{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "a2220df6",
   "metadata": {},
   "source": [
    "# Import Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "7249bea4",
   "metadata": {},
   "outputs": [],
   "source": [
    "import gradio as gr\n",
    "import torch\n",
    "import torch.nn.functional as F\n",
    "from facenet_pytorch import MTCNN, InceptionResnetV1\n",
    "import numpy as np\n",
    "from PIL import Image\n",
    "import cv2\n",
    "from pytorch_grad_cam import GradCAM\n",
    "from pytorch_grad_cam.utils.model_targets import ClassifierOutputTarget\n",
    "from pytorch_grad_cam.utils.image import show_cam_on_image\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "62f0492b-aad6-4464-ab96-1365b7f3a44e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: gradio in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (3.39.0)\n",
      "Collecting gradio\n",
      "  Downloading gradio-4.19.1-py3-none-any.whl.metadata (15 kB)\n",
      "Requirement already satisfied: aiofiles<24.0,>=22.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (23.2.1)\n",
      "Requirement already satisfied: altair<6.0,>=4.2.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (5.2.0)\n",
      "Requirement already satisfied: fastapi in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (0.109.2)\n",
      "Requirement already satisfied: ffmpy in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (0.3.2)\n",
      "Requirement already satisfied: gradio-client==0.10.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (0.10.0)\n",
      "Requirement already satisfied: httpx in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (0.26.0)\n",
      "Requirement already satisfied: huggingface-hub>=0.19.3 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (0.20.3)\n",
      "Requirement already satisfied: importlib-resources<7.0,>=1.3 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (6.1.1)\n",
      "Requirement already satisfied: jinja2<4.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (3.1.3)\n",
      "Requirement already satisfied: markupsafe~=2.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (2.1.5)\n",
      "Requirement already satisfied: matplotlib~=3.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (3.8.3)\n",
      "Requirement already satisfied: numpy~=1.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (1.26.4)\n",
      "Requirement already satisfied: orjson~=3.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (3.9.14)\n",
      "Requirement already satisfied: packaging in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (23.2)\n",
      "Requirement already satisfied: pandas<3.0,>=1.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (2.2.0)\n",
      "Requirement already satisfied: pillow<11.0,>=8.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (9.4.0)\n",
      "Requirement already satisfied: pydantic>=2.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (2.6.1)\n",
      "Requirement already satisfied: pydub in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (0.25.1)\n",
      "Requirement already satisfied: python-multipart>=0.0.9 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (0.0.9)\n",
      "Requirement already satisfied: pyyaml<7.0,>=5.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (6.0.1)\n",
      "Collecting ruff>=0.1.7 (from gradio)\n",
      "  Downloading ruff-0.2.1-py3-none-win_amd64.whl.metadata (23 kB)\n",
      "Requirement already satisfied: semantic-version~=2.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (2.10.0)\n",
      "Collecting tomlkit==0.12.0 (from gradio)\n",
      "  Downloading tomlkit-0.12.0-py3-none-any.whl.metadata (2.7 kB)\n",
      "Collecting typer<1.0,>=0.9 (from typer[all]<1.0,>=0.9->gradio)\n",
      "  Downloading typer-0.9.0-py3-none-any.whl.metadata (14 kB)\n",
      "Requirement already satisfied: typing-extensions~=4.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (4.9.0)\n",
      "Requirement already satisfied: uvicorn>=0.14.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio) (0.27.1)\n",
      "Requirement already satisfied: fsspec in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio-client==0.10.0->gradio) (2024.2.0)\n",
      "Requirement already satisfied: websockets<12.0,>=10.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from gradio-client==0.10.0->gradio) (11.0.3)\n",
      "Requirement already satisfied: jsonschema>=3.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from altair<6.0,>=4.2.0->gradio) (4.21.1)\n",
      "Requirement already satisfied: toolz in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from altair<6.0,>=4.2.0->gradio) (0.12.1)\n",
      "Requirement already satisfied: filelock in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from huggingface-hub>=0.19.3->gradio) (3.13.1)\n",
      "Requirement already satisfied: requests in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from huggingface-hub>=0.19.3->gradio) (2.31.0)\n",
      "Requirement already satisfied: tqdm>=4.42.1 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from huggingface-hub>=0.19.3->gradio) (4.66.2)\n",
      "Requirement already satisfied: zipp>=3.1.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from importlib-resources<7.0,>=1.3->gradio) (3.17.0)\n",
      "Requirement already satisfied: contourpy>=1.0.1 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from matplotlib~=3.0->gradio) (1.2.0)\n",
      "Requirement already satisfied: cycler>=0.10 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from matplotlib~=3.0->gradio) (0.12.1)\n",
      "Requirement already satisfied: fonttools>=4.22.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from matplotlib~=3.0->gradio) (4.49.0)\n",
      "Requirement already satisfied: kiwisolver>=1.3.1 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from matplotlib~=3.0->gradio) (1.4.5)\n",
      "Requirement already satisfied: pyparsing>=2.3.1 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from matplotlib~=3.0->gradio) (3.1.1)\n",
      "Requirement already satisfied: python-dateutil>=2.7 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from matplotlib~=3.0->gradio) (2.8.2)\n",
      "Requirement already satisfied: pytz>=2020.1 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from pandas<3.0,>=1.0->gradio) (2024.1)\n",
      "Requirement already satisfied: tzdata>=2022.7 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from pandas<3.0,>=1.0->gradio) (2024.1)\n",
      "Requirement already satisfied: annotated-types>=0.4.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from pydantic>=2.0->gradio) (0.6.0)\n",
      "Requirement already satisfied: pydantic-core==2.16.2 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from pydantic>=2.0->gradio) (2.16.2)\n",
      "Requirement already satisfied: click<9.0.0,>=7.1.1 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from typer<1.0,>=0.9->typer[all]<1.0,>=0.9->gradio) (8.1.7)\n",
      "Requirement already satisfied: colorama<0.5.0,>=0.4.3 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from typer[all]<1.0,>=0.9->gradio) (0.4.6)\n",
      "Collecting shellingham<2.0.0,>=1.3.0 (from typer[all]<1.0,>=0.9->gradio)\n",
      "  Downloading shellingham-1.5.4-py2.py3-none-any.whl.metadata (3.5 kB)\n",
      "Collecting rich<14.0.0,>=10.11.0 (from typer[all]<1.0,>=0.9->gradio)\n",
      "  Downloading rich-13.7.0-py3-none-any.whl.metadata (18 kB)\n",
      "Requirement already satisfied: h11>=0.8 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from uvicorn>=0.14.0->gradio) (0.14.0)\n",
      "Requirement already satisfied: starlette<0.37.0,>=0.36.3 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from fastapi->gradio) (0.36.3)\n",
      "Requirement already satisfied: anyio in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from httpx->gradio) (4.2.0)\n",
      "Requirement already satisfied: certifi in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from httpx->gradio) (2024.2.2)\n",
      "Requirement already satisfied: httpcore==1.* in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from httpx->gradio) (1.0.3)\n",
      "Requirement already satisfied: idna in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from httpx->gradio) (3.6)\n",
      "Requirement already satisfied: sniffio in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from httpx->gradio) (1.3.0)\n",
      "Requirement already satisfied: attrs>=22.2.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from jsonschema>=3.0->altair<6.0,>=4.2.0->gradio) (23.2.0)\n",
      "Requirement already satisfied: jsonschema-specifications>=2023.03.6 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from jsonschema>=3.0->altair<6.0,>=4.2.0->gradio) (2023.12.1)\n",
      "Requirement already satisfied: referencing>=0.28.4 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from jsonschema>=3.0->altair<6.0,>=4.2.0->gradio) (0.33.0)\n",
      "Requirement already satisfied: rpds-py>=0.7.1 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from jsonschema>=3.0->altair<6.0,>=4.2.0->gradio) (0.18.0)\n",
      "Requirement already satisfied: six>=1.5 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from python-dateutil>=2.7->matplotlib~=3.0->gradio) (1.16.0)\n",
      "Requirement already satisfied: markdown-it-py>=2.2.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from rich<14.0.0,>=10.11.0->typer[all]<1.0,>=0.9->gradio) (2.2.0)\n",
      "Requirement already satisfied: pygments<3.0.0,>=2.13.0 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from rich<14.0.0,>=10.11.0->typer[all]<1.0,>=0.9->gradio) (2.17.2)\n",
      "Requirement already satisfied: exceptiongroup>=1.0.2 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from anyio->httpx->gradio) (1.2.0)\n",
      "Requirement already satisfied: charset-normalizer<4,>=2 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from requests->huggingface-hub>=0.19.3->gradio) (3.3.2)\n",
      "Requirement already satisfied: urllib3<3,>=1.21.1 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from requests->huggingface-hub>=0.19.3->gradio) (2.2.0)\n",
      "Requirement already satisfied: mdurl~=0.1 in c:\\kandikits\\deepfake-detection\\deepfake-detection-env\\lib\\site-packages (from markdown-it-py>=2.2.0->rich<14.0.0,>=10.11.0->typer[all]<1.0,>=0.9->gradio) (0.1.2)\n",
      "Downloading gradio-4.19.1-py3-none-any.whl (16.9 MB)\n",
      "   ---------------------------------------- 0.0/16.9 MB ? eta -:--:--\n",
      "   ---------------------------------------- 0.0/16.9 MB 1.9 MB/s eta 0:00:09\n",
      "   ---------------------------------------- 0.2/16.9 MB 2.4 MB/s eta 0:00:08\n",
      "   - -------------------------------------- 0.5/16.9 MB 3.9 MB/s eta 0:00:05\n",
      "   -- ------------------------------------- 1.1/16.9 MB 6.1 MB/s eta 0:00:03\n",
      "   ----- ---------------------------------- 2.2/16.9 MB 10.2 MB/s eta 0:00:02\n",
      "   -------- ------------------------------- 3.7/16.9 MB 13.9 MB/s eta 0:00:01\n",
      "   ----------- ---------------------------- 5.1/16.9 MB 15.4 MB/s eta 0:00:01\n",
      "   --------------- ------------------------ 6.6/16.9 MB 18.4 MB/s eta 0:00:01\n",
      "   ------------------- -------------------- 8.4/16.9 MB 20.6 MB/s eta 0:00:01\n",
      "   --------------------- ------------------ 9.2/16.9 MB 21.1 MB/s eta 0:00:01\n",
      "   --------------------- ------------------ 9.2/16.9 MB 21.1 MB/s eta 0:00:01\n",
      "   --------------------- ------------------ 9.2/16.9 MB 21.1 MB/s eta 0:00:01\n",
      "   --------------------- ------------------ 9.3/16.9 MB 16.5 MB/s eta 0:00:01\n",
      "   ------------------------ --------------- 10.3/16.9 MB 17.2 MB/s eta 0:00:01\n",
      "   ------------------------ --------------- 10.5/16.9 MB 17.3 MB/s eta 0:00:01\n",
      "   -------------------------- ------------- 11.1/16.9 MB 18.7 MB/s eta 0:00:01\n",
      "   ---------------------------- ----------- 12.1/16.9 MB 17.7 MB/s eta 0:00:01\n",
      "   -------------------------------- ------- 13.6/16.9 MB 18.2 MB/s eta 0:00:01\n",
      "   ----------------------------------- ---- 14.9/16.9 MB 18.2 MB/s eta 0:00:01\n",
      "   ---------------------------------------  16.6/16.9 MB 18.2 MB/s eta 0:00:01\n",
      "   ---------------------------------------- 16.9/16.9 MB 16.8 MB/s eta 0:00:00\n",
      "Downloading tomlkit-0.12.0-py3-none-any.whl (37 kB)\n",
      "Downloading ruff-0.2.1-py3-none-win_amd64.whl (7.4 MB)\n",
      "   ---------------------------------------- 0.0/7.4 MB ? eta -:--:--\n",
      "   -------- ------------------------------- 1.6/7.4 MB 51.9 MB/s eta 0:00:01\n",
      "   ---------------- ----------------------- 3.1/7.4 MB 40.2 MB/s eta 0:00:01\n",
      "   ------------------------- -------------- 4.8/7.4 MB 37.9 MB/s eta 0:00:01\n",
      "   --------------------------------- ------ 6.3/7.4 MB 36.5 MB/s eta 0:00:01\n",
      "   ---------------------------------------  7.4/7.4 MB 33.7 MB/s eta 0:00:01\n",
      "   ---------------------------------------- 7.4/7.4 MB 31.6 MB/s eta 0:00:00\n",
      "Downloading typer-0.9.0-py3-none-any.whl (45 kB)\n",
      "   ---------------------------------------- 0.0/45.9 kB ? eta -:--:--\n",
      "   ---------------------------------------- 45.9/45.9 kB ? eta 0:00:00\n",
      "Downloading rich-13.7.0-py3-none-any.whl (240 kB)\n",
      "   ---------------------------------------- 0.0/240.6 kB ? eta -:--:--\n",
      "   --------------------------------------- 240.6/240.6 kB 14.4 MB/s eta 0:00:00\n",
      "Downloading shellingham-1.5.4-py2.py3-none-any.whl (9.8 kB)\n",
      "Installing collected packages: tomlkit, shellingham, ruff, typer, rich, gradio\n",
      "  Attempting uninstall: gradio\n",
      "    Found existing installation: gradio 3.39.0\n",
      "    Uninstalling gradio-3.39.0:\n",
      "      Successfully uninstalled gradio-3.39.0\n",
      "Successfully installed gradio-4.19.1 rich-13.7.0 ruff-0.2.1 shellingham-1.5.4 tomlkit-0.12.0 typer-0.9.0\n"
     ]
    }
   ],
   "source": [
    "!pip install -U gradio"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d25e1c5d",
   "metadata": {},
   "source": [
    "# Download and Load Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "237fbf44",
   "metadata": {},
   "outputs": [],
   "source": [
    "DEVICE = 'cuda:0' if torch.cuda.is_available() else 'cpu'\n",
    "\n",
    "mtcnn = MTCNN(\n",
    "    select_largest=False,\n",
    "    post_process=False,\n",
    "    device=DEVICE\n",
    ").to(DEVICE).eval()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "f3ef2b4f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "43131e0cdbdf44beb6f775f854ebbf07",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0.00/107M [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "InceptionResnetV1(\n",
       "  (conv2d_1a): BasicConv2d(\n",
       "    (conv): Conv2d(3, 32, kernel_size=(3, 3), stride=(2, 2), bias=False)\n",
       "    (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (relu): ReLU()\n",
       "  )\n",
       "  (conv2d_2a): BasicConv2d(\n",
       "    (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), bias=False)\n",
       "    (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (relu): ReLU()\n",
       "  )\n",
       "  (conv2d_2b): BasicConv2d(\n",
       "    (conv): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "    (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (relu): ReLU()\n",
       "  )\n",
       "  (maxpool_3a): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "  (conv2d_3b): BasicConv2d(\n",
       "    (conv): Conv2d(64, 80, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "    (bn): BatchNorm2d(80, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (relu): ReLU()\n",
       "  )\n",
       "  (conv2d_4a): BasicConv2d(\n",
       "    (conv): Conv2d(80, 192, kernel_size=(3, 3), stride=(1, 1), bias=False)\n",
       "    (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (relu): ReLU()\n",
       "  )\n",
       "  (conv2d_4b): BasicConv2d(\n",
       "    (conv): Conv2d(192, 256, kernel_size=(3, 3), stride=(2, 2), bias=False)\n",
       "    (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (relu): ReLU()\n",
       "  )\n",
       "  (repeat_1): Sequential(\n",
       "    (0): Block35(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (branch2): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(96, 256, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (1): Block35(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (branch2): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(96, 256, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (2): Block35(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (branch2): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(96, 256, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (3): Block35(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (branch2): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(96, 256, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (4): Block35(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (branch2): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(96, 256, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "  )\n",
       "  (mixed_6a): Mixed_6a(\n",
       "    (branch0): BasicConv2d(\n",
       "      (conv): Conv2d(256, 384, kernel_size=(3, 3), stride=(2, 2), bias=False)\n",
       "      (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (branch1): Sequential(\n",
       "      (0): BasicConv2d(\n",
       "        (conv): Conv2d(256, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (1): BasicConv2d(\n",
       "        (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (2): BasicConv2d(\n",
       "        (conv): Conv2d(192, 256, kernel_size=(3, 3), stride=(2, 2), bias=False)\n",
       "        (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "    )\n",
       "    (branch2): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "  )\n",
       "  (repeat_2): Sequential(\n",
       "    (0): Block17(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(256, 896, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (1): Block17(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(256, 896, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (2): Block17(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(256, 896, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (3): Block17(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(256, 896, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (4): Block17(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(256, 896, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (5): Block17(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(256, 896, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (6): Block17(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(256, 896, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (7): Block17(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(256, 896, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (8): Block17(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(256, 896, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (9): Block17(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(896, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(128, 128, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
       "          (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(256, 896, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "  )\n",
       "  (mixed_7a): Mixed_7a(\n",
       "    (branch0): Sequential(\n",
       "      (0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (1): BasicConv2d(\n",
       "        (conv): Conv2d(256, 384, kernel_size=(3, 3), stride=(2, 2), bias=False)\n",
       "        (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "    )\n",
       "    (branch1): Sequential(\n",
       "      (0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (1): BasicConv2d(\n",
       "        (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), bias=False)\n",
       "        (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "    )\n",
       "    (branch2): Sequential(\n",
       "      (0): BasicConv2d(\n",
       "        (conv): Conv2d(896, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (1): BasicConv2d(\n",
       "        (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (2): BasicConv2d(\n",
       "        (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), bias=False)\n",
       "        (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "    )\n",
       "    (branch3): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "  )\n",
       "  (repeat_3): Sequential(\n",
       "    (0): Block8(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(192, 192, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(192, 192, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(384, 1792, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (1): Block8(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(192, 192, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(192, 192, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(384, 1792, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (2): Block8(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(192, 192, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(192, 192, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(384, 1792, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (3): Block8(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(192, 192, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(192, 192, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(384, 1792, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (4): Block8(\n",
       "      (branch0): BasicConv2d(\n",
       "        (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (branch1): Sequential(\n",
       "        (0): BasicConv2d(\n",
       "          (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (1): BasicConv2d(\n",
       "          (conv): Conv2d(192, 192, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "        (2): BasicConv2d(\n",
       "          (conv): Conv2d(192, 192, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
       "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (conv2d): Conv2d(384, 1792, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "  )\n",
       "  (block8): Block8(\n",
       "    (branch0): BasicConv2d(\n",
       "      (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "      (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      (relu): ReLU()\n",
       "    )\n",
       "    (branch1): Sequential(\n",
       "      (0): BasicConv2d(\n",
       "        (conv): Conv2d(1792, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "        (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (1): BasicConv2d(\n",
       "        (conv): Conv2d(192, 192, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
       "        (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (2): BasicConv2d(\n",
       "        (conv): Conv2d(192, 192, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
       "        (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "    )\n",
       "    (conv2d): Conv2d(384, 1792, kernel_size=(1, 1), stride=(1, 1))\n",
       "  )\n",
       "  (avgpool_1a): AdaptiveAvgPool2d(output_size=1)\n",
       "  (dropout): Dropout(p=0.6, inplace=False)\n",
       "  (last_linear): Linear(in_features=1792, out_features=512, bias=False)\n",
       "  (last_bn): BatchNorm1d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
       "  (logits): Linear(in_features=512, out_features=1, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = InceptionResnetV1(\n",
    "    pretrained=\"vggface2\",\n",
    "    classify=True,\n",
    "    num_classes=1,\n",
    "    device=DEVICE\n",
    ")\n",
    "\n",
    "checkpoint = torch.load(\"resnetinceptionv1_epoch_32.pth\", map_location=torch.device('cpu'))\n",
    "model.load_state_dict(checkpoint['model_state_dict'])\n",
    "model.to(DEVICE)\n",
    "model.eval()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a499194a",
   "metadata": {},
   "source": [
    "# Model Inference "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "376e6cd6",
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict(input_image:Image.Image):\n",
    "    \"\"\"Predict the label of the input_image\"\"\"\n",
    "    face = mtcnn(input_image)\n",
    "    if face is None:\n",
    "        raise Exception('No face detected')\n",
    "    face = face.unsqueeze(0) # add the batch dimension\n",
    "    face = F.interpolate(face, size=(256, 256), mode='bilinear', align_corners=False)\n",
    "    \n",
    "    # convert the face into a numpy array to be able to plot it\n",
    "    prev_face = face.squeeze(0).permute(1, 2, 0).cpu().detach().int().numpy()\n",
    "    prev_face = prev_face.astype('uint8')\n",
    "\n",
    "    face = face.to(DEVICE)\n",
    "    face = face.to(torch.float32)\n",
    "    face = face / 255.0\n",
    "    face_image_to_plot = face.squeeze(0).permute(1, 2, 0).cpu().detach().int().numpy()\n",
    "\n",
    "    target_layers=[model.block8.branch1[-1]]\n",
    "    use_cuda = True if torch.cuda.is_available() else False\n",
    "    cam = GradCAM(model=model, target_layers=target_layers, use_cuda=use_cuda)\n",
    "    targets = [ClassifierOutputTarget(0)]\n",
    "\n",
    "    grayscale_cam = cam(input_tensor=face, targets=targets, eigen_smooth=True)\n",
    "    grayscale_cam = grayscale_cam[0, :]\n",
    "    visualization = show_cam_on_image(face_image_to_plot, grayscale_cam, use_rgb=True)\n",
    "    face_with_mask = cv2.addWeighted(prev_face, 1, visualization, 0.5, 0)\n",
    "\n",
    "    with torch.no_grad():\n",
    "        output = torch.sigmoid(model(face).squeeze(0))\n",
    "        prediction = \"real\" if output.item() < 0.5 else \"fake\"\n",
    "        \n",
    "        real_prediction = 1 - output.item()\n",
    "        fake_prediction = output.item()\n",
    "        \n",
    "        confidences = {\n",
    "            'real': real_prediction,\n",
    "            'fake': fake_prediction\n",
    "        }\n",
    "    return confidences, face_with_mask\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "14f47b5a",
   "metadata": {},
   "source": [
    "# Gradio Interface"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "d62177b5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running on local URL:  http://127.0.0.1:7860\n",
      "\n",
      "To create a public link, set `share=True` in `launch()`.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div><iframe src=\"http://127.0.0.1:7860/\" width=\"100%\" height=\"500\" allow=\"autoplay; camera; microphone; clipboard-read; clipboard-write;\" frameborder=\"0\" allowfullscreen></iframe></div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "interface = gr.Interface(\n",
    "    fn=predict,\n",
    "    inputs=[\n",
    "        gr.inputs.Image(label=\"Input Image\", type=\"pil\")\n",
    "    ],\n",
    "    outputs=[\n",
    "        gr.outputs.Label(label=\"Class\"),\n",
    "        gr.outputs.Image(label=\"Face with Explainability\", type=\"pil\")\n",
    "    ],\n",
    ").launch()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0c0b293c",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}