{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "22553677-3a88-4f35-a99c-a6ec7375ef7c",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/yunkeli/anaconda3/lib/python3.10/site-packages/tqdm/auto.py:22: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running on local URL:  http://127.0.0.1:7860\n",
      "Running on public URL: https://a2d6699e89843c017c.gradio.live\n",
      "\n",
      "This share link expires in 72 hours. For free permanent hosting and GPU upgrades, run `gradio deploy` from Terminal to deploy to Spaces (https://huggingface.co/spaces)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div><iframe src=\"https://a2d6699e89843c017c.gradio.live\" 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": [
    "import cv2\n",
    "import numpy as np\n",
    "from matplotlib import pyplot as plt\n",
    "import gradio as gr\n",
    "\n",
    "# Helper function to detect sky condition and get the HSV range\n",
    "def detect_sky_color(hsv_image):\n",
    "    # Crop the image to the upper half, because we assume the sky is always on the upper half of the image\n",
    "    height = hsv_image.shape[0]\n",
    "    upper_half_image = hsv_image[:height//2, :]\n",
    "\n",
    "    # Define color ranges in HSV\n",
    "    blue_lower = np.array([46, 17, 148], np.uint8)\n",
    "    blue_upper = np.array([154, 185, 249], np.uint8)\n",
    "    orange_lower = np.array([10, 100, 100], np.uint8)\n",
    "    orange_upper = np.array([25, 183, 254], np.uint8)\n",
    "    pale_lower = np.array([0, 0, 129], np.uint8)\n",
    "    pale_upper = np.array([171, 64, 225], np.uint8)\n",
    "\n",
    "    # Create masks for colors\n",
    "    blue_mask = cv2.inRange(upper_half_image, blue_lower, blue_upper)\n",
    "    orange_mask = cv2.inRange(upper_half_image, orange_lower, orange_upper)\n",
    "    pale_mask = cv2.inRange(upper_half_image, pale_lower, pale_upper)\n",
    "\n",
    "    # Calculate the percentage of cropped image covered by each color\n",
    "    blue_percentage = np.sum(blue_mask > 0) / (upper_half_image.shape[0] * upper_half_image.shape[1]) * 100\n",
    "    orange_percentage = np.sum(orange_mask > 0) / (upper_half_image.shape[0] * upper_half_image.shape[1]) * 100\n",
    "    pale_percentage = np.sum(pale_mask > 0) / (upper_half_image.shape[0] * upper_half_image.shape[1]) * 100\n",
    "\n",
    "    # Determine the predominant color in the upper half\n",
    "    max_color = max(blue_percentage, orange_percentage, pale_percentage)\n",
    "    if max_color == blue_percentage:\n",
    "        return blue_lower, blue_upper\n",
    "    elif max_color == orange_percentage:\n",
    "        return orange_lower, orange_upper\n",
    "    else:\n",
    "        return pale_lower, pale_upper\n",
    "\n",
    "\n",
    "# Main function to process image and display sky masks\n",
    "def sky_segmentation(uploaded_image):\n",
    "    # Read the image\n",
    "    image = cv2.imread(uploaded_image)\n",
    "\n",
    "    # Convert to HSV image\n",
    "    hsv = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)\n",
    "\n",
    "    # Determine HSV range based on helper function\n",
    "    (hsv_lower, hsv_upper) = detect_sky_color(hsv)\n",
    "\n",
    "    # Use hsv_lower and hsv_upper to create a mask, which isolates the sky region\n",
    "    mask_initial = cv2.inRange(hsv, hsv_lower, hsv_upper)\n",
    "\n",
    "    # Apply morphological operations to fine-tune the mask\n",
    "    kernel = np.ones((3,3), np.uint8)\n",
    "    mask_fine_tuned = cv2.erode(mask_initial, kernel, iterations=1)\n",
    "    mask_fine_tuned = cv2.dilate(mask_fine_tuned, kernel, iterations=1)\n",
    "\n",
    "    # Perform connected component analysis\n",
    "    num_labels, labels_im = cv2.connectedComponents(mask_fine_tuned)\n",
    "\n",
    "    # Create an array to hold the size of each component\n",
    "    sizes = np.bincount(labels_im.flatten())\n",
    "\n",
    "    # Set the size of the background (label 0) to zero\n",
    "    sizes[0] = 0\n",
    "\n",
    "    # Find the largest component\n",
    "    max_label = np.argmax(sizes)\n",
    "\n",
    "    # Create a mask with only the largest component\n",
    "    sky_mask = np.zeros_like(mask_fine_tuned)\n",
    "    sky_mask[labels_im == max_label] = 255    \n",
    "    \n",
    "    return sky_mask\n",
    "\n",
    "\n",
    "# Create a Gradio demo\n",
    "demo = gr.Interface(sky_segmentation, gr.Image(type='filepath'), \"image\")\n",
    "if __name__ == \"__main__\":\n",
    "    demo.launch(share=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1e4ad199-ca35-48c0-9889-66fa874c4d9d",
   "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": "3.10.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}