Aerial_Cactus_Identification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Aerial Cactus Identification.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyNO4zo8ZOc1+C2pbei+EiDr",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/sksoumik/machine-learning-projects/blob/master/Aerial_Cactus_Identification.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "E_z0sd8DbHRR"
      },
      "source": [
        "## Download image and EDA"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C_EkxP6gZulr",
        "outputId": "a694e6c7-e80e-456c-a8f2-b719d3409b36",
        "colab": {
          "resources": {
            "http://localhost:8080/nbextensions/google.colab/files.js": {
              "data": "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",
              "ok": true,
              "headers": [
                [
                  "content-type",
                  "application/javascript"
                ]
              ],
              "status": 200,
              "status_text": "OK"
            }
          },
          "base_uri": "https://localhost:8080/",
          "height": 93
        }
      },
      "source": [
        "from google.colab import files\n",
        "# upload kaggle.json file to download data using kaggle API \n",
        "files.upload()"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "     <input type=\"file\" id=\"files-d42e280a-22bf-44cf-bfcf-7e7f3c699c1d\" name=\"files[]\" multiple disabled\n",
              "        style=\"border:none\" />\n",
              "     <output id=\"result-d42e280a-22bf-44cf-bfcf-7e7f3c699c1d\">\n",
              "      Upload widget is only available when the cell has been executed in the\n",
              "      current browser session. Please rerun this cell to enable.\n",
              "      </output>\n",
              "      <script src=\"/nbextensions/google.colab/files.js\"></script> "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Saving kaggle.json to kaggle.json\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'kaggle.json': b'{\"username\":\"sadmansoumik\",\"key\":\"13ae98718981dd5b9e07ae9ed01b5bbe\"}'}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 1
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xc_JeeQwYwgV"
      },
      "source": [
        "!mkdir ~/.kaggle"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "83wYCm2-Z35X"
      },
      "source": [
        "!cp kaggle.json ~/.kaggle/"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wyq_iNrvOQF0",
        "outputId": "e7a6f0fd-889c-4722-c8ba-dc710b58408b",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "# download the data from kaggle competition \n",
        "!kaggle competitions download -c aerial-cactus-identification"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Warning: Your Kaggle API key is readable by other users on this system! To fix this, you can run 'chmod 600 /root/.kaggle/kaggle.json'\n",
            "Warning: Looks like you're using an outdated API Version, please consider updating (server 1.5.9 / client 1.5.4)\n",
            "Downloading train.csv to /content\n",
            "  0% 0.00/667k [00:00<?, ?B/s]\n",
            "100% 667k/667k [00:00<00:00, 43.2MB/s]\n",
            "Downloading sample_submission.csv to /content\n",
            "  0% 0.00/160k [00:00<?, ?B/s]\n",
            "100% 160k/160k [00:00<00:00, 51.9MB/s]\n",
            "Downloading test.zip to /content\n",
            "  0% 0.00/4.20M [00:00<?, ?B/s]\n",
            "100% 4.20M/4.20M [00:00<00:00, 68.7MB/s]\n",
            "Downloading train.zip to /content\n",
            " 68% 13.0M/19.2M [00:00<00:00, 29.3MB/s]\n",
            "100% 19.2M/19.2M [00:00<00:00, 39.2MB/s]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1D3vXcVfRzgE"
      },
      "source": [
        "# unzip train and test folder and remove the zip files \n",
        "%%capture \n",
        "!unzip test.zip\n",
        "!unzip train.zip\n",
        "!rm test.zip\n",
        "!rm train.zip "
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MSPBiE94PM9p"
      },
      "source": [
        "import pandas as pd \n",
        "import numpy as np \n",
        "import matplotlib.pyplot as plt \n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "import torch\n",
        "import torch.nn.functional as F\n",
        "import torchvision\n",
        "import torchvision.transforms as transforms\n",
        "from torch.utils.data import Dataset, DataLoader\n",
        "\n",
        "%matplotlib inline"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NuNfwUhERE0G",
        "outputId": "8c467b5d-b3e0-485b-ab20-1f987ce377c4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "import os\n",
        "\n",
        "# get current directory \n",
        "os.getcwd()"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "'/content'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "w6Jzeu5URfVn",
        "outputId": "2c9527ce-9b23-4708-8406-ea6e7a97cc8f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 143
        }
      },
      "source": [
        "# train dataframe\n",
        "\n",
        "labels = pd.read_csv(\"/content/train.csv\")\n",
        "labels.head(3) "
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>id</th>\n",
              "      <th>has_cactus</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0004be2cfeaba1c0361d39e2b000257b.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>000c8a36845c0208e833c79c1bffedd1.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>000d1e9a533f62e55c289303b072733d.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                     id  has_cactus\n",
              "0  0004be2cfeaba1c0361d39e2b000257b.jpg           1\n",
              "1  000c8a36845c0208e833c79c1bffedd1.jpg           1\n",
              "2  000d1e9a533f62e55c289303b072733d.jpg           1"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hU0JygQKS3IU"
      },
      "source": [
        "train_path = \"/content/train/\"\n",
        "test_path = \"/content/test/\""
      ],
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v_rGZL8NWNwS",
        "outputId": "103b86bc-d169-406a-af12-61c02ea9f359",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "labels['has_cactus'].value_counts()"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1    13136\n",
              "0     4364\n",
              "Name: has_cactus, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mPjQmNqRapJj"
      },
      "source": [
        "```\n",
        "has_cactus = 1 : has cactus \n",
        "has_cactus = 0 : hasn't cactus\n",
        "```"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JW1jaEOvZgFT",
        "outputId": "0c5f754a-2f68-4bc4-8501-ec2230e835e7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 357
        }
      },
      "source": [
        "label = 'Hasn\\'t Cactus', 'Has Cactus' \n",
        "plt.figure(figsize = (6,6))\n",
        "plt.pie(labels.groupby('has_cactus').size(), labels = label, autopct='%1.1f%%', shadow=False, startangle=90)\n",
        "plt.show()"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "muzjmIDDcJY9",
        "outputId": "60448938-8453-400d-a2d6-88c6ee9eb428",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 147
        }
      },
      "source": [
        "# display 5 images from train data that has cactus \n",
        "import matplotlib.image as mpimg\n",
        "\n",
        "fig, ax = plt.subplots(1,5,figsize = (10,2))\n",
        "\n",
        "for i,idx in enumerate(labels[labels['has_cactus'] == 1]['id'][-5:]):\n",
        "    path = os.path.join(train_path,idx)\n",
        "    ax[i].imshow(mpimg.imread(path))"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x144 with 5 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qPp7sGQldmmA",
        "outputId": "f2cf5113-50a3-469a-c498-0112ea607831",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 147
        }
      },
      "source": [
        "# display 5 images from train data that has no cactus \n",
        "\n",
        "fig,ax = plt.subplots(1,5,figsize = (10,2))\n",
        "for i,idx in enumerate(labels[labels['has_cactus'] == 0]['id'][:5]):\n",
        "    path = os.path.join(train_path,idx)\n",
        "    ax[i].imshow(mpimg.imread(path))"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x144 with 5 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZXuue5pcbM-i"
      },
      "source": [
        "## Pre-processing"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6j9kxruMeK4I"
      },
      "source": [
        "# standardize image by mean and standard deviation \n",
        "\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "def imshow(image, ax=None, title=None, normalize=True):\n",
        "    if ax is None:\n",
        "        fig, ax = plt.subplots()\n",
        "    image = image.numpy().transpose((1, 2, 0))\n",
        "\n",
        "    if normalize:\n",
        "        mean = np.array([0.485, 0.456, 0.406])\n",
        "        std = np.array([0.229, 0.224, 0.225])\n",
        "        image = std * image + mean\n",
        "        image = np.clip(image, 0, 1)\n",
        "\n",
        "    ax.imshow(image)\n",
        "    ax.spines['top'].set_visible(False)\n",
        "    ax.spines['right'].set_visible(False)\n",
        "    ax.spines['left'].set_visible(False)\n",
        "    ax.spines['bottom'].set_visible(False)\n",
        "    ax.tick_params(axis='both', length=0)\n",
        "    ax.set_xticklabels('')\n",
        "    ax.set_yticklabels('')\n",
        "\n",
        "    return ax"
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9IOegVEtgSCE"
      },
      "source": [
        "import torch.optim as optim \n",
        "import torch.utils.data as utils\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim \n",
        "\n",
        "import warnings\n",
        "warnings.filterwarnings(\"ignore\")"
      ],
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "N3fqzqYhkXJp"
      },
      "source": [
        "## Dataset clss"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0BFpDXQ6fFaJ"
      },
      "source": [
        "# class ImageData(Dataset):\n",
        "#     def __init__(self, df, data_dir, transform):\n",
        "#         super().__init__()\n",
        "#         self.df = df\n",
        "#         self.data_dir = data_dir\n",
        "#         self.transform = transform\n",
        "\n",
        "#     def __len__(self):\n",
        "#         return len(self.df)\n",
        "    \n",
        "#     def __getitem__(self, index):       \n",
        "#         img_name = self.df.id[index]\n",
        "#         label = self.df.has_cactus[index]\n",
        "        \n",
        "#         img_path = os.path.join(self.data_dir, img_name)\n",
        "#         image = mpimg.imread(img_path)\n",
        "#         image = self.transform(image)\n",
        "#         return image, label\n",
        "\n",
        "class CactusDataset(Dataset):\n",
        "    def __init__(self, df, path, transform = None):\n",
        "        super().__init__()\n",
        "        self.df = df.values\n",
        "        self.path = path\n",
        "        self.transform = transform\n",
        "        \n",
        "    def __len__(self):\n",
        "        return len(self.df)\n",
        "    \n",
        "    def __getitem__(self,index):\n",
        "        img_name, label = self.df[index]\n",
        "        img_path = os.path.join(self.path, img_name)\n",
        "        image = mpimg.imread(img_path)\n",
        "        if self.transform is not None:\n",
        "            image = self.transform(image)\n",
        "        return image, label"
      ],
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eCytTyWzjyjr"
      },
      "source": [
        "# data_transf = transforms.Compose([transforms.ToPILImage(), transforms.ToTensor()])\n",
        "# train_data = ImageData(df = labels, data_dir = train_path, transform = data_transf)\n",
        "# train_loader = DataLoader(dataset = train_data, batch_size = 64, shuffle=True) "
      ],
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sZvHDM_jkcfb"
      },
      "source": [
        "## Transforms"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gsQE9tbifUo9"
      },
      "source": [
        "means, std = (0.5, 0.5, 0.5), (0.5, 0.5, 0.5)\n",
        "\n",
        "train_transform = transforms.Compose([transforms.ToPILImage(),\n",
        "                                      transforms.ToTensor(),\n",
        "                                      transforms.Normalize(means,std)])"
      ],
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rrkSc1apgS7j"
      },
      "source": [
        "test_transform = transforms.Compose([transforms.ToPILImage(),\n",
        "                                     transforms.ToTensor(),\n",
        "                                     transforms.Normalize(means,std)])"
      ],
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mubmQ9XXgSta"
      },
      "source": [
        "valid_transform = transforms.Compose([transforms.ToPILImage(),\n",
        "                                     transforms.ToTensor(),\n",
        "                                     transforms.Normalize(means,std)])"
      ],
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Z4QVGAhPgcFE"
      },
      "source": [
        "## Split the dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wVDIcgTEgbgB"
      },
      "source": [
        "# split the train data into train and validation set \n",
        "\n",
        "train_df, valid_df = train_test_split(labels, \n",
        "                                     stratify=labels.has_cactus, \n",
        "                                     test_size=0.2)"
      ],
      "execution_count": 21,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "d07CWUuzjchj",
        "outputId": "1ab29329-44d4-4ae0-9091-3fe3203236f6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 112
        }
      },
      "source": [
        "train_df.head(2)"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>id</th>\n",
              "      <th>has_cactus</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>14876</th>\n",
              "      <td>d8bbaa363027f93a34c7782a8f309ec8.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1147</th>\n",
              "      <td>107b916b3eefd665c925d0a56b4cf7e7.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                         id  has_cactus\n",
              "14876  d8bbaa363027f93a34c7782a8f309ec8.jpg           1\n",
              "1147   107b916b3eefd665c925d0a56b4cf7e7.jpg           1"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "odWF6KgBjWdi"
      },
      "source": [
        "train_df, test_df = train_test_split(train_df, \n",
        "                                     stratify=train_df.has_cactus, \n",
        "                                     test_size=0.2)"
      ],
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wg0_ZBRTiqxh",
        "outputId": "32c2d974-2cfc-4c08-a1c6-9c4011a2234b",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "len(train_df), len(valid_df), len(test_df) "
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(11200, 3500, 2800)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 24
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SMDuqH_pkPaY"
      },
      "source": [
        "# train data \n",
        "train_data = CactusDataset(train_df, train_path, train_transform)"
      ],
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6zpXF9JtkPeH"
      },
      "source": [
        "# validation data \n",
        "valid_data = CactusDataset(valid_df, train_path, transform=valid_transform)"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "O4PqKoK_lYO6"
      },
      "source": [
        "# test data \n",
        "test_data = CactusDataset(test_df, train_path, transform=test_transform)"
      ],
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wi2N0N3Vm-5z"
      },
      "source": [
        "## Hyperparameters"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aj9QrUECm-KT"
      },
      "source": [
        "# define hyperparameters \n",
        "\n",
        "num_epochs = 15\n",
        "num_classes = 2\n",
        "batch_size = 25\n",
        "learning_rate = 0.001"
      ],
      "execution_count": 28,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jJJf32t7mtze"
      },
      "source": [
        "# data loaders\n",
        "\n",
        "train_loader = DataLoader(dataset=train_data, \n",
        "                          batch_size = batch_size, \n",
        "                          shuffle=True, \n",
        "                          num_workers=0)\n",
        "\n",
        "valid_loader = DataLoader(dataset=valid_data, \n",
        "                          batch_size = batch_size, \n",
        "                          shuffle=False, \n",
        "                          num_workers=0)\n",
        "\n",
        "test_loader = DataLoader(dataset=test_data, \n",
        "                         batch_size = batch_size, \n",
        "                         shuffle=False, \n",
        "                         num_workers=0)"
      ],
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rIO1-Zvwl7uR"
      },
      "source": [
        "## Custom CNN Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IMrKcj0cl656"
      },
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "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(in_channels=3, out_channels=10, kernel_size=3)\n",
        "        self.conv2 = nn.Conv2d(10, 20, kernel_size=3)\n",
        "        self.conv2_drop = nn.Dropout2d()\n",
        "        self.fc1 = nn.Linear(720, 1024)\n",
        "        self.fc2 = nn.Linear(1024, 2)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = F.relu(F.max_pool2d(self.conv1(x), 2))\n",
        "        x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))\n",
        "        x = x.view(x.shape[0],-1)\n",
        "        x = F.relu(self.fc1(x))\n",
        "        x = F.dropout(x, training=self.training)\n",
        "        x = self.fc2(x)\n",
        "        return x"
      ],
      "execution_count": 30,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "f3yX18anmA-U"
      },
      "source": [
        "device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')"
      ],
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JBS3Hj9dmA8Q",
        "outputId": "db151a04-c9f9-46b0-85b2-de9c1e87d6b6",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "model = CNN()\n",
        "print(model)"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "CNN(\n",
            "  (conv1): Conv2d(3, 10, kernel_size=(3, 3), stride=(1, 1))\n",
            "  (conv2): Conv2d(10, 20, kernel_size=(3, 3), stride=(1, 1))\n",
            "  (conv2_drop): Dropout2d(p=0.5, inplace=False)\n",
            "  (fc1): Linear(in_features=720, out_features=1024, bias=True)\n",
            "  (fc2): Linear(in_features=1024, out_features=2, bias=True)\n",
            ")\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lUkx6MckmA6l"
      },
      "source": [
        "model = CNN().to(device)\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "optimizer = torch.optim.Adam(model.parameters(),lr = learning_rate)"
      ],
      "execution_count": 33,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SbDfHybkmA3q",
        "outputId": "758ff90d-22d6-4752-be34-b1e2f4fda3df",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "%%time\n",
        "# keeping-track-of-losses \n",
        "train_losses = []\n",
        "valid_losses = []\n",
        "\n",
        "for epoch in range(1, num_epochs + 1):\n",
        "    # keep-track-of-training-and-validation-loss\n",
        "    train_loss = 0.0\n",
        "    valid_loss = 0.0\n",
        "    \n",
        "    # training-the-model\n",
        "    model.train()\n",
        "    for data, target in train_loader:\n",
        "        # move-tensors-to-GPU \n",
        "        data = data.to(device)\n",
        "        target = target.to(device)\n",
        "        \n",
        "        # clear-the-gradients-of-all-optimized-variables\n",
        "        optimizer.zero_grad()\n",
        "        # forward-pass: compute-predicted-outputs-by-passing-inputs-to-the-model\n",
        "        output = model(data)\n",
        "        # calculate-the-batch-loss\n",
        "        loss = criterion(output, target)\n",
        "        # backward-pass: compute-gradient-of-the-loss-wrt-model-parameters\n",
        "        loss.backward()\n",
        "        # perform-a-ingle-optimization-step (parameter-update)\n",
        "        optimizer.step()\n",
        "        # update-training-loss\n",
        "        train_loss += loss.item() * data.size(0)\n",
        "        \n",
        "    # validate-the-model\n",
        "    model.eval()\n",
        "    for data, target in valid_loader:\n",
        "        \n",
        "        data = data.to(device)\n",
        "        target = target.to(device)\n",
        "        \n",
        "        output = model(data)\n",
        "        \n",
        "        loss = criterion(output, target)\n",
        "        \n",
        "        # update-average-validation-loss \n",
        "        valid_loss += loss.item() * data.size(0)\n",
        "    \n",
        "    # calculate-average-losses\n",
        "    train_loss = train_loss/len(train_loader.sampler)\n",
        "    valid_loss = valid_loss/len(valid_loader.sampler)\n",
        "    train_losses.append(train_loss)\n",
        "    valid_losses.append(valid_loss)\n",
        "        \n",
        "    # print-training/validation-statistics \n",
        "    print('Epoch: {} \\tTraining Loss: {:.6f} \\tValidation Loss: {:.6f}'.format(\n",
        "        epoch, train_loss, valid_loss))"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch: 1 \tTraining Loss: 0.209069 \tValidation Loss: 0.079070\n",
            "Epoch: 2 \tTraining Loss: 0.106642 \tValidation Loss: 0.058605\n",
            "Epoch: 3 \tTraining Loss: 0.095184 \tValidation Loss: 0.063400\n",
            "Epoch: 4 \tTraining Loss: 0.078689 \tValidation Loss: 0.056614\n",
            "Epoch: 5 \tTraining Loss: 0.072361 \tValidation Loss: 0.054775\n",
            "Epoch: 6 \tTraining Loss: 0.071070 \tValidation Loss: 0.053702\n",
            "Epoch: 7 \tTraining Loss: 0.059541 \tValidation Loss: 0.051485\n",
            "Epoch: 8 \tTraining Loss: 0.060267 \tValidation Loss: 0.045912\n",
            "Epoch: 9 \tTraining Loss: 0.061541 \tValidation Loss: 0.039526\n",
            "Epoch: 10 \tTraining Loss: 0.059396 \tValidation Loss: 0.054103\n",
            "Epoch: 11 \tTraining Loss: 0.053523 \tValidation Loss: 0.040252\n",
            "Epoch: 12 \tTraining Loss: 0.052536 \tValidation Loss: 0.043402\n",
            "Epoch: 13 \tTraining Loss: 0.054019 \tValidation Loss: 0.046389\n",
            "Epoch: 14 \tTraining Loss: 0.049426 \tValidation Loss: 0.041892\n",
            "Epoch: 15 \tTraining Loss: 0.045210 \tValidation Loss: 0.037367\n",
            "CPU times: user 1min 57s, sys: 5.08 s, total: 2min 2s\n",
            "Wall time: 2min 3s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BVCETUSRvLQg",
        "outputId": "4cedbe21-b953-4dd0-c031-4ce6fa154258",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "# test-the-model\n",
        "\n",
        "model.eval()  # it-disables-dropout\n",
        "with torch.no_grad():\n",
        "    correct = 0\n",
        "    total = 0\n",
        "    for images, labels in test_loader: \n",
        "        images = images.to(device)\n",
        "        labels = labels.to(device)\n",
        "        outputs = model(images)\n",
        "        _, predicted = torch.max(outputs.data, 1)\n",
        "        total += labels.size(0)\n",
        "        correct += (predicted == labels).sum().item()\n",
        "          \n",
        "    print('Test Accuracy of the model: {} %'.format(100 * correct / total))\n",
        "# Save the model \n",
        "torch.save(model.state_dict(), 'model.ckpt')"
      ],
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test Accuracy of the model: 98.60714285714286 %\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zGUE37BhvLMH",
        "outputId": "a288eeb0-6012-496b-efa2-57dd268bdb3b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        }
      },
      "source": [
        "%matplotlib inline\n",
        "%config InlineBackend.figure_format = 'retina'\n",
        "\n",
        "plt.plot(train_losses, label='Training loss')\n",
        "plt.plot(valid_losses, label='Validation loss')\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"Loss\")\n",
        "plt.legend(frameon=False)"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f4881f6a630>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 38
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "image/png": {
              "width": 398,
              "height": 261
            },
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TLHHPVRAncuu"
      },
      "source": [
        "## Transfer Learning (EfficientNet)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "taxr_nyUAZnN"
      },
      "source": [
        "import numpy as np\n",
        "import pandas as pd\n",
        "import os\n",
        "import matplotlib.image as mpimg\n",
        "\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim \n",
        "\n",
        "import torchvision\n",
        "from torch.utils.data import DataLoader, Dataset\n",
        "import torch.utils.data as utils\n",
        "from torchvision import transforms\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "\n",
        "import warnings\n",
        "warnings.filterwarnings(\"ignore\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nbbqcSFPAHTE"
      },
      "source": [
        "train_data = pd.read_csv(\"/content/train.csv\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "d_PHEYZEECl_",
        "outputId": "5a8a70b1-2a8a-4be3-8d6e-87015017d35b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 206
        }
      },
      "source": [
        "train_data.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>id</th>\n",
              "      <th>has_cactus</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0004be2cfeaba1c0361d39e2b000257b.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>000c8a36845c0208e833c79c1bffedd1.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>000d1e9a533f62e55c289303b072733d.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0011485b40695e9138e92d0b3fb55128.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>0014d7a11e90b62848904c1418fc8cf2.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                     id  has_cactus\n",
              "0  0004be2cfeaba1c0361d39e2b000257b.jpg           1\n",
              "1  000c8a36845c0208e833c79c1bffedd1.jpg           1\n",
              "2  000d1e9a533f62e55c289303b072733d.jpg           1\n",
              "3  0011485b40695e9138e92d0b3fb55128.jpg           1\n",
              "4  0014d7a11e90b62848904c1418fc8cf2.jpg           1"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 80
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "835PXQOUAmrC"
      },
      "source": [
        "train_path = \"/content/train/\""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UkKpP1NKHtAZ"
      },
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "train_df, test_df = train_test_split(train_data,\n",
        "                                    stratify=train_data['has_cactus'], \n",
        "                                    test_size=0.20)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "75IFccfcGyz1"
      },
      "source": [
        "train_df.reset_index(inplace = True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yA2-3ZEGGyxI"
      },
      "source": [
        "test_df.reset_index(inplace = True) "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oaWrS2cgGyue",
        "outputId": "e39393a0-0869-480e-c9d1-b3d80c66b774",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 206
        }
      },
      "source": [
        "test_df.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>index</th>\n",
              "      <th>id</th>\n",
              "      <th>has_cactus</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>283</td>\n",
              "      <td>043d6f04479edc752dfdf5d62af43491.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>4283</td>\n",
              "      <td>3c67539419c47afa1d99ce9bb21352e6.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>17012</td>\n",
              "      <td>f8c0b24fa0490b66c81770b45e8bc9ef.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>539</td>\n",
              "      <td>07ee453070fd57ded2f718b58602b86a.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>502</td>\n",
              "      <td>0754c6e64a4286570f907d7916a15fe3.jpg</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   index                                    id  has_cactus\n",
              "0    283  043d6f04479edc752dfdf5d62af43491.jpg           1\n",
              "1   4283  3c67539419c47afa1d99ce9bb21352e6.jpg           1\n",
              "2  17012  f8c0b24fa0490b66c81770b45e8bc9ef.jpg           1\n",
              "3    539  07ee453070fd57ded2f718b58602b86a.jpg           1\n",
              "4    502  0754c6e64a4286570f907d7916a15fe3.jpg           1"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 85
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rCon_liOq_oJ"
      },
      "source": [
        "class ImageData(Dataset):\n",
        "    def __init__(self, df, data_dir, transform):\n",
        "        super().__init__()\n",
        "        self.df = df\n",
        "        self.data_dir = data_dir\n",
        "        self.transform = transform\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.df)\n",
        "    \n",
        "    def __getitem__(self, index):       \n",
        "        img_name = self.df.id[index]\n",
        "        label = self.df.has_cactus[index]\n",
        "        \n",
        "        img_path = os.path.join(self.data_dir, img_name)\n",
        "        image = mpimg.imread(img_path)\n",
        "        image = self.transform(image)\n",
        "        return image, label"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NvGfsV600yoF",
        "outputId": "88a7bb9d-bb6d-472a-d712-00df726fc625",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "len(train_df)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "14000"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 87
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tEQYAsiwrFNy"
      },
      "source": [
        "data_transf = transforms.Compose([transforms.ToPILImage(), transforms.ToTensor()])\n",
        "\n",
        "train_data = ImageData(df = train_df, data_dir = train_path, transform = data_transf)\n",
        "train_loader = DataLoader(dataset = train_data, batch_size = 64)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kOepKx_4zG09"
      },
      "source": [
        "test_data = ImageData(df = test_df, data_dir = train_path, transform = data_transf)\n",
        "test_loader = DataLoader(dataset = test_data, shuffle=False) "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bUe6njJ8w5GV",
        "outputId": "d22cdb71-d83b-4d0f-c8d9-04afc1b36f25",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "!pip install efficientnet_pytorch"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: efficientnet_pytorch in /usr/local/lib/python3.6/dist-packages (0.7.0)\n",
            "Requirement already satisfied: torch in /usr/local/lib/python3.6/dist-packages (from efficientnet_pytorch) (1.6.0+cu101)\n",
            "Requirement already satisfied: future in /usr/local/lib/python3.6/dist-packages (from torch->efficientnet_pytorch) (0.16.0)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from torch->efficientnet_pytorch) (1.18.5)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hDCm8f8ouYR0"
      },
      "source": [
        "## EfficientNet"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KXC6DAAsrzat"
      },
      "source": [
        "from efficientnet_pytorch import EfficientNet\n",
        "model_eff = EfficientNet.from_name('efficientnet-b1')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2RxI4wXjr18W"
      },
      "source": [
        "# Unfreeze model weights\n",
        "for param in model_eff.parameters():\n",
        "    param.requires_grad = True"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XS9Cs8_br64J"
      },
      "source": [
        "num_ftrs = model_eff._fc.in_features\n",
        "model_eff._fc = nn.Linear(num_ftrs, 1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kV8n3j22sINR"
      },
      "source": [
        "model_eff = model_eff.to('cuda')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nHw_nzQxsOPU"
      },
      "source": [
        "optimizer = optim.Adam(model_eff.parameters())\n",
        "loss_func = nn.BCELoss()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xRZKKlxIBhCZ",
        "outputId": "ec071426-8275-4371-9a55-18384efc1293",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "%%time\n",
        "# Train model\n",
        "loss_log = []\n",
        "\n",
        "for epoch in range(5):    \n",
        "    model_eff.train()    \n",
        "    for ii, (data, target) in enumerate(train_loader):\n",
        "        data, target = data.cuda(), target.cuda()\n",
        "        target = target.float()                \n",
        "\n",
        "        optimizer.zero_grad()\n",
        "        output = model_eff(data)                \n",
        "    \n",
        "        m = nn.Sigmoid()\n",
        "        loss = loss_func(m(output), target)\n",
        "        loss.backward()\n",
        "\n",
        "        optimizer.step()  \n",
        "        \n",
        "        if ii % 1000 == 0:\n",
        "            loss_log.append(loss.item())\n",
        "       \n",
        "    print('Epoch: {} - Loss: {:.6f}'.format(epoch + 1, loss.item()))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch: 1 - Loss: 0.541067\n",
            "Epoch: 2 - Loss: 0.554949\n",
            "Epoch: 3 - Loss: 0.544592\n",
            "Epoch: 4 - Loss: 0.529942\n",
            "Epoch: 5 - Loss: 0.529917\n",
            "CPU times: user 2min 9s, sys: 1.54 s, total: 2min 10s\n",
            "Wall time: 2min 11s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MOrp5RdLB1-K",
        "outputId": "a996fbbf-876f-4c58-b0cb-d101ef6ca536",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "%%time\n",
        "predict = []\n",
        "model_eff.eval()\n",
        "for i, (data, _) in enumerate(test_loader):\n",
        "    data = data.cuda()\n",
        "    output = model_eff(data)    \n",
        "\n",
        "    pred = torch.sigmoid(output)\n",
        "    predicted_vals = pred > 0.5\n",
        "    predict.append(int(predicted_vals))\n",
        "    \n",
        "test_df['prediction'] = predict"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "CPU times: user 1min 10s, sys: 143 ms, total: 1min 10s\n",
            "Wall time: 1min 10s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hyulIySlOeie",
        "outputId": "ffd225c5-4e27-4881-89d8-5fdecf8f006d",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "test_df['prediction'].value_counts()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1    3500\n",
              "Name: prediction, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 104
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2ACP4QV5PXhi"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}