Train_Segmentation_Model_FPN+xception.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "single-admission",
   "metadata": {},
   "source": [
    "# IMPORTS\n",
    "\n",
    "First we started with some python imports. For model building we used the <mark>**segmentation_models_pytorch library**</mark>. This has library has a lot segmentation model architechtures. Before you can import the library it needs to be ensured that the library is installed in the conda environment. For this in the anaconda terminal just type the below command to install the library:-\n",
    "\n",
    "<div class=\"alert alert-block alert-success\">\n",
    "<b>Conda Command:</b> conda install -q segmentation-models-pytorch\n",
    "</div>\n",
    "\n",
    "Once, the libray is installed it can be imported in the script safely. We also imported the utility script Trainer for model training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "answering-redhead",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import segmentation_models_pytorch as smp\n",
    "from Utilities.Trainer import Trainer\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "from torchvision import models\n",
    "import torch.multiprocessing as mp\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "resistant-civilian",
   "metadata": {},
   "source": [
    "The below statement helps in better error tracking when training is done in cuda. More details about this can be found in the below url:-\n",
    "\n",
    "https://pytorch.org/docs/stable/notes/cuda.html"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "passive-hampton",
   "metadata": {},
   "outputs": [],
   "source": [
    "CUDA_LAUNCH_BLOCKING=\"1\""
   ]
  },
  {
   "cell_type": "markdown",
   "id": "strong-static",
   "metadata": {},
   "source": [
    "# Model Building\n",
    "\n",
    "Next a FPN model with inceptionv3 encoder was created with imagenet weights from the smp library."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "gross-drama",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "FPN(\n",
      "  (encoder): XceptionEncoder(\n",
      "    (conv1): Conv2d(3, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "    (bn1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (relu): ReLU(inplace=True)\n",
      "    (conv2): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "    (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (block1): Block(\n",
      "      (skip): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "      (skipbn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): SeparableConv2d(\n",
      "          (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)\n",
      "          (pointwise): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (2): ReLU(inplace=True)\n",
      "        (3): SeparableConv2d(\n",
      "          (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)\n",
      "          (pointwise): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (4): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (5): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "      )\n",
      "    )\n",
      "    (block2): Block(\n",
      "      (skip): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "      (skipbn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): ReLU()\n",
      "        (1): SeparableConv2d(\n",
      "          (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)\n",
      "          (pointwise): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (3): ReLU(inplace=True)\n",
      "        (4): SeparableConv2d(\n",
      "          (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256, bias=False)\n",
      "          (pointwise): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (5): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (6): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "      )\n",
      "    )\n",
      "    (block3): Block(\n",
      "      (skip): Conv2d(256, 728, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "      (skipbn): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): ReLU()\n",
      "        (1): SeparableConv2d(\n",
      "          (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256, bias=False)\n",
      "          (pointwise): Conv2d(256, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (2): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (3): ReLU(inplace=True)\n",
      "        (4): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (5): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (6): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "      )\n",
      "    )\n",
      "    (block4): Block(\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): ReLU()\n",
      "        (1): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (2): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (3): ReLU(inplace=True)\n",
      "        (4): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (5): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (6): ReLU(inplace=True)\n",
      "        (7): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (8): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (block5): Block(\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): ReLU()\n",
      "        (1): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (2): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (3): ReLU(inplace=True)\n",
      "        (4): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (5): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (6): ReLU(inplace=True)\n",
      "        (7): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (8): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (block6): Block(\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): ReLU()\n",
      "        (1): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (2): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (3): ReLU(inplace=True)\n",
      "        (4): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (5): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (6): ReLU(inplace=True)\n",
      "        (7): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (8): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (block7): Block(\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): ReLU()\n",
      "        (1): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (2): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (3): ReLU(inplace=True)\n",
      "        (4): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (5): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (6): ReLU(inplace=True)\n",
      "        (7): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (8): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (block8): Block(\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): ReLU()\n",
      "        (1): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (2): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (3): ReLU(inplace=True)\n",
      "        (4): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (5): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (6): ReLU(inplace=True)\n",
      "        (7): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (8): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (block9): Block(\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): ReLU()\n",
      "        (1): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (2): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (3): ReLU(inplace=True)\n",
      "        (4): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (5): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (6): ReLU(inplace=True)\n",
      "        (7): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (8): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (block10): Block(\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): ReLU()\n",
      "        (1): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (2): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (3): ReLU(inplace=True)\n",
      "        (4): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (5): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (6): ReLU(inplace=True)\n",
      "        (7): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (8): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (block11): Block(\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): ReLU()\n",
      "        (1): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (2): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (3): ReLU(inplace=True)\n",
      "        (4): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (5): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (6): ReLU(inplace=True)\n",
      "        (7): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (8): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (block12): Block(\n",
      "      (skip): Conv2d(728, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "      (skipbn): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace=True)\n",
      "      (rep): Sequential(\n",
      "        (0): ReLU()\n",
      "        (1): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 728, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (2): BatchNorm2d(728, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (3): ReLU(inplace=True)\n",
      "        (4): SeparableConv2d(\n",
      "          (conv1): Conv2d(728, 728, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=728, bias=False)\n",
      "          (pointwise): Conv2d(728, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        )\n",
      "        (5): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (6): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "      )\n",
      "    )\n",
      "    (conv3): SeparableConv2d(\n",
      "      (conv1): Conv2d(1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1024, bias=False)\n",
      "      (pointwise): Conv2d(1024, 1536, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "    )\n",
      "    (bn3): BatchNorm2d(1536, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (conv4): SeparableConv2d(\n",
      "      (conv1): Conv2d(1536, 1536, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1536, bias=False)\n",
      "      (pointwise): Conv2d(1536, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "    )\n",
      "    (bn4): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "  )\n",
      "  (decoder): FPNDecoder(\n",
      "    (p5): Conv2d(2048, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "    (p4): FPNBlock(\n",
      "      (skip_conv): Conv2d(728, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (p3): FPNBlock(\n",
      "      (skip_conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (p2): FPNBlock(\n",
      "      (skip_conv): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (seg_blocks): ModuleList(\n",
      "      (0): SegmentationBlock(\n",
      "        (block): Sequential(\n",
      "          (0): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "          (1): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "          (2): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (1): SegmentationBlock(\n",
      "        (block): Sequential(\n",
      "          (0): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "          (1): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (2): SegmentationBlock(\n",
      "        (block): Sequential(\n",
      "          (0): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (3): SegmentationBlock(\n",
      "        (block): Sequential(\n",
      "          (0): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "    )\n",
      "    (merge): MergeBlock()\n",
      "    (dropout): Dropout2d(p=0.2, inplace=True)\n",
      "  )\n",
      "  (segmentation_head): SegmentationHead(\n",
      "    (0): Conv2d(128, 6, kernel_size=(1, 1), stride=(1, 1))\n",
      "    (1): UpsamplingBilinear2d(scale_factor=4.0, mode=bilinear)\n",
      "    (2): Activation(\n",
      "      (activation): Identity()\n",
      "    )\n",
      "  )\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "torch.manual_seed(42)\n",
    "ENCODER = 'xception'\n",
    "ENCODER_WEIGHTS = 'imagenet'\n",
    "CLASSES = 6\n",
    "ACTIVATION = None # could be None for logits or 'softmax2d' for multiclass segmentation\n",
    "\n",
    "# create segmentation model with pretrained encoder\n",
    "model = smp.FPN(\n",
    "    encoder_name=ENCODER, \n",
    "    encoder_weights=ENCODER_WEIGHTS, \n",
    "    in_channels=3,\n",
    "    classes=CLASSES, \n",
    "    activation=ACTIVATION,\n",
    ")\n",
    "print(model)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "alike-password",
   "metadata": {},
   "source": [
    "# Model Training\n",
    "\n",
    "Next the model training was started with 30 epochs and initial learning rate as 5e-04. The model checkpoint was saved whenever the validation dice was reported the greater than the previous best validation dice.\n",
    "\n",
    "In the below results you can see the training starting from epoch 6. This is because the kernel was initally interuptted after epoch 5. So, I had to start all over again. Since, checkpoint was saved upto epoch number 5 so it started from epoch 6."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "double-diagnosis",
   "metadata": {},
   "outputs": [],
   "source": [
    "lr = 5e-04\n",
    "epochs = 20\n",
    "path = \"/home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/Models/model_fpn_xception.pth\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "supreme-circuit",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Learning Rate =  0.0005\n",
      "Starting epoch: 0 | phase: train | ⏰: 23:25:37\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:19<00:00,  1.86s/it, loss=0.0296]\n",
      "/home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/Utilities/Meter.py:40: RuntimeWarning: Mean of empty slice\n",
      "  dice_neg = np.nanmean(self.dice_neg_scores)\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.2371 | IoU: 0.1903 | dice: 0.2562 | dice_neg: nan | dice_pos: 0.2562\n",
      "Starting epoch: 0 | phase: val | ⏰: 23:35:57\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:57<00:00,  2.58it/s, loss=0.0105]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0837 | IoU: 0.4720 | dice: 0.5928 | dice_neg: nan | dice_pos: 0.5928\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 1 | phase: train | ⏰: 23:36:55\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:17<00:00,  1.90s/it, loss=0.0161]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.1291 | IoU: 0.4112 | dice: 0.5288 | dice_neg: nan | dice_pos: 0.5288\n",
      "Starting epoch: 1 | phase: val | ⏰: 23:47:13\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:58<00:00,  2.88it/s, loss=0.00906]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0725 | IoU: 0.5589 | dice: 0.6904 | dice_neg: nan | dice_pos: 0.6904\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 2 | phase: train | ⏰: 23:48:12\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:16<00:00,  1.88s/it, loss=0.0123]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0986 | IoU: 0.4944 | dice: 0.6167 | dice_neg: nan | dice_pos: 0.6167\n",
      "Starting epoch: 2 | phase: val | ⏰: 23:58:29\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:56<00:00,  3.14it/s, loss=0.0089] \n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0712 | IoU: 0.5650 | dice: 0.6876 | dice_neg: nan | dice_pos: 0.6876\n",
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 3 | phase: train | ⏰: 23:59:26\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:16<00:00,  1.88s/it, loss=0.0106]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0846 | IoU: 0.5470 | dice: 0.6711 | dice_neg: nan | dice_pos: 0.6711\n",
      "Starting epoch: 3 | phase: val | ⏰: 00:09:42\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:55<00:00,  2.76it/s, loss=0.00802]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0642 | IoU: 0.6046 | dice: 0.7301 | dice_neg: nan | dice_pos: 0.7301\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 4 | phase: train | ⏰: 00:10:39\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:15<00:00,  1.88s/it, loss=0.0107]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0860 | IoU: 0.5570 | dice: 0.6820 | dice_neg: nan | dice_pos: 0.6820\n",
      "Starting epoch: 4 | phase: val | ⏰: 00:20:55\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:56<00:00,  3.07it/s, loss=0.00957]\n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0766 | IoU: 0.5763 | dice: 0.6991 | dice_neg: nan | dice_pos: 0.6991\n",
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 5 | phase: train | ⏰: 00:21:51\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:14<00:00,  1.88s/it, loss=0.00978]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0782 | IoU: 0.5720 | dice: 0.6971 | dice_neg: nan | dice_pos: 0.6971\n",
      "Starting epoch: 5 | phase: val | ⏰: 00:32:06\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:58<00:00,  2.80it/s, loss=0.00764]\n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0612 | IoU: 0.5973 | dice: 0.7191 | dice_neg: nan | dice_pos: 0.7191\n",
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 6 | phase: train | ⏰: 00:33:04\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:15<00:00,  1.88s/it, loss=0.0086] \n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0688 | IoU: 0.6065 | dice: 0.7301 | dice_neg: nan | dice_pos: 0.7301\n",
      "Starting epoch: 6 | phase: val | ⏰: 00:43:20\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:55<00:00,  2.62it/s, loss=0.00743]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0594 | IoU: 0.6603 | dice: 0.7808 | dice_neg: nan | dice_pos: 0.7808\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 7 | phase: train | ⏰: 00:44:16\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:14<00:00,  1.88s/it, loss=0.0079] \n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0632 | IoU: 0.6322 | dice: 0.7530 | dice_neg: nan | dice_pos: 0.7530\n",
      "Starting epoch: 7 | phase: val | ⏰: 00:54:31\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:56<00:00,  2.43it/s, loss=0.0075] \n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0600 | IoU: 0.5936 | dice: 0.7170 | dice_neg: nan | dice_pos: 0.7170\n",
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 8 | phase: train | ⏰: 00:55:28\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:14<00:00,  1.88s/it, loss=0.00742]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0594 | IoU: 0.6515 | dice: 0.7715 | dice_neg: nan | dice_pos: 0.7715\n",
      "Starting epoch: 8 | phase: val | ⏰: 01:05:42\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:56<00:00,  2.65it/s, loss=0.00782]\n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0626 | IoU: 0.6142 | dice: 0.7367 | dice_neg: nan | dice_pos: 0.7367\n",
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 9 | phase: train | ⏰: 01:06:39\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:14<00:00,  1.87s/it, loss=0.00714]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0571 | IoU: 0.6586 | dice: 0.7757 | dice_neg: nan | dice_pos: 0.7757\n",
      "Starting epoch: 9 | phase: val | ⏰: 01:16:54\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:57<00:00,  2.91it/s, loss=0.00713]\n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0570 | IoU: 0.6493 | dice: 0.7685 | dice_neg: nan | dice_pos: 0.7685\n",
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 10 | phase: train | ⏰: 01:17:52\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:13<00:00,  1.87s/it, loss=0.00705]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0564 | IoU: 0.6655 | dice: 0.7819 | dice_neg: nan | dice_pos: 0.7819\n",
      "Starting epoch: 10 | phase: val | ⏰: 01:28:05\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [01:00<00:00,  2.55it/s, loss=0.00724]\n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0579 | IoU: 0.6529 | dice: 0.7743 | dice_neg: nan | dice_pos: 0.7743\n",
      "Epoch    11: reducing learning rate of group 0 to 5.0000e-05.\n",
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 11 | phase: train | ⏰: 01:29:06\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:12<00:00,  1.87s/it, loss=0.00641]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0513 | IoU: 0.6867 | dice: 0.7993 | dice_neg: nan | dice_pos: 0.7993\n",
      "Starting epoch: 11 | phase: val | ⏰: 01:39:18\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:55<00:00,  3.28it/s, loss=0.00684]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0548 | IoU: 0.6690 | dice: 0.7864 | dice_neg: nan | dice_pos: 0.7864\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 12 | phase: train | ⏰: 01:40:15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:11<00:00,  1.87s/it, loss=0.00634]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0507 | IoU: 0.6987 | dice: 0.8096 | dice_neg: nan | dice_pos: 0.8096\n",
      "Starting epoch: 12 | phase: val | ⏰: 01:50:27\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [01:00<00:00,  3.16it/s, loss=0.00682]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0546 | IoU: 0.6730 | dice: 0.7897 | dice_neg: nan | dice_pos: 0.7897\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 13 | phase: train | ⏰: 01:51:28\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:12<00:00,  1.87s/it, loss=0.00594]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0475 | IoU: 0.7095 | dice: 0.8177 | dice_neg: nan | dice_pos: 0.8177\n",
      "Starting epoch: 13 | phase: val | ⏰: 02:01:41\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:59<00:00,  2.44it/s, loss=0.00671]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0537 | IoU: 0.6755 | dice: 0.7919 | dice_neg: nan | dice_pos: 0.7919\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 14 | phase: train | ⏰: 02:02:41\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:11<00:00,  1.87s/it, loss=0.0058] \n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0464 | IoU: 0.7125 | dice: 0.8204 | dice_neg: nan | dice_pos: 0.8204\n",
      "Starting epoch: 14 | phase: val | ⏰: 02:12:53\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:56<00:00,  2.48it/s, loss=0.007]  \n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0560 | IoU: 0.6568 | dice: 0.7739 | dice_neg: nan | dice_pos: 0.7739\n",
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 15 | phase: train | ⏰: 02:13:49\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:11<00:00,  1.87s/it, loss=0.00552]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0441 | IoU: 0.7178 | dice: 0.8236 | dice_neg: nan | dice_pos: 0.8236\n",
      "Starting epoch: 15 | phase: val | ⏰: 02:24:01\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:53<00:00,  2.54it/s, loss=0.00691]\n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0553 | IoU: 0.6599 | dice: 0.7766 | dice_neg: nan | dice_pos: 0.7766\n",
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 16 | phase: train | ⏰: 02:24:55\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:11<00:00,  1.86s/it, loss=0.00535]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0428 | IoU: 0.7274 | dice: 0.8332 | dice_neg: nan | dice_pos: 0.8332\n",
      "Starting epoch: 16 | phase: val | ⏰: 02:35:07\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:56<00:00,  2.66it/s, loss=0.00702]\n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0562 | IoU: 0.6585 | dice: 0.7759 | dice_neg: nan | dice_pos: 0.7759\n",
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 17 | phase: train | ⏰: 02:36:03\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:12<00:00,  1.87s/it, loss=0.00543]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0434 | IoU: 0.7249 | dice: 0.8305 | dice_neg: nan | dice_pos: 0.8305\n",
      "Starting epoch: 17 | phase: val | ⏰: 02:46:16\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:56<00:00,  2.39it/s, loss=0.00716]\n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0573 | IoU: 0.6507 | dice: 0.7687 | dice_neg: nan | dice_pos: 0.7687\n",
      "Epoch    18: reducing learning rate of group 0 to 5.0000e-06.\n",
      "\n",
      "Learning Rate =  5e-06\n",
      "Starting epoch: 18 | phase: train | ⏰: 02:47:13\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:12<00:00,  1.87s/it, loss=0.00568]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0454 | IoU: 0.7234 | dice: 0.8285 | dice_neg: nan | dice_pos: 0.8285\n",
      "Starting epoch: 18 | phase: val | ⏰: 02:57:26\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:57<00:00,  2.59it/s, loss=0.00721]\n",
      "  0%|          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0576 | IoU: 0.6513 | dice: 0.7685 | dice_neg: nan | dice_pos: 0.7685\n",
      "\n",
      "Learning Rate =  5e-06\n",
      "Starting epoch: 19 | phase: train | ⏰: 02:58:24\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 310/310 [10:13<00:00,  1.87s/it, loss=0.00537]\n",
      "  0%|          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0429 | IoU: 0.7235 | dice: 0.8283 | dice_neg: nan | dice_pos: 0.8283\n",
      "Starting epoch: 19 | phase: val | ⏰: 03:08:37\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 133/133 [00:59<00:00,  2.31it/s, loss=0.00707]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0566 | IoU: 0.6621 | dice: 0.7789 | dice_neg: nan | dice_pos: 0.7789\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "model_trainer = Trainer(model, lr, epochs, path)\n",
    "model_trainer.start()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "moving-enlargement",
   "metadata": {},
   "source": [
    "After the model training was done the loss, dice_scores and iou_scores were plotted. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "opponent-polyester",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "losses = model_trainer.losses\n",
    "dice_scores = model_trainer.dice_scores # overall dice\n",
    "iou_scores = model_trainer.iou_scores\n",
    "\n",
    "def plot(scores, name):\n",
    "    plt.figure(figsize=(15,5))\n",
    "    plt.plot(range(len(scores[\"train\"])), scores[\"train\"], label=f'train {name}')\n",
    "    plt.plot(range(len(scores[\"train\"])), scores[\"val\"], label=f'val {name}')\n",
    "    plt.title(f'{name} plot'); plt.xlabel('Epoch'); plt.ylabel(f'{name}');\n",
    "    plt.legend(); \n",
    "    plt.show()\n",
    "\n",
    "plot(losses, \"BCE loss\")\n",
    "plot(dice_scores, \"Dice score\")\n",
    "plot(iou_scores, \"IoU score\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "general-status",
   "metadata": {},
   "source": [
    "The above results the overall training and validation dice and IoU goes up with epochs and training and validation loss goes down with epochs."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "varying-digit",
   "metadata": {},
   "source": [
    "# Convert to ONNX\n",
    "\n",
    "The trained model is converted to ONNX format using the Utility script ONNX_Converter.py."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "based-eight",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model name :FPN\n",
      "Model checkpoint path is :/home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/Models/model_fpn_xception.pth\n",
      "Model output path is :/home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/ONNX_models/fpn_xception.onnx\n",
      "Successfully converted the model to onnx format.\n"
     ]
    }
   ],
   "source": [
    "! python3 ./Utilities/ONNX_converter.py FPN xception  imagenet  6 /home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/Models/model_fpn_xception.pth /home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/ONNX_models/fpn_xception.onnx"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "hindu-merchant",
   "metadata": {},
   "source": [
    "# Convert to OpenVINO format\n",
    "\n",
    "Lastly the trained Onnx model was converted to OpenVINO format for inferencing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "deluxe-poland",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model Optimizer arguments:\n",
      "Common parameters:\n",
      "\t- Path to the Input Model: \t/home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/ONNX_models/fpn_xception.onnx\n",
      "\t- Path for generated IR: \t/home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/OpenVINO_models\n",
      "\t- IR output name: \tfpn_xception\n",
      "\t- Log level: \tERROR\n",
      "\t- Batch: \tNot specified, inherited from the model\n",
      "\t- Input layers: \tNot specified, inherited from the model\n",
      "\t- Output layers: \tNot specified, inherited from the model\n",
      "\t- Input shapes: \tNot specified, inherited from the model\n",
      "\t- Mean values: \tNot specified\n",
      "\t- Scale values: \tNot specified\n",
      "\t- Scale factor: \tNot specified\n",
      "\t- Precision of IR: \tFP32\n",
      "\t- Enable fusing: \tTrue\n",
      "\t- Enable grouped convolutions fusing: \tTrue\n",
      "\t- Move mean values to preprocess section: \tNone\n",
      "\t- Reverse input channels: \tFalse\n",
      "ONNX specific parameters:\n",
      "\t- Inference Engine found in: \t/opt/intel/openvino/python/python3.6/openvino\n",
      "Inference Engine version: \t2.1.2021.3.0-2787-60059f2c755-releases/2021/3\n",
      "Model Optimizer version: \t    2021.3.0-2787-60059f2c755-releases/2021/3\n",
      "[ SUCCESS ] Generated IR version 10 model.\n",
      "[ SUCCESS ] XML file: /home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/OpenVINO_models/fpn_xception.xml\n",
      "[ SUCCESS ] BIN file: /home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/OpenVINO_models/fpn_xception.bin\n",
      "[ SUCCESS ] Total execution time: 31.99 seconds. \n",
      "[ SUCCESS ] Memory consumed: 740 MB. \n"
     ]
    }
   ],
   "source": [
    "! mo.py --input_model /home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/ONNX_models/fpn_xception.onnx --output_dir /home/u71926/My-Notebooks/NEU-SURFACE-DEFECT/OpenVINO_models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "buried-certification",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (OpenVINO 2021.3)",
   "language": "python",
   "name": "c003-python_3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}