worstRunEver.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "authorship_tag": "ABX9TyOIfH2VwUeUJu1uuGiSlnd9",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU",
    "gpuClass": "standard"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/threedee003/NotebooksForDL/blob/main/worstRunEver.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "9h2o4RQvTWOw",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "081574ca-491d-46d9-f28e-8bc9a26701f6"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ],
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "trainDir = \"/content/drive/MyDrive/GIS/trainTiles\"\n",
        "testDir = \"/content/drive/MyDrive/GIS/testTiles\""
      ],
      "metadata": {
        "id": "_cvvaZxBvHgw"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "import os\n",
        "import skimage.io as io\n",
        "import time\n",
        "from itertools import product\n",
        "\n",
        "\n",
        "\n",
        "import torch.nn as nn\n",
        "import torch\n",
        "import torch.optim as optim\n",
        "from torchsummary import summary\n",
        "from torch.utils.data import Dataset\n",
        "from torchvision.transforms import transforms"
      ],
      "metadata": {
        "id": "nMRNA6IgaLo_"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "class CustomDataset(Dataset):\n",
        "    def __init__(self, data_dir, transform=None):\n",
        "        self.data_dir = data_dir\n",
        "        self.transform = transform\n",
        "        self.imageFilenames = sorted(os.listdir(os.path.join(data_dir, \"tiles\")))\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.imageFilenames)\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        imagePath = os.path.join(self.data_dir, \"tiles\", self.imageFilenames[idx])\n",
        "        img = np.load(imagePath)\n",
        "        X = img[:,:,0:5]\n",
        "        Y = img[:,:,-1]\n",
        "\n",
        "\n",
        "        \n",
        "\n",
        "        if self.transform:\n",
        "            X = self.transform(X)\n",
        "            Y = self.transform(Y)\n",
        "\n",
        "        return X, Y\n"
      ],
      "metadata": {
        "id": "a4IMKIJoLWSL"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "transform = transforms.Compose([\n",
        "    transforms.ToTensor(),\n",
        "])"
      ],
      "metadata": {
        "id": "c226ZLD4XTat"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "trainSet = CustomDataset(trainDir, transform=transform)\n",
        "validationSet = CustomDataset(testDir, transform=transform)\n",
        "\n",
        "trainLoader = torch.utils.data.DataLoader(trainSet, batch_size=None, shuffle=True)\n",
        "validationLoader = torch.utils.data.DataLoader(validationSet, batch_size = None, shuffle = True)"
      ],
      "metadata": {
        "id": "2OD0pDzgXTYF"
      },
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(\"Number of training samples : {}\".format(len(trainLoader)))\n",
        "print(\"Number of validating samples : {}\".format(len(validationLoader)))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rNAe2EhjvgYx",
        "outputId": "7f871f4b-4cf7-45a9-98ca-66f693e84db2"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Number of training samples : 1120\n",
            "Number of validating samples : 440\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class Autoencoder1(nn.Module):\n",
        "  def __init__(self,channels):\n",
        "    super(Autoencoder1,self).__init__()\n",
        "    self.encoder = nn.Sequential(\n",
        "        nn.Conv2d(channels,64, kernel_size = 3),\n",
        "        nn.Conv2d(64,64,kernel_size = 3, stride = 2),\n",
        "        nn.Conv2d(64,128,kernel_size= 3),\n",
        "        nn.Conv2d(128,128,kernel_size = 3,stride = 2),\n",
        "        nn.Conv2d(128,256,kernel_size = 3),\n",
        "        nn.Conv2d(256,256,kernel_size = 3,stride = 2),\n",
        "        nn.Conv2d(256,512,kernel_size = 3)\n",
        "    )\n",
        "\n",
        "    self.decoder = nn.Sequential(\n",
        "        nn.Conv2d(512,256,kernel_size = 2),\n",
        "        nn.ConvTranspose2d(256,256,kernel_size=4, stride=2, padding=1, output_padding=0),\n",
        "        nn.Conv2d(256,128,kernel_size = 3),\n",
        "        nn.ConvTranspose2d(128,128,kernel_size=4, stride=2, padding=1, output_padding=0),\n",
        "        nn.Conv2d(128,64,kernel_size = 3),\n",
        "        nn.ConvTranspose2d(64,64,kernel_size=4, stride=2, padding=1, output_padding=0),\n",
        "        nn.Conv2d(64,1,kernel_size = 3)\n",
        "    )\n",
        "\n",
        "\n",
        "  def forward(self,x):\n",
        "    x = self.encoder(x)\n",
        "    x = self.decoder(x)\n",
        "\n",
        "    return x"
      ],
      "metadata": {
        "id": "mUvldmB7Xepl"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "class Autoencoder2(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(Autoencoder2, self).__init__()\n",
        "        self.encoder = nn.Sequential(\n",
        "            nn.Conv2d(5, 32, kernel_size=3, stride=1, padding=1),\n",
        "            nn.ReLU(),\n",
        "            nn.MaxPool2d(kernel_size=2, stride=2),\n",
        "            nn.Conv2d(32, 16, kernel_size=3, stride=1, padding=1),\n",
        "            nn.ReLU(),\n",
        "            nn.MaxPool2d(kernel_size=2, stride=2)\n",
        "        )\n",
        "        self.decoder = nn.Sequential(\n",
        "            nn.ConvTranspose2d(16, 32, kernel_size=3, stride=2, padding=1, output_padding=1),\n",
        "            nn.ReLU(),\n",
        "            nn.ConvTranspose2d(32, 1, kernel_size=3, stride=2, padding=1, output_padding=1),\n",
        "            nn.Sigmoid()\n",
        "        )\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.encoder(x)\n",
        "        x = self.decoder(x)\n",
        "        return x"
      ],
      "metadata": {
        "id": "jQ4F4_W5agWh"
      },
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "class Autoencoder3(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(Autoencoder3, self).__init__()\n",
        "        \n",
        "        # Encoder\n",
        "        self.conv1 = nn.Conv2d(5, 32, kernel_size=3, stride=1, padding=1)\n",
        "        self.relu1 = nn.ReLU(inplace=True)\n",
        "        self.conv2 = nn.Conv2d(32, 64, kernel_size=3, stride=2, padding=1)\n",
        "        self.relu2 = nn.ReLU(inplace=True)\n",
        "        self.conv3 = nn.Conv2d(64, 128, kernel_size=3, stride=2, padding=1)\n",
        "        self.relu3 = nn.ReLU(inplace=True)\n",
        "        self.conv4 = nn.Conv2d(128, 256, kernel_size=3, stride=2, padding=1)\n",
        "        self.relu4 = nn.ReLU(inplace=True)\n",
        "        \n",
        "        # Decoder\n",
        "        self.conv5 = nn.ConvTranspose2d(256, 128, kernel_size=3, stride=2, padding=1, output_padding=1)\n",
        "        self.relu5 = nn.ReLU(inplace=True)\n",
        "        self.conv6 = nn.ConvTranspose2d(128, 64, kernel_size=3, stride=2, padding=1, output_padding=1)\n",
        "        self.relu6 = nn.ReLU(inplace=True)\n",
        "        self.conv7 = nn.ConvTranspose2d(64, 32, kernel_size=3, stride=2, padding=1, output_padding=1)\n",
        "        self.relu7 = nn.ReLU(inplace=True)\n",
        "        self.conv8 = nn.ConvTranspose2d(32, 1, kernel_size=3, stride=1, padding=1)\n",
        "        self.sigmoid = nn.Sigmoid()\n",
        "        \n",
        "    def forward(self, x):\n",
        "        # Encoder\n",
        "        x = self.relu1(self.conv1(x))\n",
        "        x = self.relu2(self.conv2(x))\n",
        "        x = self.relu3(self.conv3(x))\n",
        "        x = self.relu4(self.conv4(x))\n",
        "        \n",
        "        # Decoder\n",
        "        x = self.relu5(self.conv5(x))\n",
        "        x = self.relu6(self.conv6(x))\n",
        "        x = self.relu7(self.conv7(x))\n",
        "        x = self.sigmoid(self.conv8(x))\n",
        "        return x"
      ],
      "metadata": {
        "id": "dFFRBpyk7L_J"
      },
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "autoenc = Autoencoder3()"
      ],
      "metadata": {
        "id": "NfD2SYtexkmo"
      },
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "summary(autoenc.to('cuda'),(5,64,64))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9EtIpUFp7vlg",
        "outputId": "894261b4-a9f8-409c-b6d9-b448a2d39199"
      },
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "----------------------------------------------------------------\n",
            "        Layer (type)               Output Shape         Param #\n",
            "================================================================\n",
            "            Conv2d-1           [-1, 32, 64, 64]           1,472\n",
            "              ReLU-2           [-1, 32, 64, 64]               0\n",
            "            Conv2d-3           [-1, 64, 32, 32]          18,496\n",
            "              ReLU-4           [-1, 64, 32, 32]               0\n",
            "            Conv2d-5          [-1, 128, 16, 16]          73,856\n",
            "              ReLU-6          [-1, 128, 16, 16]               0\n",
            "            Conv2d-7            [-1, 256, 8, 8]         295,168\n",
            "              ReLU-8            [-1, 256, 8, 8]               0\n",
            "   ConvTranspose2d-9          [-1, 128, 16, 16]         295,040\n",
            "             ReLU-10          [-1, 128, 16, 16]               0\n",
            "  ConvTranspose2d-11           [-1, 64, 32, 32]          73,792\n",
            "             ReLU-12           [-1, 64, 32, 32]               0\n",
            "  ConvTranspose2d-13           [-1, 32, 64, 64]          18,464\n",
            "             ReLU-14           [-1, 32, 64, 64]               0\n",
            "  ConvTranspose2d-15            [-1, 1, 64, 64]             289\n",
            "          Sigmoid-16            [-1, 1, 64, 64]               0\n",
            "================================================================\n",
            "Total params: 776,577\n",
            "Trainable params: 776,577\n",
            "Non-trainable params: 0\n",
            "----------------------------------------------------------------\n",
            "Input size (MB): 0.08\n",
            "Forward/backward pass size (MB): 7.31\n",
            "Params size (MB): 2.96\n",
            "Estimated Total Size (MB): 10.35\n",
            "----------------------------------------------------------------\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "def trainer(model,trainLoader,validationLoader,epochs,lossFn,optimizer,trainLoss,validLoss,device,batchSize):\n",
        "  print(\"Training has begun.....\")\n",
        "  startTime = time.time()\n",
        "  model = model.to(device)\n",
        "  for epoch in range(epochs):\n",
        "    runningLoss = 0.0\n",
        "    for x, y in trainLoader:\n",
        "      x = x.to(device)\n",
        "      y = y.to(device)\n",
        "      optimizer.zero_grad()\n",
        "      pred = model(x)\n",
        "      loss = lossFn(pred,y)\n",
        "      loss.backward()\n",
        "      optimizer.step()\n",
        "      runningLoss += loss.item()\n",
        "    trainLoss.append(runningLoss/(len(trainLoader)/batchSize))\n",
        "\n",
        "    with torch.no_grad():\n",
        "      valLoss = 0.0\n",
        "      for x, y in validationLoader:\n",
        "        x = x.to(device)\n",
        "        y = y.to(device)\n",
        "        pred = model(x)\n",
        "        loss = lossFn(pred,y)\n",
        "        valLoss += loss.item()\n",
        "      validLoss.append(valLoss/(len(validationLoader)/batchSize))\n",
        "\n",
        "    print(f\"Epoch {epoch+1}/{epochs}, Training Loss: {runningLoss/len(trainLoader)}, Validation Loss: {valLoss/len(validationLoader)}\")\n",
        "  print(f\"Training is complete......\")\n",
        "  print(\"--- %s seconds ---\" % (time.time() - startTime))\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "      \n"
      ],
      "metadata": {
        "id": "yiWvAmOsXTQ2"
      },
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "criterion = nn.BCELoss()\n",
        "optimizer = optim.Adam(autoenc.parameters(), lr=0.001)\n",
        "batchSize = 1\n",
        "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
        "eps = 50\n",
        "trainArr = []\n",
        "validArr = []"
      ],
      "metadata": {
        "id": "y3mUQ_3PbUWI"
      },
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "if device.type == 'cuda':\n",
        "    print(torch.cuda.get_device_name(0))\n",
        "    print('Memory Usage:')\n",
        "    print('Allocated:', round(torch.cuda.memory_allocated(0)/1024**3,1), 'GB')\n",
        "    print('Cached:   ', round(torch.cuda.memory_reserved(0)/1024**3,1), 'GB')"
      ],
      "metadata": {
        "id": "XYmsbh0sbpFR",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2fa57ee3-b371-4cfb-89da-bc3f5b33262e"
      },
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Tesla T4\n",
            "Memory Usage:\n",
            "Allocated: 0.0 GB\n",
            "Cached:    0.0 GB\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "trainer(autoenc,trainLoader,validationLoader,eps,criterion,optimizer,trainArr,validArr,device,batchSize)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uxlNxeGVcdIH",
        "outputId": "45f3009a-2657-489b-a4f8-4fc72d680b5d"
      },
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Training has begun.....\n",
            "Epoch 1/50, Training Loss: 0.5348944054118224, Validation Loss: 0.5269723996520043\n",
            "Epoch 2/50, Training Loss: 0.5348944053053856, Validation Loss: 0.5269724001938646\n",
            "Epoch 3/50, Training Loss: 0.5348944056246961, Validation Loss: 0.5269724004647949\n",
            "Epoch 4/50, Training Loss: 0.5348944055714777, Validation Loss: 0.5269723999229344\n",
            "Epoch 5/50, Training Loss: 0.5348944051457303, Validation Loss: 0.5269724000583995\n",
            "Epoch 6/50, Training Loss: 0.5348944056779146, Validation Loss: 0.5269723997874693\n",
            "Epoch 7/50, Training Loss: 0.5348944054650409, Validation Loss: 0.5269723999229344\n",
            "Epoch 8/50, Training Loss: 0.5348944056246961, Validation Loss: 0.5269724003293298\n",
            "Epoch 9/50, Training Loss: 0.5348944054650409, Validation Loss: 0.5269724003293298\n",
            "Epoch 10/50, Training Loss: 0.5348944054118224, Validation Loss: 0.5269724003293298\n",
            "Epoch 11/50, Training Loss: 0.534894405358604, Validation Loss: 0.5269724003293298\n",
            "Epoch 12/50, Training Loss: 0.5348944054650409, Validation Loss: 0.5269724003293298\n",
            "Epoch 13/50, Training Loss: 0.5348944056246961, Validation Loss: 0.5269724004647949\n",
            "Epoch 14/50, Training Loss: 0.5348944055714777, Validation Loss: 0.5269724004647949\n",
            "Epoch 15/50, Training Loss: 0.5348944054118224, Validation Loss: 0.52697240060026\n",
            "Epoch 16/50, Training Loss: 0.534894405358604, Validation Loss: 0.5269724003293298\n",
            "Epoch 17/50, Training Loss: 0.5348944052521671, Validation Loss: 0.5269724001938646\n",
            "Epoch 18/50, Training Loss: 0.5348944056779146, Validation Loss: 0.5269724004647949\n",
            "Epoch 19/50, Training Loss: 0.5348944055714777, Validation Loss: 0.5269724003293298\n",
            "Epoch 20/50, Training Loss: 0.534894405358604, Validation Loss: 0.5269724001938646\n",
            "Epoch 21/50, Training Loss: 0.5348944052521671, Validation Loss: 0.5269724004647949\n",
            "Epoch 22/50, Training Loss: 0.5348944054118224, Validation Loss: 0.5269724001938646\n",
            "Epoch 23/50, Training Loss: 0.5348944054650409, Validation Loss: 0.5269723999229344\n",
            "Epoch 24/50, Training Loss: 0.5348944056246961, Validation Loss: 0.52697240060026\n",
            "Epoch 25/50, Training Loss: 0.5348944054650409, Validation Loss: 0.5269724001938646\n",
            "Epoch 26/50, Training Loss: 0.534894405358604, Validation Loss: 0.5269724003293298\n",
            "Epoch 27/50, Training Loss: 0.5348944051989487, Validation Loss: 0.5269724000583995\n",
            "Epoch 28/50, Training Loss: 0.5348944055182593, Validation Loss: 0.5269724003293298\n",
            "Epoch 29/50, Training Loss: 0.534894404986075, Validation Loss: 0.5269724004647949\n",
            "Epoch 30/50, Training Loss: 0.5348944054118224, Validation Loss: 0.5269724001938646\n",
            "Epoch 31/50, Training Loss: 0.5348944052521671, Validation Loss: 0.5269724000583995\n",
            "Epoch 32/50, Training Loss: 0.5348944054650409, Validation Loss: 0.5269724001938646\n",
            "Epoch 33/50, Training Loss: 0.5348944051989487, Validation Loss: 0.5269723997874693\n",
            "Epoch 34/50, Training Loss: 0.5348944054650409, Validation Loss: 0.5269724001938646\n",
            "Epoch 35/50, Training Loss: 0.5348944056779146, Validation Loss: 0.5269723997874693\n",
            "Epoch 36/50, Training Loss: 0.5348944055182593, Validation Loss: 0.5269723999229344\n",
            "Epoch 37/50, Training Loss: 0.5348944054118224, Validation Loss: 0.5269724003293298\n",
            "Epoch 38/50, Training Loss: 0.5348944050925118, Validation Loss: 0.5269724004647949\n",
            "Epoch 39/50, Training Loss: 0.5348944051989487, Validation Loss: 0.5269724004647949\n",
            "Epoch 40/50, Training Loss: 0.5348944054118224, Validation Loss: 0.5269724004647949\n",
            "Epoch 41/50, Training Loss: 0.5348944052521671, Validation Loss: 0.5269724003293298\n",
            "Epoch 42/50, Training Loss: 0.5348944054650409, Validation Loss: 0.5269724001938646\n",
            "Epoch 43/50, Training Loss: 0.5348944054118224, Validation Loss: 0.5269723999229344\n",
            "Epoch 44/50, Training Loss: 0.5348944051457303, Validation Loss: 0.5269724000583995\n",
            "Epoch 45/50, Training Loss: 0.534894405358604, Validation Loss: 0.5269724001938646\n",
            "Epoch 46/50, Training Loss: 0.534894405358604, Validation Loss: 0.5269724001938646\n",
            "Epoch 47/50, Training Loss: 0.5348944054650409, Validation Loss: 0.5269724001938646\n",
            "Epoch 48/50, Training Loss: 0.5348944054118224, Validation Loss: 0.5269724000583995\n",
            "Epoch 49/50, Training Loss: 0.534894405731133, Validation Loss: 0.5269724001938646\n",
            "Epoch 50/50, Training Loss: 0.534894405358604, Validation Loss: 0.5269724003293298\n",
            "Training is complete......\n",
            "--- 481.83296179771423 seconds ---\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "plt.plot(trainArr, label='training loss')\n",
        "plt.plot(validArr, label='validation loss')\n",
        "plt.title('Loss at the end of each epoch')\n",
        "plt.legend();"
      ],
      "metadata": {
        "id": "lS3ob3qicuAC",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 281
        },
        "outputId": "8d0f20ea-bc24-4f5f-a2fd-7642484708f9"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "4wlsFGCDhg87"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}