{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5d4984fd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Note: you may need to restart the kernel to use updated packages.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\n",
      "gradio 5.43.1 requires fastapi<1.0,>=0.115.2, but you have fastapi 0.115.0 which is incompatible.\n",
      "gradio 5.43.1 requires pydantic<2.12,>=2.0, but you have pydantic 2.12.5 which is incompatible.\n",
      "gradio 5.43.1 requires starlette<1.0,>=0.40.0; sys_platform != \"emscripten\", but you have starlette 0.38.6 which is incompatible.\n"
     ]
    }
   ],
   "source": [
    "%pip install -q torch torchvision pillow opencv-python segmentation-models-pytorch albumentations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "8e934e27",
   "metadata": {},
   "outputs": [
    {
     "ename": "OSError",
     "evalue": "[WinError 126] The specified module could not be found. Error loading \"c:\\Users\\abhay\\anaconda3\\Lib\\site-packages\\torch\\lib\\fbgemm.dll\" or one of its dependencies.",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mOSError\u001b[0m                                   Traceback (most recent call last)",
      "Cell \u001b[1;32mIn[2], line 1\u001b[0m\n\u001b[1;32m----> 1\u001b[0m \u001b[38;5;28;01mimport\u001b[39;00m \u001b[38;5;21;01mtorch\u001b[39;00m\n\u001b[0;32m      2\u001b[0m \u001b[38;5;28;01mimport\u001b[39;00m \u001b[38;5;21;01mtorchvision\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mtransforms\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m \u001b[38;5;21;01mtransforms\u001b[39;00m\n\u001b[0;32m      3\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01mPIL\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m Image\n",
      "File \u001b[1;32mc:\\Users\\abhay\\anaconda3\\Lib\\site-packages\\torch\\__init__.py:148\u001b[0m\n\u001b[0;32m    146\u001b[0m                 err \u001b[38;5;241m=\u001b[39m ctypes\u001b[38;5;241m.\u001b[39mWinError(ctypes\u001b[38;5;241m.\u001b[39mget_last_error())\n\u001b[0;32m    147\u001b[0m                 err\u001b[38;5;241m.\u001b[39mstrerror \u001b[38;5;241m+\u001b[39m\u001b[38;5;241m=\u001b[39m \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m Error loading \u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;132;01m{\u001b[39;00mdll\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m or one of its dependencies.\u001b[39m\u001b[38;5;124m'\u001b[39m\n\u001b[1;32m--> 148\u001b[0m                 \u001b[38;5;28;01mraise\u001b[39;00m err\n\u001b[0;32m    150\u001b[0m     kernel32\u001b[38;5;241m.\u001b[39mSetErrorMode(prev_error_mode)\n\u001b[0;32m    153\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21m_preload_cuda_deps\u001b[39m(lib_folder, lib_name):\n",
      "\u001b[1;31mOSError\u001b[0m: [WinError 126] The specified module could not be found. Error loading \"c:\\Users\\abhay\\anaconda3\\Lib\\site-packages\\torch\\lib\\fbgemm.dll\" or one of its dependencies."
     ]
    }
   ],
   "source": [
    "import torch\n",
    "import torchvision.transforms as transforms\n",
    "from PIL import Image\n",
    "import cv2\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import segmentation_models_pytorch as smp\n",
    "import albumentations as A\n",
    "from albumentations.pytorch import ToTensorV2\n",
    "\n",
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "print(\"Using device:\", device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f37c3d10",
   "metadata": {},
   "outputs": [],
   "source": [
    "# ── Load trained model ──────────────────────────────────────────────────────\n",
    "MODEL_PATH = \"best_model.pth\"   # path to saved weights\n",
    "\n",
    "model = smp.Unet(\n",
    "    encoder_name=\"efficientnet-b3\",\n",
    "    encoder_weights=None,          # weights loaded from checkpoint\n",
    "    in_channels=3,\n",
    "    classes=1,\n",
    "    activation=None\n",
    ")\n",
    "model.load_state_dict(torch.load(MODEL_PATH, map_location=device))\n",
    "model.to(device)\n",
    "model.eval()\n",
    "print(\"Model loaded successfully.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c63c88a3",
   "metadata": {},
   "outputs": [],
   "source": [
    "# ── Preprocess & run inference ───────────────────────────────────────────────\n",
    "IMAGE_PATH = r\"test images\\test 1.png\"   # test image\n",
    "PATCH_SIZE  = 256\n",
    "\n",
    "transform = A.Compose([\n",
    "    A.Normalize(mean=(0.485, 0.456, 0.406),\n",
    "                std=(0.229, 0.224, 0.225)),\n",
    "    ToTensorV2()\n",
    "])\n",
    "\n",
    "# Load image\n",
    "img_bgr = cv2.imread(IMAGE_PATH)\n",
    "assert img_bgr is not None, f\"Could not read image: {IMAGE_PATH}\"\n",
    "img_rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)\n",
    "\n",
    "h, w = img_rgb.shape[:2]\n",
    "\n",
    "# Pad so dimensions are divisible by PATCH_SIZE\n",
    "pad_h = (PATCH_SIZE - h % PATCH_SIZE) % PATCH_SIZE\n",
    "pad_w = (PATCH_SIZE - w % PATCH_SIZE) % PATCH_SIZE\n",
    "img_padded = np.pad(img_rgb, ((0, pad_h), (0, pad_w), (0, 0)), mode='reflect')\n",
    "H, W = img_padded.shape[:2]\n",
    "\n",
    "# Stitch patch predictions into a full mask\n",
    "full_mask = np.zeros((H, W), dtype=np.float32)\n",
    "\n",
    "with torch.no_grad():\n",
    "    for i in range(0, H, PATCH_SIZE):\n",
    "        for j in range(0, W, PATCH_SIZE):\n",
    "            patch = img_padded[i:i+PATCH_SIZE, j:j+PATCH_SIZE]\n",
    "            tensor = transform(image=patch)[\"image\"].unsqueeze(0).to(device)\n",
    "            pred   = torch.sigmoid(model(tensor)).squeeze().cpu().numpy()\n",
    "            full_mask[i:i+PATCH_SIZE, j:j+PATCH_SIZE] = pred\n",
    "\n",
    "# Crop back to original size\n",
    "pred_mask = (full_mask[:h, :w] > 0.5).astype(np.uint8)\n",
    "print(f\"Inference done. Image size: {h}×{w} | Buildings detected: {pred_mask.sum()>0}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ffbae7d8",
   "metadata": {},
   "outputs": [],
   "source": [
    "# ── Zoning mask & illegal building detection ─────────────────────────────────\n",
    "# Default zoning: right half is restricted. Modify as needed.\n",
    "def create_zoning_mask(shape):\n",
    "    \"\"\"Returns a binary mask (1 = restricted zone).\"\"\"\n",
    "    zm = np.zeros(shape, dtype=np.uint8)\n",
    "    zm[:, shape[1] // 2:] = 1\n",
    "    return zm\n",
    "\n",
    "def detect_illegal_buildings(building_mask, zoning_mask):\n",
    "    num_labels, labels = cv2.connectedComponents(building_mask)\n",
    "    illegal, legal = [], []\n",
    "    for lbl in range(1, num_labels):\n",
    "        pixels = (labels == lbl)\n",
    "        if (pixels & (zoning_mask == 1)).any():\n",
    "            illegal.append(lbl)\n",
    "        else:\n",
    "            legal.append(lbl)\n",
    "    return illegal, legal, labels\n",
    "\n",
    "def overlay_illegal(image_rgb, labels, illegal_buildings):\n",
    "    out = image_rgb.copy()\n",
    "    for lbl in illegal_buildings:\n",
    "        out[labels == lbl] = [255, 0, 0]   # red highlight\n",
    "    return out\n",
    "\n",
    "zoning_mask = create_zoning_mask(pred_mask.shape)\n",
    "illegal, legal, labels = detect_illegal_buildings(pred_mask, zoning_mask)\n",
    "overlay = overlay_illegal(img_rgb, labels, illegal)\n",
    "\n",
    "print(f\"Total buildings : {len(illegal) + len(legal)}\")\n",
    "print(f\"Illegal buildings: {len(illegal)}\")\n",
    "print(f\"Legal  buildings : {len(legal)}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6692053a",
   "metadata": {},
   "outputs": [],
   "source": [
    "# ── Visualize results ────────────────────────────────────────────────────────\n",
    "fig, axes = plt.subplots(1, 4, figsize=(20, 5))\n",
    "\n",
    "axes[0].imshow(img_rgb)\n",
    "axes[0].set_title(\"Input Image\")\n",
    "axes[0].axis(\"off\")\n",
    "\n",
    "axes[1].imshow(pred_mask, cmap=\"gray\")\n",
    "axes[1].set_title(\"Building Mask\")\n",
    "axes[1].axis(\"off\")\n",
    "\n",
    "axes[2].imshow(zoning_mask, cmap=\"gray\")\n",
    "axes[2].set_title(\"Zoning Mask\\n(white = restricted)\")\n",
    "axes[2].axis(\"off\")\n",
    "\n",
    "axes[3].imshow(overlay)\n",
    "axes[3].set_title(f\"Illegal Buildings (red)\\nIllegal: {len(illegal)} | Legal: {len(legal)}\")\n",
    "axes[3].axis(\"off\")\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.12.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}