Natural_image_Classification_Pipeline.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/JNPauli/Emotional_Machines/blob/main/Natural_image_Classification_Pipeline.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "f3485813"
      },
      "source": [
        "0. This notebook is the continuation to the preprocessing.ipynb notebook."
      ],
      "id": "f3485813"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "b60fcc27"
      },
      "source": [
        "1. Import all neccessary modules"
      ],
      "id": "b60fcc27"
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "73af1d9d"
      },
      "outputs": [],
      "source": [
        "import os\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import gc\n",
        "from torch.utils.data import random_split"
      ],
      "id": "73af1d9d"
    },
    {
      "cell_type": "code",
      "source": [
        "from sklearn.model_selection import train_test_split"
      ],
      "metadata": {
        "id": "laD4gCN1AnAp"
      },
      "id": "laD4gCN1AnAp",
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xwWpOPN-SkYk",
        "outputId": "6c363506-bda5-444e-bfb7-03eeb99b34ee"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ],
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "id": "xwWpOPN-SkYk"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0bab0d03"
      },
      "source": [
        "2. Import X and y"
      ],
      "id": "0bab0d03"
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "0e61570a"
      },
      "outputs": [],
      "source": [
        "import torchvision\n",
        "from torchvision import transforms\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "from torch.utils.data import Dataset, DataLoader\n",
        "from PIL import Image"
      ],
      "id": "0e61570a"
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "Q822x4HZTzzr"
      },
      "outputs": [],
      "source": [
        "os.chdir('/content/drive/MyDrive/Thesis')"
      ],
      "id": "Q822x4HZTzzr"
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "99428f03"
      },
      "outputs": [],
      "source": [
        "y = np.load('train_con.npy')\n",
        "y_idx = pd.read_csv('idx_y.csv')"
      ],
      "id": "99428f03"
    },
    {
      "cell_type": "markdown",
      "source": [
        "Remove elements from y that were not RBG but grayscale by index"
      ],
      "metadata": {
        "id": "FrYmCdafe8o2"
      },
      "id": "FrYmCdafe8o2"
    },
    {
      "cell_type": "code",
      "source": [
        "y = y[0:10001]"
      ],
      "metadata": {
        "id": "MHK_2z0ifFex"
      },
      "id": "MHK_2z0ifFex",
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "y_ = np.delete(y,y_idx.iloc[:,1],axis=0)"
      ],
      "metadata": {
        "id": "yHITyHqPepSa"
      },
      "id": "yHITyHqPepSa",
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "y_.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MLdZQPfnemie",
        "outputId": "19d93ff3-a844-41b9-c90d-a211f44914af"
      },
      "id": "MLdZQPfnemie",
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(9976, 3)"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "daf7012a"
      },
      "source": [
        "NEW APPROACH:\n",
        "\n",
        "Fit CNN on orginal images, so no salience maps"
      ],
      "id": "daf7012a"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4fdc2364"
      },
      "source": [
        "1. Read `n-rows`from X."
      ],
      "id": "4fdc2364"
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "r367TL-IeTgB",
        "outputId": "2663d564-acf7-435f-fe76-8686a32bc50f"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "cuda\n"
          ]
        }
      ],
      "source": [
        "if torch.cuda.is_available():\n",
        "  device = torch.device(\"cuda\")\n",
        "else:\n",
        "  device = torch.device(\"cpu\")\n",
        "\n",
        "print(device)"
      ],
      "id": "r367TL-IeTgB"
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "id": "27e2352f"
      },
      "outputs": [],
      "source": [
        "X = np.load('mscoco_og.npy')"
      ],
      "id": "27e2352f"
    },
    {
      "cell_type": "code",
      "source": [
        "X = np.swapaxes(X,0,3)"
      ],
      "metadata": {
        "id": "i2tQHntfhJRz"
      },
      "id": "i2tQHntfhJRz",
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "X = np.swapaxes(X,1,3)"
      ],
      "metadata": {
        "id": "ib4OPsRrht3q"
      },
      "id": "ib4OPsRrht3q",
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "X.shape"
      ],
      "metadata": {
        "id": "OJdPI4eQf4Wy",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4384d600-bd1b-4b9d-9335-669e94c5b59b"
      },
      "id": "OJdPI4eQf4Wy",
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(9976, 3, 224, 224)"
            ]
          },
          "metadata": {},
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "X_ = torch.from_numpy(X).to(device)"
      ],
      "metadata": {
        "id": "jGWIHxD73jAz"
      },
      "id": "jGWIHxD73jAz",
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "y_ = torch.from_numpy(y_).to(device)"
      ],
      "metadata": {
        "id": "zbl3QPuD3_aN"
      },
      "id": "zbl3QPuD3_aN",
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "del X\n",
        "gc.collect()"
      ],
      "metadata": {
        "id": "9tdWWyMBjpga",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "df79e4c6-2722-4008-9496-94def193c974"
      },
      "id": "9tdWWyMBjpga",
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "12"
            ]
          },
          "metadata": {},
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "43fc4d4d"
      },
      "source": [
        "Converting X to appropriate Datatype..."
      ],
      "id": "43fc4d4d"
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "id": "83ce427f",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "35ba3273-6f75-49ff-e7d6-aa7870d2b750"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "<ipython-input-17-e30e4ab0470a>:1: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).\n",
            "  X_ = torch.tensor(X_,dtype = torch.float32)\n"
          ]
        }
      ],
      "source": [
        "X_ = torch.tensor(X_,dtype = torch.float32)\n",
        "\n",
        "y_ = y_.type(torch.LongTensor)"
      ],
      "id": "83ce427f"
    },
    {
      "cell_type": "markdown",
      "source": [
        "To get max value of classN-1"
      ],
      "metadata": {
        "id": "T2DjZNYUrFdA"
      },
      "id": "T2DjZNYUrFdA"
    },
    {
      "cell_type": "code",
      "source": [
        "y_ = y_-1"
      ],
      "metadata": {
        "id": "eFGPr_GdqDwY"
      },
      "id": "eFGPr_GdqDwY",
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "X_.to('cpu');"
      ],
      "metadata": {
        "id": "SUD7b30_4hQQ"
      },
      "id": "SUD7b30_4hQQ",
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "X_train,X_test,y_train,y_test = train_test_split(X_,y_,test_size=0.2,random_state=42)"
      ],
      "metadata": {
        "id": "NSHPQ6UcAepw"
      },
      "id": "NSHPQ6UcAepw",
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "X_train = torch\n",
        ",X_test"
      ],
      "metadata": {
        "id": "nM9fpU1oA9at"
      },
      "id": "nM9fpU1oA9at",
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5379ae48"
      },
      "source": [
        "Setup custom dataset class"
      ],
      "id": "5379ae48"
    },
    {
      "cell_type": "code",
      "execution_count": 28,
      "metadata": {
        "id": "76e13d73"
      },
      "outputs": [],
      "source": [
        "class TrainDataset(Dataset):\n",
        "    def __init__(self, X, y,transform=None):\n",
        "        self.X = X\n",
        "        self.y = y\n",
        "        self.transform = transforms.Compose([transforms.Normalize(torch.mean(X), torch.std(X))])\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.X)\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        X = self.X[idx]\n",
        "        y = self.y[idx]\n",
        "\n",
        "        if self.transform:\n",
        "          X = self.transform(X)\n",
        "\n",
        "        return X, y"
      ],
      "id": "76e13d73"
    },
    {
      "cell_type": "code",
      "source": [
        "class TestDataset(Dataset):\n",
        "    def __init__(self, X, y,transform=None):\n",
        "        self.X = X\n",
        "        self.y = y\n",
        "        self.transform = transforms.Compose([transforms.Normalize(torch.mean(X), torch.std(X))])\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.X)\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        X = self.X[idx]\n",
        "        y = self.y[idx]\n",
        "\n",
        "        if self.transform:\n",
        "          X = self.transform(X)\n",
        "\n",
        "        return X, y"
      ],
      "metadata": {
        "id": "oLANWcxwAzU9"
      },
      "id": "oLANWcxwAzU9",
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "class SimpleDataset(Dataset):\n",
        "    def __init__(self, X, y,transform=None):\n",
        "        self.X = X\n",
        "        self.y = y\n",
        "        self.transform = transforms.Compose([transforms.Normalize(torch.mean(X), torch.std(X))])\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.X)\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        X = self.X[idx]\n",
        "        y = self.y[idx]\n",
        "\n",
        "        if self.transform:\n",
        "          X = self.transform(X)\n",
        "\n",
        "        return X, y"
      ],
      "metadata": {
        "id": "liQyAEKFBMQX"
      },
      "id": "liQyAEKFBMQX",
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "dataset = SimpleDataset(X_,y_,transform=False)"
      ],
      "metadata": {
        "id": "XxSIhM3R4sC7"
      },
      "id": "XxSIhM3R4sC7",
      "execution_count": 32,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(\"length of the dataset is:\", len(dataset))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "E-9ZTvQs41F-",
        "outputId": "41629a24-b9f3-4ad3-dea8-459e59e91180"
      },
      "id": "E-9ZTvQs41F-",
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "length of the dataset is: 9976\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_data, test_data = random_split(dataset, [7000, 2976])"
      ],
      "metadata": {
        "id": "YMFf8uir5FrP"
      },
      "id": "YMFf8uir5FrP",
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "dataloader_train = DataLoader(train_data, batch_size=32, shuffle=True)\n",
        "dataloader_test = DataLoader(test_data, batch_size=32, shuffle=True)"
      ],
      "metadata": {
        "id": "yFIoQgYp5mdz"
      },
      "id": "yFIoQgYp5mdz",
      "execution_count": 36,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "mean = 0.\n",
        "std = 0.\n",
        "for images, _ in dataloader_train:\n",
        "    batch_samples = images.size(0) # batch size (the last batch can have smaller size!)\n",
        "    images = images.view(batch_samples, images.size(1), -1)\n",
        "    mean += images.mean(2).sum(0)\n",
        "    std += images.std(2).sum(0)\n",
        "\n",
        "mean /= len(dataloader_train.dataset)\n",
        "std /= len(dataloader_train.dataset)"
      ],
      "metadata": {
        "id": "XFA2EUgcAVoK"
      },
      "id": "XFA2EUgcAVoK",
      "execution_count": 37,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_mean=mean\n",
        "train_std=std"
      ],
      "metadata": {
        "id": "_BOcvdLpBE_u"
      },
      "id": "_BOcvdLpBE_u",
      "execution_count": 39,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "class SimpleDataset(Dataset):\n",
        "    def __init__(self, X, y,transform=True):\n",
        "        self.X = X\n",
        "        self.y = y\n",
        "        self.transform = transforms.Normalize(train_mean,train_std)\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.X)\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        X = self.X[idx]\n",
        "        y = self.y[idx]\n",
        "\n",
        "        if self.transform:\n",
        "          X = self.transform(X)\n",
        "\n",
        "        return X, y"
      ],
      "metadata": {
        "id": "rKIhChoGBHIo"
      },
      "id": "rKIhChoGBHIo",
      "execution_count": 51,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_data, test_data = random_split(dataset, [7000, 2976])"
      ],
      "metadata": {
        "id": "-IKpmVVVBvpd"
      },
      "id": "-IKpmVVVBvpd",
      "execution_count": 52,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "dataloader_train = DataLoader(train_data, batch_size=32, shuffle=True)\n",
        "dataloader_test = DataLoader(test_data, batch_size=32, shuffle=True)"
      ],
      "metadata": {
        "id": "KqTGHZCeB16K"
      },
      "id": "KqTGHZCeB16K",
      "execution_count": 53,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import torch.nn.functional as F"
      ],
      "metadata": {
        "id": "x_bbuiQtB3_z"
      },
      "id": "x_bbuiQtB3_z",
      "execution_count": 43,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9c4df5ea"
      },
      "source": [
        "Setting up the CNN."
      ],
      "id": "9c4df5ea"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "c5b23ab6"
      },
      "outputs": [],
      "source": [
        "import torch.nn.functional as F\n",
        "\n",
        "class CNN(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(CNN, self).__init__()\n",
        "        self.conv1 = nn.Conv2d(3, 32, 3)\n",
        "        self.batch = nn.BatchNorm2d(32)\n",
        "        self.conv2 = nn.Conv2d(32,64,3) #remove layer\n",
        "        self.dropout1 = nn.Dropout(0.25)\n",
        "        self.dropout2 = nn.Dropout(0.25)\n",
        "        self.fc1 = nn.Linear(774400, 128)\n",
        "        self.fc2 = nn.Linear(128,64)\n",
        "        self.Arousal = nn.Linear(64,10) #3\n",
        "        self.Valence = nn.Linear(64,10)\n",
        "        self.Dominance = nn.Linear(64,10)\n",
        "        self.pool = nn.MaxPool2d(2)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.conv1(x)\n",
        "        x = self.batch(x)\n",
        "        x = F.relu(x)\n",
        "        x = self.conv2(x)\n",
        "        x = F.relu(x)\n",
        "        x = self.pool(x)\n",
        "        x = self.dropout1(x)\n",
        "        x = torch.flatten(x, 1)\n",
        "        x = self.fc1(x)\n",
        "        x = F.relu(x)\n",
        "        x = self.dropout2(x)\n",
        "        x = self.fc2(x)\n",
        "        Arousal = self.Arousal(x)\n",
        "        Valence = self.Valence(x)\n",
        "        Dominance = self.Dominance(x)\n",
        "        return Arousal, Valence, Dominance\n",
        "#removed"
      ],
      "id": "c5b23ab6"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ad94c3ad",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "bd93a88c-8ac8-4a7b-8664-69f0ebe37411"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "CNN(\n",
              "  (conv1): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1))\n",
              "  (batch): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
              "  (conv2): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1))\n",
              "  (dropout1): Dropout(p=0.25, inplace=False)\n",
              "  (dropout2): Dropout(p=0.25, inplace=False)\n",
              "  (fc1): Linear(in_features=774400, out_features=128, bias=True)\n",
              "  (fc2): Linear(in_features=128, out_features=64, bias=True)\n",
              "  (Arousal): Linear(in_features=64, out_features=10, bias=True)\n",
              "  (Valence): Linear(in_features=64, out_features=10, bias=True)\n",
              "  (Dominance): Linear(in_features=64, out_features=10, bias=True)\n",
              "  (pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
              ")"
            ]
          },
          "metadata": {},
          "execution_count": 26
        }
      ],
      "source": [
        "SalienceNet = CNN()\n",
        "SalienceNet.to(device)"
      ],
      "id": "ad94c3ad"
    },
    {
      "cell_type": "code",
      "execution_count": 44,
      "metadata": {
        "id": "eMAlEzRt0VTw"
      },
      "outputs": [],
      "source": [
        "criterion = nn.CrossEntropyLoss()"
      ],
      "id": "eMAlEzRt0VTw"
    },
    {
      "cell_type": "code",
      "execution_count": 45,
      "metadata": {
        "id": "9997283d"
      },
      "outputs": [],
      "source": [
        "from tqdm import tqdm\n",
        "def train(net, loss_fn, dataloader_train,dataloader_val,\n",
        "          epoch=100, learning_rate=0.0001):\n",
        "    optimizer = torch.optim.Adam(net.parameters(), lr=learning_rate,weight_decay=1e-4) # weight decay as a regularization technique to defy overfitting\n",
        "   # for param in model_CNN.parameters():\n",
        "   #     param.requires_grad = False\n",
        "   # for param in model_CNN.Arousal.parameters():\n",
        "   #   param.requires_grad = True\n",
        "   # for param in model_CNN.Valence.parameters():\n",
        "   #   param.requires_grad = True\n",
        "   # for param in model_CNN.Dominance.parameters():\n",
        "      #param.requires_grad = True\n",
        "    train_loss = []\n",
        "    val_loss = []\n",
        "    train_accuracy = []\n",
        "    val_accuracy = []\n",
        "    train_epoch_loss = []\n",
        "    val_epoch_loss = []\n",
        "    for i in range(epoch):\n",
        "        train_loop = tqdm(dataloader_train)\n",
        "        val_loop = tqdm(dataloader_val)\n",
        "        total = 0\n",
        "        correct_a,correct_v,correct_d=0,0,0\n",
        "        net.to(device)\n",
        "        net.train()\n",
        "        for data, label in train_loop:\n",
        "            data, label = data.to(device), label.to(device)\n",
        "            #make prediction\n",
        "            Arousal,Valence,Dominance = net(data.float())\n",
        "\n",
        "            _a, y_pred_train_arousal = torch.max(Arousal,1)\n",
        "            _v, y_pred_train_valence = torch.max(Valence,1)\n",
        "            _d, y_pred_train_dominance = torch.max(Dominance,1)\n",
        "\n",
        "            total += label[:,0].size(0)+label[:,1].size(0)+label[:,2].size(0)\n",
        "\n",
        "            correct_a += (y_pred_train_arousal == label[:,0]).sum().item()\n",
        "            correct_v += (y_pred_train_valence == label[:,1]).sum().item()\n",
        "            correct_d += (y_pred_train_dominance == label[:,2]).sum().item()\n",
        "\n",
        "            sum_correct = (correct_a+correct_v+correct_d)\n",
        "            # Zero out the gradients\n",
        "            optimizer.zero_grad()\n",
        "\n",
        "            # Calculate loss.\n",
        "            #loss = loss_fn(output,label)#torch.max(label.long(), 1)[0])\n",
        "            loss_arousal = loss_fn(Arousal,label[:,0])\n",
        "            loss_valence = loss_fn(Valence,label[:,1])\n",
        "            loss_dominance = loss_fn(Dominance,label[:,2])\n",
        "            loss=loss_arousal+loss_valence+loss_dominance\n",
        "\n",
        "\n",
        "            # Backpropagation and gradient update.\n",
        "            loss.backward()# Calculate gradients. #retrain_graph=True\n",
        "\n",
        "            optimizer.step() # Apply gradient update\n",
        "\n",
        "            train_loss.append(loss.item())\n",
        "            train_loop.set_description(f\"Epoch [{i+1}/{epoch}]\")\n",
        "\n",
        "        train_accuracy.append(sum_correct/total)\n",
        "        train_epoch_loss.append((np.array(train_loss).mean()))\n",
        "\n",
        "        #validation\n",
        "        net.eval()\n",
        "        val_total = 0\n",
        "        val_arousal,val_valence, val_dominance = 0,0,0\n",
        "        for data, labels in dataloader_val:\n",
        "            data, labels = data.to(device), labels.to(device)\n",
        "\n",
        "            Arousal_v,Valence_v,Dominance_v = net(data.float())\n",
        "\n",
        "            _a, y_pred_val_arousal = torch.max(Arousal_v,1)\n",
        "            _v, y_pred_val_valence = torch.max(Valence_v,1)\n",
        "            _d, y_pred_val_dominance = torch.max(Dominance_v,1)\n",
        "\n",
        "            val_total += labels[:,0].size(0)+labels[:,1].size(0)+labels[:,2].size(0)\n",
        "\n",
        "            val_arousal += (y_pred_val_arousal == labels[:,0]).sum().item()\n",
        "            val_valence += (y_pred_val_valence == labels[:,1]).sum().item()\n",
        "            val_dominance += (y_pred_val_dominance == labels[:,2]).sum().item()\n",
        "\n",
        "            sum_correct_val = (val_arousal+val_valence+val_dominance)\n",
        "\n",
        "            loss_arousal_v = loss_fn(Arousal_v,labels[:,0])\n",
        "            loss_valence_v = loss_fn(Valence_v,labels[:,1])\n",
        "            loss_dominance_v = loss_fn(Dominance_v,labels[:,2])\n",
        "\n",
        "            val_l=loss_arousal_v+loss_valence_v+loss_dominance_v\n",
        "\n",
        "            val_loss.append(val_l.item())\n",
        "\n",
        "        val_epoch_loss.append((np.array(val_loss).mean()))\n",
        "\n",
        "        val_accuracy.append(sum_correct_val/val_total)\n",
        "\n",
        "\n",
        "    return train_loss,train_epoch_loss,val_loss,val_epoch_loss, train_accuracy,val_accuracy\n"
      ],
      "id": "9997283d"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "JvEWX6FyZGIk"
      },
      "outputs": [],
      "source": [
        "def test(net,loss_fn,dataloader_test):\n",
        "  net.eval()\n",
        "  test_loss = []\n",
        "  test_total,test_correct = 0,0\n",
        "  test_acc = []\n",
        "  test_loop = tqdm(dataloader_test)\n",
        "  i=0\n",
        "  for batch, (data, labels) in enumerate(test_loop):\n",
        "      print('testing...')\n",
        "\n",
        "      data, labels = data.to(device), labels.to(device)\n",
        "      arousal,valence,dominance = net(data.float())\n",
        "\n",
        "      arousal_t = loss_fn(arousal,labels[:,0])\n",
        "      valence_t = loss_fn(valence,labels[:,1])\n",
        "      dominance_t = loss_fn(dominance,labels[:,2])\n",
        "\n",
        "      test_l = arousal_t+valence_t+dominance_t\n",
        "\n",
        "      test_loss.append(test_l.item())\n",
        "      test_loop.set_description(f\"Batch [{i+1}/{batch}]\")\n",
        "      i=+1\n",
        "\n",
        "        #test_total += labels.size(0)\n",
        "        #test_correct += (y_pred_test==torch.max(labels,1)[0]).sum().item()\n",
        "\n",
        "    #test_acc.append(test_correct/test_total)\n",
        "    #y_pred_test = torch.softmax(outputs,dim=1).argmax(dim=1)\n",
        "\n",
        "   # print('test acc is {}'.format(test_acc)\n",
        "  return test_loss"
      ],
      "id": "JvEWX6FyZGIk"
    },
    {
      "cell_type": "code",
      "execution_count": 46,
      "metadata": {
        "id": "7e318c98"
      },
      "outputs": [],
      "source": [
        "import matplotlib.pyplot as plt"
      ],
      "id": "7e318c98"
    },
    {
      "cell_type": "code",
      "execution_count": 47,
      "metadata": {
        "id": "kGxhNDnUbZMD",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 17
        },
        "outputId": "64709e8f-5e3b-4ce3-af09-4ea5ad50b1d9"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "google.colab.output.setIframeHeight(0, true, {maxHeight: 300})"
            ]
          },
          "metadata": {}
        }
      ],
      "source": [
        "from IPython.display import Javascript\n",
        "\n",
        "display(Javascript('''google.colab.output.setIframeHeight(0, true, {maxHeight: 300})'''))"
      ],
      "id": "kGxhNDnUbZMD"
    },
    {
      "cell_type": "code",
      "source": [
        "from torchvision import models"
      ],
      "metadata": {
        "id": "912Sex6VMmmd"
      },
      "id": "912Sex6VMmmd",
      "execution_count": 48,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model = models.alexnet(pretrained=True)"
      ],
      "metadata": {
        "id": "Hod-SeceMqqE",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f873803b-4943-42e2-aede-f47b8fdd3225"
      },
      "id": "Hod-SeceMqqE",
      "execution_count": 49,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.\n",
            "  warnings.warn(\n",
            "/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=AlexNet_Weights.IMAGENET1K_V1`. You can also use `weights=AlexNet_Weights.DEFAULT` to get the most up-to-date weights.\n",
            "  warnings.warn(msg)\n",
            "Downloading: \"https://download.pytorch.org/models/alexnet-owt-7be5be79.pth\" to /root/.cache/torch/hub/checkpoints/alexnet-owt-7be5be79.pth\n",
            "100%|██████████| 233M/233M [00:02<00:00, 88.9MB/s]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class CNN1(nn.Module):\n",
        "    def __init__(self, pretrained):\n",
        "        super(CNN1, self).__init__()\n",
        "        if pretrained is True:\n",
        "            self.model = models.__dict__[\"alexnet\"](pretrained=\"imagenet\")\n",
        "        else:\n",
        "            self.model = models.__dict__[\"alexnet\"](pretrained=None)\n",
        "        self.Arousal= nn.Linear(256, 10)\n",
        "        self.Valence = nn.Linear(256,10)\n",
        "        self.Dominance = nn.Linear(256, 10)\n",
        "\n",
        "    def forward(self, x):\n",
        "        bs, _, _, _ = x.shape\n",
        "        x = self.model.features(x)\n",
        "        x = F.adaptive_avg_pool2d(x, 1).reshape(bs, -1)\n",
        "        Arousal = self.Arousal(x)\n",
        "        Valence = self.Valence(x)\n",
        "        Dominance = self.Dominance(x)\n",
        "        return Arousal, Valence, Dominance\n",
        "        return {'label1': label1, 'label2': label2, 'label3': label3}"
      ],
      "metadata": {
        "id": "ayX1tkQ6p5r6"
      },
      "id": "ayX1tkQ6p5r6",
      "execution_count": 50,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "with torch.no_grad():\n",
        "  torch.cuda.empty_cache()"
      ],
      "metadata": {
        "id": "JF9XdnAI5c4N"
      },
      "id": "JF9XdnAI5c4N",
      "execution_count": 56,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model_CNN = CNN1(True).to('cpu')"
      ],
      "metadata": {
        "id": "LN5RTKpPqPQG",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "c24610e1-f657-435e-db02-fc2607cde938"
      },
      "id": "LN5RTKpPqPQG",
      "execution_count": 54,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.\n",
            "  warnings.warn(\n",
            "/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=AlexNet_Weights.IMAGENET1K_V1`. You can also use `weights=AlexNet_Weights.DEFAULT` to get the most up-to-date weights.\n",
            "  warnings.warn(msg)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "for param in model_CNN.parameters():\n",
        "        param.requires_grad = False\n",
        "for param in model_CNN.Arousal.parameters():\n",
        "  param.requires_grad = True\n",
        "for param in model_CNN.Valence.parameters():\n",
        "  param.requires_grad = True\n",
        "for param in model_CNN.Dominance.parameters():\n",
        "  param.requires_grad = True"
      ],
      "metadata": {
        "id": "-cX76_sdu0bo"
      },
      "id": "-cX76_sdu0bo",
      "execution_count": 55,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": 57,
      "metadata": {
        "id": "c5ddd06d",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "5f13f9d0-122d-472e-bc83-cdfa8ef2539b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "Epoch [1/20]: 100%|██████████| 219/219 [00:09<00:00, 21.99it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "\n",
            "  0%|          | 0/93 [00:11<?, ?it/s]\n",
            "Epoch [2/20]: 100%|██████████| 219/219 [00:03<00:00, 61.24it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [3/20]: 100%|██████████| 219/219 [00:03<00:00, 58.00it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "\n",
            "  0%|          | 0/93 [00:05<?, ?it/s]\n",
            "Epoch [4/20]: 100%|██████████| 219/219 [00:03<00:00, 61.15it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [5/20]: 100%|██████████| 219/219 [00:03<00:00, 60.42it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [6/20]: 100%|██████████| 219/219 [00:03<00:00, 56.91it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "  0%|          | 0/93 [00:05<?, ?it/s]\n",
            "Epoch [7/20]: 100%|██████████| 219/219 [00:03<00:00, 61.34it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [8/20]: 100%|██████████| 219/219 [00:03<00:00, 61.05it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [9/20]: 100%|██████████| 219/219 [00:03<00:00, 57.39it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [10/20]: 100%|██████████| 219/219 [00:03<00:00, 61.32it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [11/20]: 100%|██████████| 219/219 [00:03<00:00, 55.71it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "\n",
            "  0%|          | 0/93 [00:05<?, ?it/s]\n",
            "Epoch [12/20]: 100%|██████████| 219/219 [00:03<00:00, 57.68it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [13/20]: 100%|██████████| 219/219 [00:03<00:00, 61.39it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [14/20]: 100%|██████████| 219/219 [00:03<00:00, 59.88it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [15/20]: 100%|██████████| 219/219 [00:03<00:00, 59.57it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [16/20]: 100%|██████████| 219/219 [00:03<00:00, 60.99it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [17/20]: 100%|██████████| 219/219 [00:03<00:00, 60.04it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [18/20]: 100%|██████████| 219/219 [00:03<00:00, 60.63it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [19/20]: 100%|██████████| 219/219 [00:03<00:00, 61.59it/s]\n",
            "  0%|          | 0/219 [00:00<?, ?it/s]\n",
            "\n",
            "  0%|          | 0/93 [00:04<?, ?it/s]\n",
            "Epoch [20/20]: 100%|██████████| 219/219 [00:03<00:00, 59.36it/s]\n",
            "  0%|          | 0/93 [00:05<?, ?it/s]\n"
          ]
        }
      ],
      "source": [
        "np.random.seed(1)\n",
        "torch.manual_seed(1)\n",
        "\n",
        "train_loss,train_epoch_loss,val_loss,val_epoch_loss,train_acc,val_acc=train(net=model_CNN,loss_fn=criterion,dataloader_train=dataloader_train,\n",
        "                  dataloader_val=dataloader_test,epoch=20)\n",
        "\n",
        "\n"
      ],
      "id": "c5ddd06d"
    },
    {
      "cell_type": "code",
      "execution_count": 61,
      "metadata": {
        "id": "fk-6haR7G5-q",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 472
        },
        "outputId": "1bd85182-2731-427e-bcb8-faa46f18c24d"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "plt.plot(train_epoch_loss, label = 'Trainingsverlust')\n",
        "plt.plot(val_epoch_loss, label = 'Testverlust')\n",
        "plt.xlabel('Epoche')\n",
        "plt.ylabel('Cross-Entropie')\n",
        "plt.title('Cross-Entropie pro Epoche')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "id": "fk-6haR7G5-q"
    },
    {
      "cell_type": "code",
      "source": [
        "plt.plot(train_acc, label = 'Trainingsgenauigkeit')\n",
        "plt.plot(val_acc, label = 'Testgenauigkeit')\n",
        "plt.xlabel('Epoche')\n",
        "plt.ylabel('Genauigkeit')\n",
        "plt.title('Genauigkeit pro Epoche')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "metadata": {
        "id": "s9cEBXo1dJNj",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 472
        },
        "outputId": "8429388c-057f-49e8-e8f2-970740c572d3"
      },
      "id": "s9cEBXo1dJNj",
      "execution_count": 60,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "provenance": [],
      "include_colab_link": true
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "YOLO",
      "language": "python",
      "name": "yolo"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.10.9"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 5
}