfYtm6uQ7JG0I/L5wDH0r0N4i6QSalbzHtp8W1ikZQE8LtCvJkEtBkrajmeo01TH6vAVoqTiuAna0/fv2eF3gctvbF45jMc3w12Bx9itdDHNIWgfYEbihXbX6eGCu7atKx9IXkjagWcn7I9s/aVfyPqNkXUHSjUx/6PzyBFp6B8pRiv6Gi+UrzLa0vSVweHuuC5+c8FwJjxv4fuOOYpgqzp4i6V+BC4B3dBTL82gWNd0h6UCaMfROFhZJOkzSRmp8RtJlkl4+/idnlu17bJ9p+yft8a2li8S2t2qT9tuBHdqC7eeAK+nJ5lwZcikrW4BO90Hgcknn0dz5vBA4snQQtr/ZLiaaKs7+bYfF2eOBHSTtQPNL/zPAyTTDIKW9wfYn2gLtH9AsQDsF6GLGTV8cZfv0tkD7EpoCbS+eg5s79LJukPRuSfPar6PofgvQqa+iBdqB+fdn0iTRM2mmCT7P9hdLxTFkF2C39uu5q71y7XrAzVjovsC/2D6O0TWPEqYVaKG7hyD3xLQCLc3/V53LGHpBWrEF6FQym9oC9I5V/9Rai6XTLUAlLba9s3rypJcRxdkDgEs7KM7Szsn/JnAQzSeWXwNXlt5SuI3lc8BmNAXaHWg+xX3X9s6lY+kL9Wir5WFJ6AVJmk8zVjuPFcNdLl0AbGPptEAr6WLgKuBPgdOGX7dddCVeX4qzbdt/CPwPml8oF6rZ/XE32yd3EEsKtEP6UKBdlST0giRdT7PC7GoGpqHZvrmDWDp9RmO7mOhlwLHAe4Zft31SiTgG4rmKJmn+tj1+PM2daPGE3ift0NgVtv+zLdDuBHyiiz4b46UoWtYy22d1HUSr0wJtW3A8TdJ1tq8s1e5q9KI4CyDpblZMh1sfeATwH7a7mAHUpwJtjJGEXtZ7JX0G+DYrL0joYhe9vmwBeq+kbwNPtL2dpO2BfWwXWQU4MPf+TOC7NOPoAG+3/csSMQyzvbwA2q4Q3ZfuirQP2LakqQLtZyUd3FEsMUaGXApq5zc/HbiGFUMutl18682+FGjbAuARwAm2n9meu9r2doXa71VxdlUkXT71/hRutzcF2hgvd+hlPcv207oOorU1sDnN1NX1gJfSzKktPWa8ge1LhrYqeaBg+/dLWgDMlfTPwy+WLs4CDO2fsg7N8zx/VzqO1l/SFGgPtv3LtkD7kY5iiTGS0Mv6vqRtbV/bdSA00/OmFWg7cJukrWnHjCW9Gri1YPuvoinOvoJm97w+GNw/5QHgJpphl+LaYaePDxz/jGYMPXooQy4FSbqO5s74RpoxdNHdtMWLbD+/dLsj4ngKKx6wezvNe/M/S8+ikLRDT4qzvdKzAm2MkYReULvh0zQdTVt8Kc3imU4KtMPb9wKPphle+M82jqLb+Kp5aMHxdFScHYrlUcDBTH8wc6ePORss0NruZAZQrF6W/hdk++ZRXx2FcxDNgpE9aD7i700z/FDK1JYD82menrQJzSZdb6KbJzh9mmYzrvsB2oUz+3cQBzQzj/6QZhjofGAuzdOLOuXGV2niih7KGPrs1WmB1vbRAJIuAHayfXd7/D6ax7+V1nVxdtAf2X6NpH1tnyTpVJpZSMX1rEAbYyShz159KdA+Ebhv4Pi+9lxpXRdnB93f/nlHu0XDL4GuHrbRmwJtjJeEPns9F7ii3bS/ywLtycAlkr7SHv8p8PnCMQD8NU1x9umSbqEtznYQB8CCdp3AUcBCmt0x391FILYP6qLdeHBSFJ2lelag3Ql4QXt4ge3LC7bdq+JsG9PhrJhZMjUGdAew2PYVhWPpZYE2Rssd+izVp82VbF8GXNZR81PL7J9Gs+z/azRJ9LXAJR3FtDPNWPXUvj+votmZ8k2SzrD94YKxnAL8mKYQegzNp5brCrYfayB36BEsL87uNVCc3RA42/YLO4plz6k96ts968+mmZG02Pa2BWO53PYzJV1le3tJjwAutN3lA0BiFTJtMaLRl+IsNAXQ/xo4vp9mfvy9Q+dLGC7Qbkx3BdoYI0MuEY2+FGeh2Zbhh5K+1h7vDZzabm9celZSbwq0MV6GXCJaXRZnR8QynxU7YX7P9qKO4uhNgTbGS0KPiFVqFzWNKtDOA0oXaGOMJPSIWKU+FWhjvBRFI2J1+lSgjTFSFI2I1elTgTbGyJBLRKxWXwq0MV4SekREJTKGHhFRiST0iIhKJKFHRFQiCT0iohJJ6BERlfj/K4ORoxIYjTkAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "cellView": "form",
        "id": "QAtIS9R3nnCb"
      },
      "source": [
        "#@title Load fresh VGG19 model\n",
        "#! git clone 'https://github.com/huyvnphan/PyTorch_CIFAR10.git' \n",
        "\n",
        "#! python ./PyTorch_CIFAR10/train.py --download_weights 1\n",
        "\n",
        "from PyTorch_CIFAR10.cifar10_models.vgg import vgg11_bn, vgg13_bn, vgg16_bn, vgg19_bn\n",
        "\n",
        "# Untrained model\n",
        "model = vgg19_bn()\n",
        "\n",
        "# Pretrained model\n",
        "model = vgg19_bn(pretrained=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 684
        },
        "id": "kglkZo-J7L2H",
        "cellView": "form",
        "outputId": "a8e08d8a-4873-4fb8-894b-c2ce027eaafd"
      },
      "source": [
        "#@title Run first finetuning and testing - only motion blur model \n",
        "\n",
        "# path = 'drive/MyDrive/NMA_DL_project/model-fold-2.pth'\n",
        "# model = CNOTP()\n",
        "# model.load_state_dict(torch.load(path))\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "dataset_list_train = [motion_low_train, motion_high_train]\n",
        "dataset_list_test = [motion_low_test, motion_high_test]\n",
        "\n",
        "model_save_name = 'motion_blur_all_severity_model'\n",
        "\n",
        "finetuning (model, optimizer, criterion, dataset_list_train, dataset_list_test,model_save_name, batch_size=100, num_epochs =50)\n",
        "\n",
        "fig_savepath = '/content/drive/MyDrive/NMA_DL_project/'+ model_save_name+ '_test_acc.png'\n",
        "\n",
        "test_accuracies = test_function(model,fig_savepath,res_savepath,\n",
        "                               clean = testset, \n",
        "                               motion_low = motion_low_test, \n",
        "                               motion_high = motion_high_test, \n",
        "                               defocus_low = defocus_low_test,\n",
        "                               defocus_high = defocus_high_test, \n",
        "                               gaussian_low = gaussian_low_test,\n",
        "                               gaussian_high = gaussian_high_test,\n",
        "                               shot_low = shot_low_test,\n",
        "                               shot_high = shot_high_test)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "20\n",
            "[1] epoch. Loss: 83.620. Training accuracy: 0.800 . Test accuracy: 0.865\n",
            "[2] epoch. Loss: 46.093. Training accuracy: 0.970 . Test accuracy: 0.887\n",
            "[3] epoch. Loss: 30.342. Training accuracy: 0.970 . Test accuracy: 0.892\n",
            "[4] epoch. Loss: 24.673. Training accuracy: 0.920 . Test accuracy: 0.897\n",
            "[5] epoch. Loss: 20.371. Training accuracy: 0.970 . Test accuracy: 0.897\n",
            "[6] epoch. Loss: 16.507. Training accuracy: 0.970 . Test accuracy: 0.910\n",
            "[7] epoch. Loss: 14.165. Training accuracy: 0.970 . Test accuracy: 0.907\n",
            "[8] epoch. Loss: 11.460. Training accuracy: 0.980 . Test accuracy: 0.913\n",
            "[9] epoch. Loss: 12.806. Training accuracy: 0.990 . Test accuracy: 0.906\n",
            "[10] epoch. Loss: 8.321. Training accuracy: 0.990 . Test accuracy: 0.919\n",
            "[11] epoch. Loss: 9.662. Training accuracy: 0.980 . Test accuracy: 0.890\n",
            "[12] epoch. Loss: 9.860. Training accuracy: 1.000 . Test accuracy: 0.908\n",
            "[13] epoch. Loss: 10.687. Training accuracy: 1.000 . Test accuracy: 0.917\n",
            "[14] epoch. Loss: 8.745. Training accuracy: 0.990 . Test accuracy: 0.908\n",
            "[15] epoch. Loss: 6.137. Training accuracy: 1.000 . Test accuracy: 0.926\n",
            "[16] epoch. Loss: 6.528. Training accuracy: 0.990 . Test accuracy: 0.912\n",
            "[17] epoch. Loss: 7.157. Training accuracy: 0.980 . Test accuracy: 0.929\n",
            "[18] epoch. Loss: 8.116. Training accuracy: 0.980 . Test accuracy: 0.929\n",
            "[19] epoch. Loss: 8.618. Training accuracy: 0.990 . Test accuracy: 0.919\n",
            "[20] epoch. Loss: 5.983. Training accuracy: 1.000 . Test accuracy: 0.922\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "cellView": "form",
        "id": "w0Ky_se1p2AM"
      },
      "source": [
        "#@title Load fresh VGG19 model\n",
        "#! git clone 'https://github.com/huyvnphan/PyTorch_CIFAR10.git' \n",
        "\n",
        "#! python ./PyTorch_CIFAR10/train.py --download_weights 1\n",
        "\n",
        "from PyTorch_CIFAR10.cifar10_models.vgg import vgg11_bn, vgg13_bn, vgg16_bn, vgg19_bn\n",
        "\n",
        "# Untrained model\n",
        "model = vgg19_bn()\n",
        "\n",
        "# Pretrained model\n",
        "model = vgg19_bn(pretrained=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C9rYke6L-fBX",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 684
        },
        "cellView": "form",
        "outputId": "aaa30192-7407-4b14-8d8d-773b4e2a2ffb"
      },
      "source": [
        "#@title Run first finetuning and testing - only high and low gaussian noise model\n",
        "\n",
        "# path = 'drive/MyDrive/NMA_DL_project/model-fold-2.pth'\n",
        "# model = CNOTP()\n",
        "# model.load_state_dict(torch.load(path))\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "dataset_list_train = [gaussian_low_train, gaussian_high_train]\n",
        "dataset_list_test = [gaussian_low_test, gaussian_high_test]\n",
        "\n",
        "model_save_name = 'gaussian_blur_all_severity_model'\n",
        "\n",
        "finetuning (model, optimizer, criterion, dataset_list_train, dataset_list_test,model_save_name, batch_size=100, num_epochs =50)\n",
        "\n",
        "fig_savepath = '/content/drive/MyDrive/NMA_DL_project/'+ model_save_name+ '_test_acc.png'\n",
        "\n",
        "test_accuracies = test_function(model,fig_savepath,res_savepath,\n",
        "                               clean = testset, \n",
        "                               motion_low = motion_low_test, \n",
        "                               motion_high = motion_high_test, \n",
        "                               defocus_low = defocus_low_test,\n",
        "                               defocus_high = defocus_high_test, \n",
        "                               gaussian_low = gaussian_low_test,\n",
        "                               gaussian_high = gaussian_high_test,\n",
        "                               shot_low = shot_low_test,\n",
        "                               shot_high = shot_high_test)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "20\n",
            "[1] epoch. Loss: 110.258. Training accuracy: 0.840 . Test accuracy: 0.805\n",
            "[2] epoch. Loss: 57.624. Training accuracy: 0.900 . Test accuracy: 0.846\n",
            "[3] epoch. Loss: 38.721. Training accuracy: 0.920 . Test accuracy: 0.839\n",
            "[4] epoch. Loss: 27.616. Training accuracy: 0.970 . Test accuracy: 0.864\n",
            "[5] epoch. Loss: 22.612. Training accuracy: 0.920 . Test accuracy: 0.839\n",
            "[6] epoch. Loss: 20.620. Training accuracy: 0.940 . Test accuracy: 0.894\n",
            "[7] epoch. Loss: 15.660. Training accuracy: 0.980 . Test accuracy: 0.881\n",
            "[8] epoch. Loss: 14.898. Training accuracy: 0.980 . Test accuracy: 0.893\n",
            "[9] epoch. Loss: 10.859. Training accuracy: 1.000 . Test accuracy: 0.875\n",
            "[10] epoch. Loss: 11.361. Training accuracy: 0.980 . Test accuracy: 0.887\n",
            "[11] epoch. Loss: 9.704. Training accuracy: 0.980 . Test accuracy: 0.882\n",
            "[12] epoch. Loss: 10.025. Training accuracy: 0.970 . Test accuracy: 0.868\n",
            "[13] epoch. Loss: 9.217. Training accuracy: 0.980 . Test accuracy: 0.870\n",
            "[14] epoch. Loss: 8.510. Training accuracy: 0.980 . Test accuracy: 0.902\n",
            "[15] epoch. Loss: 6.197. Training accuracy: 1.000 . Test accuracy: 0.884\n",
            "[16] epoch. Loss: 7.350. Training accuracy: 0.980 . Test accuracy: 0.891\n",
            "[17] epoch. Loss: 8.292. Training accuracy: 1.000 . Test accuracy: 0.887\n",
            "[18] epoch. Loss: 9.037. Training accuracy: 1.000 . Test accuracy: 0.887\n",
            "[19] epoch. Loss: 6.495. Training accuracy: 0.990 . Test accuracy: 0.884\n",
            "[20] epoch. Loss: 8.384. Training accuracy: 0.990 . Test accuracy: 0.895\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "cellView": "form",
        "id": "OAxnWvIKqBHe"
      },
      "source": [
        "#@title Load fresh VGG19 model\n",
        "#! git clone 'https://github.com/huyvnphan/PyTorch_CIFAR10.git' \n",
        "\n",
        "#! python ./PyTorch_CIFAR10/train.py --download_weights 1\n",
        "\n",
        "from PyTorch_CIFAR10.cifar10_models.vgg import vgg11_bn, vgg13_bn, vgg16_bn, vgg19_bn\n",
        "\n",
        "# Untrained model\n",
        "model = vgg19_bn()\n",
        "\n",
        "# Pretrained model\n",
        "model = vgg19_bn(pretrained=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KBcxLGPG-qh6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 684
        },
        "cellView": "form",
        "outputId": "b83f9588-6cc6-497f-89ed-39a1baacf1c4"
      },
      "source": [
        "#@title Run first finetuning and testing - all noise model\n",
        "\n",
        "# path = 'drive/MyDrive/NMA_DL_project/model-fold-2.pth'\n",
        "# model = CNOTP()\n",
        "# model.load_state_dict(torch.load(path))\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "dataset_list_train = [gaussian_low_train, gaussian_high_train, shot_low_train, shot_high_train]\n",
        "dataset_list_test = [gaussian_low_test, gaussian_high_test, shot_low_test, shot_high_test]\n",
        "\n",
        "model_save_name = 'all_noise_all_severity_model'\n",
        "\n",
        "finetuning (model, optimizer, criterion, dataset_list_train, dataset_list_test,model_save_name, batch_size=100, num_epochs =50)\n",
        "\n",
        "fig_savepath = '/content/drive/MyDrive/NMA_DL_project/'+ model_save_name+ '_test_acc.png'\n",
        "\n",
        "test_accuracies = test_function(model,fig_savepath,res_savepath,\n",
        "                               clean = testset, \n",
        "                               motion_low = motion_low_test, \n",
        "                               motion_high = motion_high_test, \n",
        "                               defocus_low = defocus_low_test,\n",
        "                               defocus_high = defocus_high_test, \n",
        "                               gaussian_low = gaussian_low_test,\n",
        "                               gaussian_high = gaussian_high_test,\n",
        "                               shot_low = shot_low_test,\n",
        "                               shot_high = shot_high_test)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "20\n",
            "[1] epoch. Loss: 176.401. Training accuracy: 0.900 . Test accuracy: 0.865\n",
            "[2] epoch. Loss: 82.088. Training accuracy: 0.970 . Test accuracy: 0.918\n",
            "[3] epoch. Loss: 49.335. Training accuracy: 0.970 . Test accuracy: 0.934\n",
            "[4] epoch. Loss: 36.620. Training accuracy: 0.950 . Test accuracy: 0.939\n",
            "[5] epoch. Loss: 33.013. Training accuracy: 0.980 . Test accuracy: 0.944\n",
            "[6] epoch. Loss: 22.347. Training accuracy: 0.990 . Test accuracy: 0.953\n",
            "[7] epoch. Loss: 18.597. Training accuracy: 0.960 . Test accuracy: 0.950\n",
            "[8] epoch. Loss: 18.944. Training accuracy: 1.000 . Test accuracy: 0.961\n",
            "[9] epoch. Loss: 19.220. Training accuracy: 0.970 . Test accuracy: 0.959\n",
            "[10] epoch. Loss: 14.948. Training accuracy: 0.980 . Test accuracy: 0.968\n",
            "[11] epoch. Loss: 13.844. Training accuracy: 1.000 . Test accuracy: 0.948\n",
            "[12] epoch. Loss: 14.992. Training accuracy: 0.980 . Test accuracy: 0.957\n",
            "[13] epoch. Loss: 10.806. Training accuracy: 0.990 . Test accuracy: 0.962\n",
            "[14] epoch. Loss: 12.320. Training accuracy: 0.990 . Test accuracy: 0.965\n",
            "[15] epoch. Loss: 11.699. Training accuracy: 1.000 . Test accuracy: 0.953\n",
            "[16] epoch. Loss: 13.594. Training accuracy: 1.000 . Test accuracy: 0.967\n",
            "[17] epoch. Loss: 9.603. Training accuracy: 0.980 . Test accuracy: 0.961\n",
            "[18] epoch. Loss: 10.564. Training accuracy: 0.970 . Test accuracy: 0.970\n",
            "[19] epoch. Loss: 10.065. Training accuracy: 1.000 . Test accuracy: 0.970\n",
            "[20] epoch. Loss: 8.090. Training accuracy: 0.970 . Test accuracy: 0.969\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAE2CAYAAABx82k0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAeaElEQVR4nO3debglVX3u8e8LiIoCYmiN0kAjQQ1BQGgncMABRRDIoETuxUfRK48xRBKRiIooxKvilGsMQVBRIBIERW0FxVxFBhWhm0kGubYMCqI2ypSAAeJ7/6g63fvss/vsDZxeVazzfp7nPN27dnXXj93Fe+qsX61Vsk1ERDz0rdV1ARERMTcS6BERlUigR0RUIoEeEVGJBHpERCUS6BERlVinqwNvvPHGXrRoUVeHj4h4SFq2bNkttheMeq+zQF+0aBFLly7t6vAREQ9Jkm5Y3XsZcomIqEQCPSKiEgn0iIhKjA10ScdL+rWkK1bzviT9k6Tlki6XtMPclxkREeNMcoX+OWC3Wd5/ObBV+3UAcMyDLysiIu6vsYFu+1zgt7PssjdwohsXAI+R9IS5KjAiIiYzF2PomwA/H3h9Y7ttBkkHSFoqaemKFSvm4NARETGlaFPU9nG2F9tevGDByPviIyLiAZqLiUU3AZsOvF7YbltjFh16xpr862e4/oN7FD1eRMQDMReBvgQ4UNIpwLOA223fPAd/bzxE5RtuRDfGBrqkfwN2ATaWdCPwHuBhALY/CZwJ7A4sB+4C9l9TxUY8VOWb3Ez5TObe2EC3ve+Y9w389ZxVFBERD0hmikZEVCKBHhFRiQR6REQlOlsPvRZp7EREXyTQI+aZkhchD5ULkFouzDLkEhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVGKiQJe0m6RrJC2XdOiI9zeTdLakSyRdLmn3uS81IiJmMzbQJa0NHA28HNga2FfS1kO7HQacavvpwKuBf5nrQiMiYnaTXKE/E1hu+1rb9wCnAHsP7WNgg/b3GwK/mLsSIyJiEpME+ibAzwde39huG/ReYD9JNwJnAn8z6i+SdICkpZKWrlix4gGUGxERqzNXTdF9gc/ZXgjsDpwkacbfbfs424ttL16wYMEcHToiImCyQL8J2HTg9cJ226A3AKcC2P4B8Ahg47koMCIiJjNJoF8EbCVpC0nr0jQ9lwzt8zPgxQCS/pgm0DOmEhFR0NhAt30fcCBwFnA1zd0sV0o6UtJe7W4HA2+UdBnwb8DrbHtNFR0RETOtM8lOts+kaXYObjt84PdXATvPbWkRD96iQ88odqzrP7hHsWNFjJKZohERlUigR0RUIoEeEVGJBHpERCUS6BERlUigR0RUIoEeEVGJBHpERCUS6BERlUigR0RUIoEeEVGJBHpERCUS6BERlUigR0RUIoEeEVGJidZDj/4rue43ZO3viD7KFXpERCUS6BERlUigR0RUIoEeEVGJBHpERCUS6BERlUigR0RUIoEeEVGJBHpERCUS6BERlUigR0RUIoEeEVGJBHpERCUS6BERlUigR0RUIoEeEVGJiQJd0m6SrpG0XNKhq9lnH0lXSbpS0slzW2ZERIwz9olFktYGjgZ2BW4ELpK0xPZVA/tsBbwD2Nn2rZIet6YKjoiI0Sa5Qn8msNz2tbbvAU4B9h7a543A0bZvBbD967ktMyIixpkk0DcBfj7w+sZ226AnA0+W9D1JF0jaba4KjIiIyczVQ6LXAbYCdgEWAudKeprt2wZ3knQAcADAZpttNkeHjogImOwK/SZg04HXC9ttg24Elti+1/Z1wP+jCfhpbB9ne7HtxQsWLHigNUdExAiTBPpFwFaStpC0LvBqYMnQPl+huTpH0sY0QzDXzmGdERExxthAt30fcCBwFnA1cKrtKyUdKWmvdrezgN9Iugo4GzjE9m/WVNERETHTRGPots8EzhzadvjA7w28tf2KiIgOZKZoREQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlJgp0SbtJukbSckmHzrLfX0iypMVzV2JERExibKBLWhs4Gng5sDWwr6StR+y3PnAQ8MO5LjIiIsab5Ar9mcBy29favgc4Bdh7xH7/ABwF/G4O64uIiAlNEuibAD8feH1ju20lSTsAm9o+Yw5ri4iI++FBN0UlrQV8DDh4gn0PkLRU0tIVK1Y82ENHRMSASQL9JmDTgdcL221T1ge2Ab4r6Xrg2cCSUY1R28fZXmx78YIFCx541RERMcMkgX4RsJWkLSStC7waWDL1pu3bbW9se5HtRcAFwF62l66RiiMiYqSxgW77PuBA4CzgauBU21dKOlLSXmu6wIiImMw6k+xk+0zgzKFth69m310efFkREXF/ZaZoREQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlEugREZVIoEdEVCKBHhFRiQR6REQlJgp0SbtJukbSckmHjnj/rZKuknS5pG9L2nzuS42IiNmMDXRJawNHAy8Htgb2lbT10G6XAIttbwt8EfjQXBcaERGzm+QK/ZnActvX2r4HOAXYe3AH22fbvqt9eQGwcG7LjIiIcSYJ9E2Anw+8vrHdtjpvAL4x6g1JB0haKmnpihUrJq8yIiLGmtOmqKT9gMXAh0e9b/s424ttL16wYMFcHjoiYt5bZ4J9bgI2HXi9sN02jaSXAO8CXmD7v+amvIiImNQkV+gXAVtJ2kLSusCrgSWDO0h6OnAssJftX899mRERMc7YQLd9H3AgcBZwNXCq7SslHSlpr3a3DwOPBk6TdKmkJav56yIiYg2ZZMgF22cCZw5tO3zg9y+Z47oiIuJ+ykzRiIhKJNAjIiqRQI+IqEQCPSKiEgn0iIhKJNAjIiqRQI+IqEQCPSKiEgn0iIhKJNAjIiqRQI+IqEQCPSKiEgn0iIhKJNAjIiqRQI+IqEQCPSKiEgn0iIhKJNAjIiqRQI+IqEQCPSKiEgn0iIhKJNAjIiqRQI+IqEQCPSKiEgn0iIhKJNAjIiqRQI+IqEQCPSKiEgn0iIhKJNAjIiqRQI+IqEQCPSKiEgn0iIhKTBToknaTdI2k5ZIOHfH+wyV9oX3/h5IWzXWhERExu7GBLmlt4Gjg5cDWwL6Sth7a7Q3Arbb/CPhH4Ki5LjQiImY3yRX6M4Hltq+1fQ9wCrD30D57Aye0v/8i8GJJmrsyIyJiHNmefQfplcButv9X+/o1wLNsHziwzxXtPje2r3/a7nPL0N91AHBA+/IpwDVz9R8yoY2BW8buVUZfaulLHdCfWvpSB6SWUfpSB3RTy+a2F4x6Y52SVdg+Djiu5DEHSVpqe3FXxx/Ul1r6Ugf0p5a+1AGppc91QL9qgcmGXG4CNh14vbDdNnIfSesAGwK/mYsCIyJiMpME+kXAVpK2kLQu8GpgydA+S4DXtr9/JfAdjxvLiYiIOTV2yMX2fZIOBM4C1gaOt32lpCOBpbaXAJ8BTpK0HPgtTej3UWfDPSP0pZa+1AH9qaUvdUBqGaUvdUC/ahnfFI2IiIeGzBSNiKhEAj0iohIJ9IiISiTQC5L0YkmP7LqO6LecJzNJekTXNTwUzIumaLsezeMZuKvH9s86qOME4Dk0dwKdB5wLnG/71g5q2Rl4L7A5zeciwLafVLiO84FzaD6P79m+s+Txh2rZCVjE9PPkxA7qyHkys47lwK9oPo/zaD6P20vWMFDLAuCNzDxXXt9FPYOqD3RJfwO8h+Zk+H272ba37bCmJ9Lcr/824Im2i87YbWv4MfB3wDLgv6e22y46IUzSFsDz2q9nA/8FnGf77wrXcRKwJXApqz4P235LyTqGasp5Mr2WzWjOk52B3YHbbG/fQR3fp/mmMvyZfKl0LcOKnyAdOAh4Shcn4DBJ+9GckE+jWf/hn2lOjC7cbvsbHR17JdvXSfodcE/79ULgjzsoZTGwdR8mxOU8mUnSQpogfx6wHXAlcH5H5axn++0dHXtW8+EK/WxgV9v39aCWW4CfAp8EzrZ9fQc17ND+dh+aiWKn01wVA2D74sL1/JQmtE6mCa1Lbf9+9j+1Ruo4DXiL7ZtLH3tELTlPZtbze5pZ6++3/dWSxx5Ry/uA79s+s8s6RpkPgf4ZmpUdz2D6Cfmxjur5E+D5wHOBrYBrbL+m4PHPnuVt235RqVoAJB1E81lsCvyYZjz9XNs/LXT8rwEG1ge2By5k+nmyV4k6RtSV82SApO1oPovnA5sBPwHOsf2ZgjXcSXOuCHgUzXlyL6v6ChuUqmV15sOQy8/ar3Xbr85I2oDmZNycpqGyIavG9Yuw/cKSxxvH9seBj0t6NLA/TQNuIc1VYQkfKXScieU8mcn2Ze1Pcz+lGXbZD3gBzbIjpWpYv9SxHqjqr9D7RNLlNON+59Nchd7YYS1vHbH5dmCZ7UsL1vFRmiuvRwM/oL2Lwfa1pWrom5wnI+tYCjwcmGpInmf7hlLHH6plhxGbbwdu6Hpot/pAb28x+nvgT4CV97KW/pFxqKZHtzX8R4c1nEzTCPxau+kVwOU0V4Sn2f5QoTpeSfM/569KHG+WOqZ+nB50O7AUOLiLbzA5T6bVscD2ihLHGkfSBcAOwI/aTU8DrqD5SeqvbH+rq9rmw8Siz9OMzW4BHAFcT9NcKU7SNpIuoenQXyVpmaRtuqiFZlhjB9sH2z4Y2BF4HM0Y5etKFWH7i8CzJH2k/dqz1LGH/B/gEGATms/mbTSN2lOA40sWkvNkpHskfUzS0vbro5I2LHj8Qb8Anm57R9s70vRergV2BYp8g1ud+RDof9A2Tu61fU57839XV+fHAW+1vbntzYCD6W75zccx0Pyjae483vbdQ9vXKEkfoLm19Kr26y2S3l/q+AP2sn2s7Ttt39E+Xetltr8AbFS4lpwnMx0P3Elz180+wB3AZwsef9CTbV859cL2VcBT+zBMOB+aove2v94saQ+a766P7aiWR9leefeA7e9KelRHtXwe+KGkqVvA9gRObuu5qmAdewDbT92q2M6SvAR4Z8EaAO6StA/NQ86hmdDzu/b3pcclc57MtKXtvxh4fYSkYmP4Q66UdAzNT28Af0nzk9TDWZU3nZgPY+ivoGmibAp8AtgAOKJ9MEfpWr4MXAyc1G7aD9jR9p+VrqWtZzHNZA1opt0v7aCGy4FdbP+2ff1Y4LulZ/JKehLwcZop9wYuoJkheRPNv1GxSSw5T0bW8APgkKl/h3ZJgo/Yfk4HtTwSeDNNMx/ge8C/0FwArNdpz6P2QO8TSRvRjONPnQjnAe8tuUaHpA1s39EG5wxTwVqwnn2BDwJn09zP+3zg0HaoY17KeTKynu2BE2gaj6JZ5+Z1ti8rWUffVR/okp4MHEMz7reNpG1pxkvf13FpnZD0dduvkHQd04cSOll0qa3pCcAz2pcX2v5lwWP/ve0PSfoEI4ZWulzLpUt9PE/aujagKeCODo59qu19JP2I0edKZ+tDTZkPgX4Ozd0Lx9p+ervtCtvF7hoYmI04UlezEbu0mnt5Vyo1tVzSnra/Jum1o963fUKJOtpacp4MWc198CuVnPEt6Qm2b5a0+Wpq6eS++EHzoSm6nu0LJQ1uK33zf+9mIwJI2oRVy6ICYPvcQof/6CzvmUJ3Itn+WvtrseCeRc6TmXozO3NqnZ8+BPfqzIdAv0XSlrRXPu1ElqILMNk+Z5L9JH1pqJO/xkg6irY7z8BysTRrb69xk04tl7Sr7X9f0/W0Q3NvY+Ya18Vucc15MpPtIybZT9I7bH9gTdfTHuvPgaNobukUPVrLZT4MuTyJ5h7enYBbgeuA/dzBCnbjSLpkaliowLGuAba1XfJe4vtN0sW2Zx2emaPjXEazuuHwGtfL1vSx76+cJzOVOk/aYy0H9rR9dYnj3R/VX6G3N/u/pL1vdi13+EScCZT87not8DDKTg55IDR+lzlxn+1jCh3rwcp5MlOp8wTgV30Mc6g40FfXTJkaSy/ZTOmTgbs57gIulfRtpi8X27e7OtZoeA3clvc1SW8Gvsz0z6Po7Xl9kfNkpnaoBWCppC8AX2H6Z3L6mq5hnGoDnVXNlKn1iwf1dZypxFXG1KSQZUDxyVU9tIzp58ghA+8Z6OT2vDFynsxU4jMZXGfoLuClA69N8xCQbtmu+otmMsJjBl5vBBzfdV2rqfWlXdcwUMuXuq6hreP0rmto69i16xoGapl35wmw82zbgHd2/VkM1PKOro49H5qiMxpIJZtKQ8ftxRPUJ1HqM5L0KuCbtu+UdBjNsqTvc+FHnI1TuOmW82TmcWZ8/iX/Te6PLuuqechlylqSNnI7bbodM+3qv/szjHiCek+V+k7/btunSXou8BLgwzQze59V6PiTKtl0y3nSkvQcmjvUFgz1xTag3FOt7q+S58o08yHQPwr8QM1DgAFeBfzvjmrpxRPUe2YqsPYAjrN9hpqH8PZNyR9lc56ssi7N06zWYfokoztoVsTso86GPaofcgGQtDWrZh5+x836xV3U8UF68AT1SRT8UfrrNCsa7koz3HI3zXou263pY98fhYdccp7MPM7mtm9QD57iNE5XQ7owP67QaQO8kxAfMjWMsHhgW7Fp7vfT2wsdZx9gN5qlUG9rF+o6ZMyf6cL1BY+V82Sm9dunOD0WQNItwGttX1Ho+CtJ2tn292bZdtqIP1bEvLhCj5n60niTtNmo7bZ/VriOh0RztrQenSffB97l9sEfknYB3m97p5J1tMfubYN2Xlyh94WaZyC+h2bNb4BzgCNt395BOX1pvJ3BqvvAH0Hz7NdraB7qXVKvmrNqnq41/GDzIzsopS/nSedPcXooNGgT6GUdT/N08H3a16+heS7in6/2T6w5vWi82X7a4Ot2Wd03d1BKb5qzkj4JrAe8EPg0TfPvwi5qoSfnCXCtpHcz/SlOpZ/h2fsGbYZcCpJ0qe3tx20rVEtvG2+SfjQc9AWO2ZvmrKTLbW878OujgW/Yfl4HtfTiPOnDU5wGaultgzZX6GXdLem5nv5cxLs7qqUXjbehH13XognTX5SsodWn5uzUOXGXpCcCvwGe0FEtvThP2uDuy/oxvWnQDkugl/VXwAntWPrK5yJ2UYgnXI+8gMEfXe+jGVP/Ugd1bEy7fslAo/bHHdQB8HVJj6EZx7+YJkA/3UUhfTlP+rBe/YDjgLcONWinlujuVIZcOtDlcxEHauhTg7ZzA8+JnNactV26OTtc18OBR3T579KHBm2f1quXdNnwUNyobV3IFXoBkvaz/a/DS/p2vJRvLxq0kv4deJXt29rXGwGn2H5ZyTr60JyV9CLb3xlYpnXwPdzB8qw9atD2ab36PjRoR0qglzF1e9Wo5yN29SPSlp7+GLMjJF3aQR0LpsIcmrFSSY/roI5pbF8sqfQtiy8AvsP0ZVpXlkQ3y7PuNNCgPULSR4Fid730dL3619M0aKf+Pc5rt3UugV6A7WPb3/7fUTPMOigJ+tOg/W9Jm01NJFLzRPXi3+T60Jy1/Z721/1LHneMrhu0o9arHzw/iq9A2bMG7TQJ9LI+QRMU47aV0JcG7buA8yWd09bxPOCADuroS3MWSQfRDH/dCXyK5vw41Pa3Oiin0wat7S0AJO1DM5P3jna4YwfgH0rVMahnDdpp0hQtYGCG2d8C/zjw1gbAn3XZTOlJg3Zj4Nntywts39JVLX0w1WCT9DLgTcBhwEldTy3vskE7cE/+c2mC/CPA4baLz+TtU4N2WK7Qy+jNDLOeNmh3YtXdNgBfL11AX5qzU+W0v+4OnGj7Sk39A5UqoH8N2sGZvJ/qeJnlPjVop0mgF2D7HOAcSZ/rwQyzXjVo25mIzwA+3246SNJOtt9ZuJQ+NWeXSfoWza2T75C0PvD7wjX0rUF7k6RjaWbyHtX+tLBWyQJ62qCdJkMuBUnahuZWp6kTo89LgJaq43Jge9u/b1+vDVxie9vCdSyjGf4abM5+uYthDklrAdsD17azVh8LLLR9eela+kLSejQzeX9k+yftTN6nlewrSLqOmQ+dXxmgpVegHKXod7hYOcNsc9ubAwe327rwiQm3lfCYgd9v2FENU83ZkyT9K3Au8I6OankOzaSm2yTtRzOG3snEIkkHSdpAjU9LuljSS8f/ybll+y7bp9v+Sfv65tJNYttbtKH9dmC7tmH7WeAyerI4V4ZcysoSoDN9ALhE0tk0Vz7PBw4tXYTtb7aTiaaas3/bYXP2GGA7SdvRfNP/NHAizTBIaa+3/fG2QfsHNBPQTgK6uOOmLw6zfWrboH0RTYO2F8/BzRV6WddKerekRe3XYXS/BOjUV9EG7cD996fThOjpNLcJPsf2F0rVMWQnYJf269mz7rlm3edmLHRv4J9tH83onkcJMxq00N1DkHtiRoOW5v+rzmUMvSCtWgJ0KsymlgC9bfV/ao3V0ukSoJKW2d5RPXnSy4jm7L7ARR00Z2nvyf8msD/NTyy/Bi4rvaRwW8tngU1oGrTb0fwU913bO5aupS/Uo6WWhyXQC5K0mGasdhGrhrtcugHY1tJpg1bSBcDlwJ8Cpwy/b7voTLy+NGfbY/8h8D9ovqGcp2b1x11sn9hBLWnQDulDg3Z1EugFSbqGZobZFQzchmb7hg5q6fQZje1kopcARwGHD79v+4QSdQzUczlNaP62ff1YmivR4oHeJ+3Q2KW2/7Nt0O4AfLyLczbGS1O0rBW2v9Z1Ea1OG7Rtw/EUSVfbvqzUcWfRi+YsgKQ7WXU73LrAw4D/sN3FHUB9atDGGAn0st4j6dPAt5k+IaGLVfT6sgTo3ZK+DTze9jaStgX2sl1kFuDAvfenA9+lGUcHeLvtX5aoYZjtlQ3Qdobo3nTXpL3PtiVNNWg/I+kNHdUSY2TIpaD2/uanAleyasjFtosvvdmXBm3bADwEONb209ttV9jeptDxe9WcXR1Jl0x9PoWP25sGbYyXK/SynmH7KV0X0doS2JTm1tV1gBfT3FNbesx4PdsXDi1Vcl/B498r6ThgoaR/Gn6zdHMWYGj9lLVonuf5u9J1tP6SpkH7Btu/bBu0H+6olhgjgV7W9yVtbfuqrguhuT1vRoO2A7dI2pJ2zFjSK4GbCx7/FTTN2ZfRrJ7XB4Prp9wHXE8z7FJcO+z0sYHXP6MZQ48eypBLQZKuprkyvo5mDF10d9vi+bafW/q4I+p4EqsesHsrzWfzP0vfRSFpu540Z3ulZw3aGCOBXlC74NMMHd22+GKayTOdNGiHl+8FHkkzvPCfbR1Fl/FV89CCY+ioOTtUyyOANzDzwcydPuZssEFru5M7gGJ2mfpfkO0bRn11VM7+NBNGdqP5EX9PmuGHUqaWHFhM8/SkjWgW6XoT3TzB6VM0i3HdC9BOnHl1B3VAc+fRH9IMA50DLKR5elGn3PgKTV3RQxlDn786bdDaPgJA0rnADrbvbF+/l+bxb6V13Zwd9Ee2XyVpb9snSDqZ5i6k4nrWoI0xEujzV18atI8H7hl4fU+7rbSum7OD7m1/va1douGXQFcP2+hNgzbGS6DPX88GLm0X7e+yQXsicKGkL7ev/xT4XOEaAP6apjn7VEk30TZnO6gD4Lh2nsBhwBKa1THf3UUhtvfv4rjxwKQpOk/1rEG7A/C89uW5ti8peOxeNWfbmg5m1Z0lU2NAtwHLbF9auJZeNmhjtFyhz1N9WlzJ9sXAxR0dfmqa/VNopv1/lSZEXwNc2FFNO9KMVU+t+/MKmpUp3yTpNNsfKljLScCPaRqhR9L81HJ1wePH/ZAr9AhWNmf3GGjOrg+cYfv5HdWy+9Qa9e2a9WfQ3JG0zPbWBWu5xPbTJV1ue1tJDwPOs93lA0BiNXLbYkSjL81ZaBqg/zXw+l6a++PvHtpewnCDdkO6a9DGGBlyiWj0pTkLzbIMP5T01fb1nsDJ7fLGpe9K6k2DNsbLkEtEq8vm7IhaFrNqJczv2V7aUR29adDGeAn0iFitdlLTqAbtIqB0gzbGSKBHxGr1qUEb46UpGhGz6VODNsZIUzQiZtOnBm2MkSGXiJhVXxq0MV4CPSKiEhlDj4ioRAI9IqISCfSIiEok0CMiKpFAj4ioxP8H75aTgt7P2+cAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "cellView": "form",
        "id": "SaPUx1cxqCJd"
      },
      "source": [
        "#@title Load fresh VGG19 model\n",
        "#! git clone 'https://github.com/huyvnphan/PyTorch_CIFAR10.git' \n",
        "\n",
        "#! python ./PyTorch_CIFAR10/train.py --download_weights 1\n",
        "\n",
        "from PyTorch_CIFAR10.cifar10_models.vgg import vgg11_bn, vgg13_bn, vgg16_bn, vgg19_bn\n",
        "\n",
        "# Untrained model\n",
        "model = vgg19_bn()\n",
        "\n",
        "# Pretrained model\n",
        "model = vgg19_bn(pretrained=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "P6wgpP_w_ABh",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 684
        },
        "cellView": "form",
        "outputId": "26b05a36-f94a-4d79-dcd7-45da973f822a"
      },
      "source": [
        "#@title Run first finetuning and testing - all blur model\n",
        "\n",
        "# path = 'drive/MyDrive/NMA_DL_project/model-fold-2.pth'\n",
        "# model = CNOTP()\n",
        "# model.load_state_dict(torch.load(path))\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "dataset_list_train = [motion_low_train,motion_high_train, defocus_low_train, defocus_high_train]\n",
        "dataset_list_test = [motion_low_test, motion_high_test, defocus_low_test, defocus_high_test]\n",
        "\n",
        "model_save_name = 'all_blur_all_severity_model'\n",
        "\n",
        "finetuning (model, optimizer, criterion, dataset_list_train, dataset_list_test,model_save_name, batch_size=100, num_epochs =50)\n",
        "\n",
        "fig_savepath = '/content/drive/MyDrive/NMA_DL_project/'+ model_save_name+ '_test_acc.png'\n",
        "\n",
        "test_accuracies = test_function(model,fig_savepath,res_savepath,\n",
        "                               clean = testset, \n",
        "                               motion_low = motion_low_test, \n",
        "                               motion_high = motion_high_test, \n",
        "                               defocus_low = defocus_low_test,\n",
        "                               defocus_high = defocus_high_test, \n",
        "                               gaussian_low = gaussian_low_test,\n",
        "                               gaussian_high = gaussian_high_test,\n",
        "                               shot_low = shot_low_test,\n",
        "                               shot_high = shot_high_test)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "20\n",
            "[1] epoch. Loss: 145.250. Training accuracy: 0.900 . Test accuracy: 0.886\n",
            "[2] epoch. Loss: 69.254. Training accuracy: 0.960 . Test accuracy: 0.927\n",
            "[3] epoch. Loss: 45.113. Training accuracy: 0.960 . Test accuracy: 0.945\n",
            "[4] epoch. Loss: 35.376. Training accuracy: 0.970 . Test accuracy: 0.944\n",
            "[5] epoch. Loss: 29.234. Training accuracy: 0.970 . Test accuracy: 0.951\n",
            "[6] epoch. Loss: 21.126. Training accuracy: 0.950 . Test accuracy: 0.945\n",
            "[7] epoch. Loss: 20.853. Training accuracy: 0.970 . Test accuracy: 0.956\n",
            "[8] epoch. Loss: 18.724. Training accuracy: 0.990 . Test accuracy: 0.958\n",
            "[9] epoch. Loss: 19.704. Training accuracy: 0.980 . Test accuracy: 0.958\n",
            "[10] epoch. Loss: 14.183. Training accuracy: 0.980 . Test accuracy: 0.974\n",
            "[11] epoch. Loss: 12.024. Training accuracy: 1.000 . Test accuracy: 0.960\n",
            "[12] epoch. Loss: 14.972. Training accuracy: 0.980 . Test accuracy: 0.956\n",
            "[13] epoch. Loss: 13.132. Training accuracy: 1.000 . Test accuracy: 0.966\n",
            "[14] epoch. Loss: 14.855. Training accuracy: 0.980 . Test accuracy: 0.971\n",
            "[15] epoch. Loss: 11.906. Training accuracy: 1.000 . Test accuracy: 0.960\n",
            "[16] epoch. Loss: 9.511. Training accuracy: 1.000 . Test accuracy: 0.969\n",
            "[17] epoch. Loss: 12.712. Training accuracy: 0.990 . Test accuracy: 0.965\n",
            "[18] epoch. Loss: 9.514. Training accuracy: 1.000 . Test accuracy: 0.971\n",
            "[19] epoch. Loss: 9.491. Training accuracy: 1.000 . Test accuracy: 0.960\n",
            "Early stepped at [20] epoch. Loss: 8.787. Training accuracy: 1.000 . Test accuracy: 0.959\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "z7V6oc4WqDBW"
      },
      "source": [
        "#@title Load fresh VGG19 model without pretraining\n",
        "#! git clone 'https://github.com/huyvnphan/PyTorch_CIFAR10.git' \n",
        "\n",
        "#! python ./PyTorch_CIFAR10/train.py --download_weights 1\n",
        "\n",
        "from PyTorch_CIFAR10.cifar10_models.vgg import vgg11_bn, vgg13_bn, vgg16_bn, vgg19_bn\n",
        "\n",
        "# Untrained model\n",
        "model = vgg19_bn()\n",
        "\n",
        "# Pretrained model\n",
        "model = vgg19_bn(pretrained=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VHJ-45uf_RUI",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "9af0ce32-324e-4950-af49-e11e85f2aedc"
      },
      "source": [
        "#@title Run first finetuning and testing - impaired vision model \n",
        "\n",
        "# path = 'drive/MyDrive/NMA_DL_project/model-fold-2.pth'\n",
        "# model = CNOTP()\n",
        "optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "dataset_list_train = [motion_low_train,motion_high_train, defocus_low_train, defocus_high_train, gaussian_low_train, gaussian_high_train, shot_low_train, shot_high_train]\n",
        "dataset_list_test = [motion_low_test, motion_high_test, defocus_low_test, defocus_high_test,gaussian_low_test, gaussian_high_test, shot_low_test, shot_high_test]\n",
        "\n",
        "model_save_name = 'impaired_vision_model'\n",
        "\n",
        "finetuning (model, optimizer, criterion, dataset_list_train, dataset_list_test,model_save_name, batch_size=100, num_epochs =200)\n",
        "\n",
        "fig_savepath = '/content/drive/MyDrive/NMA_DL_project/'+ model_save_name+ '_test_acc.png'\n",
        "\n",
        "test_accuracies = test_function(model,fig_savepath,res_savepath,\n",
        "                               clean = testset, \n",
        "                               motion_low = motion_low_test, \n",
        "                               motion_high = motion_high_test, \n",
        "                               defocus_low = defocus_low_test,\n",
        "                               defocus_high = defocus_high_test, \n",
        "                               gaussian_low = gaussian_low_test,\n",
        "                               gaussian_high = gaussian_high_test,\n",
        "                               shot_low = shot_low_test,\n",
        "                               shot_high = shot_high_test)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "200\n",
            "[1] epoch. Loss: 838.172. Training accuracy: 0.710 . Test accuracy: 0.724\n",
            "[2] epoch. Loss: 308.467. Training accuracy: 0.890 . Test accuracy: 0.831\n",
            "[3] epoch. Loss: 119.021. Training accuracy: 0.950 . Test accuracy: 0.935\n",
            "[4] epoch. Loss: 71.637. Training accuracy: 0.980 . Test accuracy: 0.943\n",
            "[5] epoch. Loss: 53.700. Training accuracy: 0.980 . Test accuracy: 0.955\n",
            "[6] epoch. Loss: 46.762. Training accuracy: 0.970 . Test accuracy: 0.964\n",
            "[7] epoch. Loss: 41.226. Training accuracy: 0.990 . Test accuracy: 0.968\n",
            "[8] epoch. Loss: 36.461. Training accuracy: 0.980 . Test accuracy: 0.963\n",
            "[9] epoch. Loss: 35.485. Training accuracy: 0.990 . Test accuracy: 0.956\n",
            "[10] epoch. Loss: 30.585. Training accuracy: 0.990 . Test accuracy: 0.973\n",
            "[11] epoch. Loss: 30.994. Training accuracy: 0.960 . Test accuracy: 0.978\n",
            "[12] epoch. Loss: 27.877. Training accuracy: 1.000 . Test accuracy: 0.975\n",
            "[13] epoch. Loss: 27.711. Training accuracy: 1.000 . Test accuracy: 0.964\n",
            "[14] epoch. Loss: 24.063. Training accuracy: 0.990 . Test accuracy: 0.979\n",
            "[15] epoch. Loss: 24.101. Training accuracy: 0.990 . Test accuracy: 0.968\n",
            "[16] epoch. Loss: 24.009. Training accuracy: 0.990 . Test accuracy: 0.979\n",
            "[17] epoch. Loss: 20.324. Training accuracy: 1.000 . Test accuracy: 0.981\n",
            "[18] epoch. Loss: 18.718. Training accuracy: 0.990 . Test accuracy: 0.981\n",
            "[19] epoch. Loss: 20.160. Training accuracy: 1.000 . Test accuracy: 0.978\n",
            "[20] epoch. Loss: 17.485. Training accuracy: 0.990 . Test accuracy: 0.982\n",
            "[21] epoch. Loss: 18.816. Training accuracy: 0.990 . Test accuracy: 0.968\n",
            "[22] epoch. Loss: 17.886. Training accuracy: 0.990 . Test accuracy: 0.982\n",
            "[23] epoch. Loss: 15.698. Training accuracy: 1.000 . Test accuracy: 0.982\n",
            "[24] epoch. Loss: 17.349. Training accuracy: 0.990 . Test accuracy: 0.971\n",
            "[25] epoch. Loss: 14.899. Training accuracy: 1.000 . Test accuracy: 0.982\n",
            "[26] epoch. Loss: 14.583. Training accuracy: 1.000 . Test accuracy: 0.984\n",
            "[27] epoch. Loss: 15.354. Training accuracy: 1.000 . Test accuracy: 0.983\n",
            "[28] epoch. Loss: 14.788. Training accuracy: 0.990 . Test accuracy: 0.983\n",
            "[29] epoch. Loss: 13.699. Training accuracy: 0.990 . Test accuracy: 0.986\n",
            "[30] epoch. Loss: 13.317. Training accuracy: 1.000 . Test accuracy: 0.983\n",
            "[31] epoch. Loss: 13.251. Training accuracy: 1.000 . Test accuracy: 0.985\n",
            "[32] epoch. Loss: 11.604. Training accuracy: 1.000 . Test accuracy: 0.985\n",
            "[33] epoch. Loss: 13.821. Training accuracy: 0.970 . Test accuracy: 0.984\n",
            "[34] epoch. Loss: 12.267. Training accuracy: 0.960 . Test accuracy: 0.974\n",
            "[35] epoch. Loss: 12.834. Training accuracy: 1.000 . Test accuracy: 0.991\n",
            "[36] epoch. Loss: 9.125. Training accuracy: 0.990 . Test accuracy: 0.989\n",
            "[37] epoch. Loss: 12.292. Training accuracy: 1.000 . Test accuracy: 0.986\n",
            "[38] epoch. Loss: 10.859. Training accuracy: 0.990 . Test accuracy: 0.988\n",
            "[39] epoch. Loss: 10.594. Training accuracy: 1.000 . Test accuracy: 0.988\n",
            "[40] epoch. Loss: 9.783. Training accuracy: 0.990 . Test accuracy: 0.988\n",
            "[41] epoch. Loss: 9.578. Training accuracy: 1.000 . Test accuracy: 0.989\n",
            "[42] epoch. Loss: 10.623. Training accuracy: 1.000 . Test accuracy: 0.988\n",
            "[43] epoch. Loss: 9.784. Training accuracy: 0.990 . Test accuracy: 0.986\n",
            "[44] epoch. Loss: 10.884. Training accuracy: 1.000 . Test accuracy: 0.990\n",
            "Early stepped at [45] epoch. Loss: 8.049. Training accuracy: 1.000 . Test accuracy: 0.985\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}