Extract_Bar_Table.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "46a82f28",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import json\n",
    "\n",
    "from IPython.display import Image, display\n",
    "# from paddleocr import PaddleOCR,draw_ocr\n",
    "from tqdm import tqdm\n",
    "# ocr = PaddleOCR(use_angle_cls=True, lang='en', use_gpu=False, ocr_version=\"PP-OCRv3\", enable_mkldnn=True) # need to run only once to download and load model into memory"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0d06ca56",
   "metadata": {},
   "source": [
    "## Line Table Extraction\n",
    "In this section, we develop code to extract the data table from the Line Image. \n",
    "### Load the Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "feabaf57",
   "metadata": {},
   "outputs": [],
   "source": [
    "split = \"test2019\"\n",
    "# data_path = '/dvmm-filer2/projects/mingyang/semafor/chart_table/data/bar/annotations'\n",
    "data_path =  '/Users/kevinz/Desktop/semafor/chart_table/data_download/bar/annotations'\n",
    "test_data_path = '/'.join([data_path, \"instancesBar(1031)_{}.json\".format(split)])\n",
    "#image_path = '/dvmm-filer2/projects/mingyang/semafor/chart_table/data/bar/images/{}'.format(split)\n",
    "image_path = '/Users/kevinz/Desktop/semafor/chart_table/data_download/bar/images/{}'.format(split)\n",
    "with open(test_data_path, 'r') as f:\n",
    "    test_annotation = json.load(f)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "284217cb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['licenses', 'images', 'annotations', 'categories'])\n"
     ]
    }
   ],
   "source": [
    "print(test_annotation.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "8bcde079",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'list'>\n",
      "<class 'list'>\n",
      "<class 'list'>\n",
      "6262\n",
      "143084\n"
     ]
    }
   ],
   "source": [
    "print(type(test_annotation['images']))\n",
    "print(type(test_annotation['annotations']))\n",
    "print(type(test_annotation['categories']))\n",
    "\n",
    "print(len(test_annotation['images']))\n",
    "print(len(test_annotation['annotations']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "87362453",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'file_name': '8763309d9af01abde80aed0d82bd4a97_d3d3LnZvbGtzZ2V6b25kaGVpZGVuem9yZy5pbmZvCTEzMS4yMjQuMjUwLjEyMA==.xls-7-1.png', 'height': 508, 'width': 548, 'id': 0}\n",
      "{'image_id': 0, 'category_id': 0, 'bbox': [65.0, 399.0, 4.0, 53.0], 'area': 212.0, 'id': 0}\n",
      "{'supercategory': 'Series', 'id': 0, 'name': 'Series'}\n"
     ]
    }
   ],
   "source": [
    "print(test_annotation['images'][0])\n",
    "print(test_annotation['annotations'][0])\n",
    "print(test_annotation['categories'][0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "544fd60c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[{'image_id': 23, 'category_id': 0, 'bbox': [70.0, 141.0, 19.0, 98.0], 'area': 1862.0, 'id': 417}, {'image_id': 23, 'category_id': 0, 'bbox': [133.0, 161.0, 19.0, 78.0], 'area': 1482.0, 'id': 418}, {'image_id': 23, 'category_id': 0, 'bbox': [195.0, 150.0, 19.0, 89.0], 'area': 1691.0, 'id': 419}, {'image_id': 23, 'category_id': 0, 'bbox': [257.0, 172.0, 19.0, 67.0], 'area': 1273.0, 'id': 420}, {'image_id': 23, 'category_id': 0, 'bbox': [319.0, 171.0, 19.0, 68.0], 'area': 1292.0, 'id': 421}, {'image_id': 23, 'category_id': 0, 'bbox': [381.0, 151.0, 19.0, 88.0], 'area': 1672.0, 'id': 422}]\n"
     ]
    }
   ],
   "source": [
    "## Line Annotation\n",
    "#image_info = [x['id']: {'file_name':x['file_name'], 'height':x['height'], 'width':x['width']} for x in test_annotation['images']]\n",
    "#good sample: [28]\n",
    "sample_info = test_annotation['images'][23]\n",
    "sample_id = sample_info['id']\n",
    "#Find all the annotations:\n",
    "\n",
    "#Display the image\n",
    "sample_img_path = '/'.join([image_path, sample_info['file_name']])\n",
    "display(Image(filename=sample_img_path))\n",
    "\n",
    "#Map Annotations to Image ID\n",
    "AnnotationMap = {}\n",
    "for x in test_annotation['annotations']:\n",
    "    img_id = x['image_id']\n",
    "    AnnotationMap[img_id] = AnnotationMap.get(img_id, []) + [x]\n",
    "    \n",
    "sample_annotation = AnnotationMap[sample_id]\n",
    "print(sample_annotation)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25d3f114",
   "metadata": {},
   "source": [
    "### Detect Text Elements of the Image\n",
    "Extract the Text Elements from Line Image. \n",
    "\n",
    "1. **Extract Legend**\n",
    "    - Disregard the Numerical Values\n",
    "    - Disregard the bounding box that is not Vertically or Horizontally aligned\n",
    "    \n",
    "2. **Extract X-labels**\n",
    "   - Filter the texts that is above the 1/2 of the height (assume the x-axies is at the bottom)\n",
    "   - Using sliding line in vertical direction and record the intersected bounding box. \n",
    "   - The maximum number of intersection is considered as the x-labels\n",
    "   \n",
    "3. **Extract Y-labels** \n",
    "   - Filter the texts that is above the 1/2 of the height (assume the y-axies is to the left side)\n",
    "   - Using sliding line in vertical direction and record the intersected bounding box. \n",
    "   - The maximum number of intersection is considered as the y-labels\n",
    "\n",
    "\n",
    "2. **Extract Title**\n",
    "    - Detected Text that is closest to the Top. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "dded3892",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[[[713.0, 262.0], [778.0, 264.0], [777.0, 279.0], [713.0, 276.0]], ['dnaeaenl', 0.8970167636871338]], [[[715.0, 299.0], [860.0, 299.0], [860.0, 309.0], [715.0, 309.0]], ['Doageaanooacstanemnc', 0.967365562915802]], [[[713.0, 331.0], [845.0, 331.0], [845.0, 344.0], [713.0, 344.0]], ['Oreeancsior-omnanaii', 0.9261882901191711]], [[[715.0, 365.0], [776.0, 365.0], [776.0, 376.0], [715.0, 376.0]], ['Otseaodco', 0.950512170791626]], [[[715.0, 397.0], [900.0, 397.0], [900.0, 408.0], [715.0, 408.0]], ['eetess es sznroineecdrorieae', 0.9497971534729004]], [[[724.0, 411.0], [764.0, 411.0], [764.0, 422.0], [724.0, 422.0]], ['ianarss', 0.9753205180168152]]]\n"
     ]
    }
   ],
   "source": [
    "#Get the OCR of this image\n",
    "ocr_dir = \"/Users/kevinz/Desktop/semafor/chart_table/data_download/bar/ocr_results/test2019/{}.json\".format(sample_id)\n",
    "with open(ocr_dir, \"r\") as f:\n",
    "    ocr_texts = json.load(f)\n",
    "\n",
    "print(ocr_texts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "ea413ab9",
   "metadata": {},
   "outputs": [
    {
     "ename": "ValueError",
     "evalue": "max() arg is an empty sequence",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[0;32m/var/folders/m8/435h1kds3r15ck86lqk5fgtw0000gn/T/ipykernel_36307/2519815138.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m    160\u001b[0m \u001b[0;31m#print(y_labels)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    161\u001b[0m \u001b[0;31m#Estimate the Data Range Value\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 162\u001b[0;31m \u001b[0my_max\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_min\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_scale\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_max_pos\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_min_pos\u001b[0m\u001b[0;34m=\u001b[0m \u001b[0mestimate_range\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_labels\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    163\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    164\u001b[0m \u001b[0;31m#Get the legends\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/var/folders/m8/435h1kds3r15ck86lqk5fgtw0000gn/T/ipykernel_36307/2519815138.py\u001b[0m in \u001b[0;36mestimate_range\u001b[0;34m(y_labels)\u001b[0m\n\u001b[1;32m     76\u001b[0m     \u001b[0my_labels\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msort\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mlambda\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     77\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 78\u001b[0;31m     \u001b[0my_max\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmax\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mfloat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0my\u001b[0m \u001b[0;32min\u001b[0m \u001b[0my_labels\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     79\u001b[0m     \u001b[0my_max_candidate\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfloat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_labels\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     80\u001b[0m     \u001b[0;32mwhile\u001b[0m \u001b[0my_max\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0my_max_candidate\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mValueError\u001b[0m: max() arg is an empty sequence"
     ]
    }
   ],
   "source": [
    "import re\n",
    "\n",
    "def ocr_preprocessing(ocr_texts):\n",
    "    filtered_ocr_texts = []\n",
    "    thred = 0.8\n",
    "    for line in ocr_texts:\n",
    "        if line[1][1] > thred:\n",
    "            filtered_ocr_texts.append(line)\n",
    "    return filtered_ocr_texts\n",
    "\n",
    "def hv_alignment(a, b):\n",
    "    #compare the bounding box a and b to check if they are horizontal or vertically aligned\n",
    "    align_thred = 5\n",
    "    align_diff = [abs(a[0][0]-b[0][0]), abs(a[0][1]-b[0][1]), abs(a[2][0]-b[1][0]), abs(a[2][1]-b[2][1])]\n",
    "    if min(align_diff) < align_thred:\n",
    "        return True\n",
    "    else:\n",
    "        return False\n",
    "\n",
    "\n",
    "#Extract xlabels\n",
    "def extract_xlabels(ocr_texts, sample_info):\n",
    "    img_height = int(sample_info['height'])\n",
    "    #print(img_height)\n",
    "    x_label_candidates = []\n",
    "    \n",
    "    #Only consider the texts under the middle point. \n",
    "    starting_point = int(0.5*img_height)\n",
    "    #moving pace is the min of the bounding box size\n",
    "    step = int(min([x[0][2][1]-x[0][0][1] for x in ocr_texts]))\n",
    "    \n",
    "    intersected_list = []\n",
    "    for l_pos in range(starting_point, img_height, step):\n",
    "        #find the intersected texts\n",
    "        intersected_texts = [x for x in ocr_texts if x[0][0][1] < l_pos and x[0][2][1] > l_pos]\n",
    "        intersected_list.append(intersected_texts)\n",
    "    intersected_list = sorted(intersected_list, key=len)\n",
    "    x_label_candidates = intersected_list[-1]   \n",
    "        \n",
    "    return x_label_candidates\n",
    "\n",
    "#Extract ylabels\n",
    "def extract_ylabels(ocr_texts, sample_info):\n",
    "    img_width = int(sample_info['width'])\n",
    "    #print(img_height)\n",
    "    y_label_candidates = []\n",
    "    \n",
    "    #Only consider the texts under the middle point. \n",
    "    end_point = int(0.5*img_width)\n",
    "    #moving pace is the min of the bounding box width\n",
    "    step = int(min([x[0][2][0]-x[0][0][0] for x in ocr_texts]))\n",
    "    \n",
    "    intersected_list = []\n",
    "    for l_pos in range(0, end_point, step):\n",
    "        #find the intersected texts\n",
    "        intersected_texts = [x for x in ocr_texts if x[0][0][0] < l_pos and x[0][2][0] > l_pos]\n",
    "        intersected_list.append(intersected_texts)\n",
    "    intersected_list = sorted(intersected_list, key=len)\n",
    "    y_label_candidates = intersected_list[-1]  \n",
    "    #Get rid of y_labels that are not numbers\n",
    "    final_y_label_candidates = []\n",
    "    for y_label in y_label_candidates:\n",
    "        try:\n",
    "            y_label_text = y_label[1][0]\n",
    "            y_label_value = float(y_label[1][0])\n",
    "            final_y_label_candidates.append(y_label)\n",
    "        except:\n",
    "            continue\n",
    "        \n",
    "    return final_y_label_candidates\n",
    "def estimate_range(y_labels):\n",
    "    \"\"\"\n",
    "    Make sure the order of the value is following the order of the position. \n",
    "    \"\"\"\n",
    "    #sorted(intersected_list, key=len)(key=x[0][0][1])\n",
    "    y_labels.sort(key=lambda x: x[0][0][1])\n",
    "    \n",
    "    y_max = max([float(y[1][0]) for y in y_labels])\n",
    "    y_max_candidate = float(y_labels[0][1][0])\n",
    "    while y_max != y_max_candidate:\n",
    "        #pop the first item\n",
    "        del y_labels[0]\n",
    "        y_max = max([float(y[1][0]) for y in y_labels])\n",
    "        y_max_candidate = float(y_labels[0][1][0])\n",
    "    y_max_pos = 0.5*(y_labels[0][0][0][1]+y_labels[0][0][2][1])\n",
    "    \n",
    "    y_min = min([float(y[1][0]) for y in y_labels])\n",
    "    y_min_candidate = float(y_labels[-1][1][0])\n",
    "    while y_min != y_min_candidate:\n",
    "        #pop the first item\n",
    "        del y_labels[-1]\n",
    "        y_min = min([float(y[1][0]) for y in y_labels])\n",
    "        y_min_candidate = float(y_labels[-1][1][0])\n",
    "    y_min_pos = 0.5*(y_labels[-1][0][0][1]+y_labels[-1][0][2][1])\n",
    "    \n",
    "    #y_values = [float(y[1][0]) for y in y_labels]\n",
    "    y_scale = (y_max - y_min)/(0.5*(y_labels[-1][0][0][1]+y_labels[-1][0][2][1]) - 0.5*(y_labels[0][0][0][1]+y_labels[0][0][2][1]))\n",
    "    return y_max, y_min, y_scale, y_max_pos, y_min_pos\n",
    "\n",
    "#Extract Legend\n",
    "def extract_legend(ocr_texts):\n",
    "    numerical_pattern = \"(\\d+(\\.\\d+))\"\n",
    "\n",
    "    #Step 1, Get rid of Numerical Value\n",
    "    legend_candidates = []\n",
    "    for candidate in ocr_texts:\n",
    "        candidate_text = candidate[1][0]\n",
    "        if not re.search(numerical_pattern, candidate_text):\n",
    "            legend_candidates.append(candidate)\n",
    "\n",
    "    #print(legend_candidates)\n",
    "    #Step 2, Group Text based on their Vertical or Horizontal Alignment.\n",
    "    align_map = {}\n",
    "    for i_a,c_a in enumerate(legend_candidates):\n",
    "        align_map[i_a] = []\n",
    "        i_b = i_a+1\n",
    "        for c_b in legend_candidates[i_a+1:]:\n",
    "            if hv_alignment(c_a[0],c_b[0]):\n",
    "                align_map[i_a].append(i_b)\n",
    "            i_b +=1\n",
    "    \n",
    "    legend_groups = []\n",
    "    legend_group = []\n",
    "\n",
    "    for k,v in align_map.items():\n",
    "        if k in legend_group:\n",
    "            continue\n",
    "        if v:\n",
    "            legend_group += v + [k]\n",
    "            for x in v:\n",
    "                legend_group += align_map[x]\n",
    "            legend_group = list(set(legend_group))\n",
    "            legend_groups.append(legend_group)\n",
    "            legend_group = []\n",
    "    if len(legend_groups) == 0:\n",
    "        return []\n",
    "    #Get the final legends\n",
    "    if len(legend_groups) == 1 and len(legend_groups[0]) > 0:\n",
    "        return [legend_candidates[x] for x in legend_groups[0]]\n",
    "    else:\n",
    "        #Pick the one with the maximum elements or maybe the group that has the mapped elements as the gt sectors\n",
    "        legend_groups.sort(key=len)\n",
    "        return [legend_candidates[x] for x in legend_groups[-1]]\n",
    "\n",
    "def expand_legend_box(detected_legends):\n",
    "    #Assume the mark is to the left\n",
    "    for c_legend in detected_legends:\n",
    "        width = c_legend[0][2][0]-c_legend[0][0][0]\n",
    "        expand_width = int(0.2*width)\n",
    "        c_legend[0][0][0] -= expand_width\n",
    "        c_legend[0][3][0] -= expand_width\n",
    "        \n",
    "ocr_texts = ocr_preprocessing(ocr_texts)\n",
    "#Get the x__labels\n",
    "x_labels = extract_xlabels(ocr_texts, sample_info)\n",
    "\n",
    "#Get the y_labels\n",
    "#print(\"\\n\")\n",
    "y_labels = extract_ylabels(ocr_texts, sample_info)\n",
    "#print(y_labels)\n",
    "#Estimate the Data Range Value\n",
    "y_max, y_min, y_scale, y_max_pos, y_min_pos= estimate_range(y_labels)\n",
    "\n",
    "#Get the legends\n",
    "filtered_ocr_texts = []\n",
    "for c_ocr in ocr_texts:\n",
    "    if c_ocr in x_labels or c_ocr in y_labels:\n",
    "        continue\n",
    "    filtered_ocr_texts.append(c_ocr)\n",
    "\n",
    "detected_legends = extract_legend(filtered_ocr_texts)\n",
    "\n",
    "print(detected_legends)\n",
    "#Augment the length of detected_legends\n",
    "expand_legend_box(detected_legends)\n",
    "\n",
    "print(detected_legends)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "26b24090",
   "metadata": {},
   "source": [
    "### Read Data Values\n",
    "In this step, we compute the value of each point based on its locations compared tot he y_min"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "id": "2d1b08a2",
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "\n",
    "# def determinant(v1, v2):\n",
    "#   return v1[0]*v2[1] - v1[1]*v2[0]\n",
    "\n",
    "# def dotproduct(v1, v2):\n",
    "#   return sum((a*b) for a, b in zip(v1, v2))\n",
    "\n",
    "# def length(v):\n",
    "#   return math.sqrt(dotproduct(v, v))\n",
    "\n",
    "# def angle(v1, v2):\n",
    "#   return math.atan2(determinant(v1,v2), dotproduct(v1, v2))\n",
    "\n",
    "def compute_unit(p1,p2):\n",
    "    distance = [p1[0] - p2[0], -(p1[1] - p2[1])]\n",
    "    norm = math.sqrt(distance[0] ** 2 + distance[1] ** 2)\n",
    "    direction = [distance[0] / norm, distance[1] / norm]\n",
    "    return direction\n",
    "\n",
    "def compute_distance(p1,p2):\n",
    "    distance = (p1[0] - p2[0])**2 + (p1[1] - p2[1])**2\n",
    "    return distance\n",
    "    \n",
    "\n",
    "for line in sample_annotation:\n",
    "    line_pts = [(line['bbox'][i],line['bbox'][i+1])for i in range(0,len(line['bbox']),2)]\n",
    "    line_values = [0]*len(line_pts)\n",
    "    for i,pt in enumerate(line_pts):\n",
    "        pt_value = y_scale*(y_min_pos - pt[1]) + y_min\n",
    "        line_values[i] = pt_value\n",
    "    line['values'] = line_values\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bfc927e4",
   "metadata": {},
   "source": [
    "### Map the Legend or the Column Text\n",
    "In this final step we will map the legend text or the other detected Text to each line."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "id": "5799560a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[(387.0, 246.0), (380.0, 243.0), (376.0, 239.0), (370.0, 241.0), (361.0, 246.0), (355.0, 254.0), (348.0, 252.0), (335.0, 252.0), (329.0, 248.0), (322.0, 248.0), (316.0, 250.0), (310.0, 252.0), (303.0, 248.0), (297.0, 242.0), (291.0, 234.0), (284.0, 221.0), (278.0, 212.0), (271.0, 210.0), (265.0, 192.0), (258.0, 190.0), (252.0, 191.0), (246.0, 191.0), (239.0, 192.0), (233.0, 190.0), (227.0, 193.0), (220.0, 169.0), (214.0, 169.0), (206.0, 169.0), (201.0, 148.0), (194.0, 148.0), (188.0, 148.0), (182.0, 148.0), (175.0, 148.0), (169.0, 143.0), (162.0, 158.0), (156.0, 158.0), (150.0, 178.0), (143.0, 153.0), (137.0, 127.0), (130.0, 110.0), (124.0, 122.0), (118.0, 143.0), (110.0, 145.0), (105.0, 112.0), (98.0, 112.0), (92.0, 77.0), (86.0, 82.0), (79.0, 82.0), (73.0, 82.0), (66.0, 87.0)]\n",
      "[(387.0, 256.0), (380.0, 247.0), (376.0, 247.0), (370.0, 251.0), (361.0, 257.0), (355.0, 268.0), (348.0, 269.0), (335.0, 269.0), (329.0, 255.0), (322.0, 266.0), (316.0, 260.0), (310.0, 261.0), (303.0, 257.0), (297.0, 268.0), (291.0, 252.0), (284.0, 239.0), (278.0, 236.0), (271.0, 236.0), (265.0, 236.0), (258.0, 228.0), (252.0, 209.0), (246.0, 216.0), (239.0, 229.0), (233.0, 227.0), (227.0, 228.0), (220.0, 227.0), (214.0, 221.0), (206.0, 211.0), (201.0, 207.0), (194.0, 214.0), (188.0, 211.0), (182.0, 225.0), (175.0, 228.0), (169.0, 229.0), (162.0, 223.0), (156.0, 209.0), (150.0, 204.0), (143.0, 197.0), (137.0, 178.0), (130.0, 174.0), (124.0, 156.0), (118.0, 151.0), (110.0, 151.0), (105.0, 113.0), (98.0, 124.0), (92.0, 128.0), (86.0, 139.0), (79.0, 142.0), (73.0, 128.0), (66.0, 162.0)]\n"
     ]
    }
   ],
   "source": [
    "import cv2\n",
    "from matplotlib import pyplot as plt\n",
    "import numpy as np\n",
    "from sklearn.cluster import KMeans\n",
    "from collections import Counter\n",
    "\n",
    "def get_dominant_color(image, k=4, image_processing_size = None):\n",
    "    \"\"\"\n",
    "    takes an image as input\n",
    "    returns the dominant color of the image as a list\n",
    "    \n",
    "    dominant color is found by running k means on the \n",
    "    pixels & returning the centroid of the largest cluster\n",
    "\n",
    "    processing time is sped up by working with a smaller image; \n",
    "    this resizing can be done with the image_processing_size param \n",
    "    which takes a tuple of image dims as input\n",
    "\n",
    "    >>> get_dominant_color(my_image, k=4, image_processing_size = (25, 25))\n",
    "    [56.2423442, 34.0834233, 70.1234123]\n",
    "    \"\"\"\n",
    "    #resize image if new dims provided\n",
    "    if image_processing_size is not None:\n",
    "        image = cv2.resize(image, image_processing_size, \n",
    "                            interpolation = cv2.INTER_AREA)\n",
    "    \n",
    "    #reshape the image to be a list of pixels\n",
    "    image = image.reshape((image.shape[0] * image.shape[1], 3))\n",
    "\n",
    "    #cluster and assign labels to the pixels \n",
    "    clt = KMeans(n_clusters = k)\n",
    "    labels = clt.fit_predict(image)\n",
    "\n",
    "    #count labels to find most popular\n",
    "    label_counts = Counter(labels)\n",
    "    \n",
    "    \n",
    "    #subset out most popular centroid\n",
    "    #dominant_color = clt.cluster_centers_[label_counts.most_common(1)[0][0]]\n",
    "    dominant_colors = clt.cluster_centers_\n",
    "    return list(dominant_colors)\n",
    "\n",
    "#If Legend Exist and the Length Matched\n",
    "#if detected_legends and len(detected_legends) == len(sample_annotation):\n",
    "#Get the dominant Color for each Legends\n",
    "legends = {}\n",
    "sample_image = cv2.imread(sample_img_path)\n",
    "sample_image = cv2.cvtColor(sample_image, cv2.COLOR_BGR2RGB)\n",
    "for i, dl in enumerate(detected_legends):\n",
    "    legends[i] = {'bbox': dl[0], 'text': dl[1]}\n",
    "    start_point = (int(dl[0][0][0]), int(dl[0][0][1]))\n",
    "    end_point = (int(dl[0][2][0]), int(dl[0][2][1]))\n",
    "    \n",
    "    cropped_image = sample_image[start_point[1]:end_point[1], start_point[0]:end_point[0]]\n",
    "    plt.imshow(cropped_image)\n",
    "    plt.show()\n",
    "#     print(cropped_image[5,3])\n",
    "#     dominant_colors = get_dominant_color(cropped_image, k=3)\n",
    "#     print(dominant_colors)\n",
    "    legends[i]['cropped_image'] = cropped_image\n",
    "\n",
    "\n",
    "#Find the Color for Each Line, locate a center point, assume the legend will be there\n",
    "for line in sample_annotation:\n",
    "    line_pts = [(line['bbox'][i],line['bbox'][i+1])for i in range(0,len(line['bbox']),2)]\n",
    "    line['column_value'] = [\"None\"]*len(line_pts)\n",
    "    pick_point = line_pts[int(0.5*len(line_pts))]\n",
    "    pick_point = (int(pick_point[0]), int(pick_point[1]))\n",
    "    \n",
    "    sample_image_hsv = cv2.cvtColor(sample_image, cv2.COLOR_RGB2HSV)\n",
    "    pick_point_color = sample_image_hsv[pick_point[1],pick_point[0]]\n",
    "    \n",
    "    pp_color_min = np.array([max(0, pick_point_color[0]-5),max(50,pick_point_color[1]-10),max(50, pick_point_color[2]-10)], np.uint8)\n",
    "    pp_color_max = np.array([pick_point_color[0]+5,min(255, pick_point_color[1]+10),min(255, pick_point_color[2]+10)], np.uint8)\n",
    "    \n",
    "    line['row_value'] = \"None\"\n",
    "    for k, cl in legends.items():\n",
    "        cl_image = cl[\"cropped_image\"]\n",
    "        cl_hsv_image = cv2.cvtColor(cl_image, cv2.COLOR_RGB2HSV)\n",
    "        frame_threshed = cv2.inRange(cl_hsv_image, pp_color_min, pp_color_max)\n",
    "        \n",
    "\n",
    "        #might satisty two saturation\n",
    "        \n",
    "        if np.count_nonzero(frame_threshed)/(frame_threshed.shape[0]*frame_threshed.shape[1]) > 0.01:\n",
    "            #Find a Map\n",
    "            line['row_value'] = cl['text'][0]\n",
    "            break\n",
    "     #Map the line column values \n",
    "    print(line_pts)\n",
    "    for i, pt in enumerate(line_pts):\n",
    "        pt_x = pt[0]\n",
    "        thred = 5 #maybe set it to the length of the bbox\n",
    "        for x_label in x_labels:\n",
    "            x_label_text = x_label[1][0]\n",
    "            c_x = 0.5*(x_label[0][0][0]+x_label[0][2][0])\n",
    "            if abs(c_x-pt_x) < thred:\n",
    "                line['column_value'][i] = x_label_text\n",
    "\n",
    "\n",
    "        \n",
    "        \n",
    "                             \n",
    "        \n",
    "\n",
    "                             \n",
    "    \n",
    "#plt.imshow('Binary',thresh_img)\n",
    "#print(sample_annotation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "id": "6a26dcb9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(None,PNW SW): 218.7782805429864\n",
      "(None,PNW SW): 224.2081447963801\n",
      "(None,PNW SW): 231.44796380090497\n",
      "(None,PNW SW): 227.82805429864254\n",
      "(None,PNW SW): 218.7782805429864\n",
      "(None,PNW SW): 204.29864253393666\n",
      "(Dec-08,PNW SW): 207.9185520361991\n",
      "(None,PNW SW): 207.9185520361991\n",
      "(None,PNW SW): 215.15837104072398\n",
      "(None,PNW SW): 215.15837104072398\n",
      "(None,PNW SW): 211.53846153846155\n",
      "(None,PNW SW): 207.9185520361991\n",
      "(None,PNW SW): 215.15837104072398\n",
      "(None,PNW SW): 226.01809954751133\n",
      "(O ct-03,PNW SW): 240.4977375565611\n",
      "(None,PNW SW): 264.02714932126696\n",
      "(None,PNW SW): 280.31674208144796\n",
      "(None,PNW SW): 283.9366515837104\n",
      "(None,PNW SW): 316.5158371040724\n",
      "(None,PNW SW): 320.1357466063348\n",
      "(None,PNW SW): 318.32579185520365\n",
      "(None,PNW SW): 318.32579185520365\n",
      "(None,PNW SW): 316.5158371040724\n",
      "(Aug-08,PNW SW): 320.1357466063348\n",
      "(None,PNW SW): 314.7058823529412\n",
      "(None,PNW SW): 358.1447963800905\n",
      "(None,PNW SW): 358.1447963800905\n",
      "(None,PNW SW): 358.1447963800905\n",
      "(None,PNW SW): 396.1538461538462\n",
      "(None,PNW SW): 396.1538461538462\n",
      "(None,PNW SW): 396.1538461538462\n",
      "(Jun-06,PNW SW): 396.1538461538462\n",
      "(Jun-06,PNW SW): 396.1538461538462\n",
      "(None,PNW SW): 405.20361990950227\n",
      "(None,PNW SW): 378.0542986425339\n",
      "(None,PNW SW): 378.0542986425339\n",
      "(None,PNW SW): 341.8552036199095\n",
      "(None,PNW SW): 387.10407239819006\n",
      "(None,PNW SW): 434.1628959276018\n",
      "(None,PNW SW): 464.9321266968326\n",
      "(Apr-06,PNW SW): 443.2126696832579\n",
      "(Apr-06,PNW SW): 405.20361990950227\n",
      "(None,PNW SW): 401.5837104072398\n",
      "(None,PNW SW): 461.3122171945701\n",
      "(None,PNW SW): 461.3122171945701\n",
      "(None,PNW SW): 524.6606334841629\n",
      "(None,PNW SW): 515.6108597285067\n",
      "(None,PNW SW): 515.6108597285067\n",
      "(None,PNW SW): 515.6108597285067\n",
      "(Feb-08,PNW SW): 506.56108597285066\n",
      "(None,GuIfSRW): 200.6787330316742\n",
      "(None,GuIfSRW): 216.9683257918552\n",
      "(None,GuIfSRW): 216.9683257918552\n",
      "(None,GuIfSRW): 209.72850678733033\n",
      "(None,GuIfSRW): 198.86877828054298\n",
      "(None,GuIfSRW): 178.95927601809956\n",
      "(Dec-08,GuIfSRW): 177.14932126696831\n",
      "(None,GuIfSRW): 177.14932126696831\n",
      "(None,GuIfSRW): 202.4886877828054\n",
      "(None,GuIfSRW): 182.579185520362\n",
      "(None,GuIfSRW): 193.43891402714934\n",
      "(None,GuIfSRW): 191.6289592760181\n",
      "(None,GuIfSRW): 198.86877828054298\n",
      "(None,GuIfSRW): 178.95927601809956\n",
      "(O ct-03,GuIfSRW): 207.9185520361991\n",
      "(None,GuIfSRW): 231.44796380090497\n",
      "(None,GuIfSRW): 236.87782805429865\n",
      "(None,GuIfSRW): 236.87782805429865\n",
      "(None,GuIfSRW): 236.87782805429865\n",
      "(None,GuIfSRW): 251.3574660633484\n",
      "(None,GuIfSRW): 285.74660633484166\n",
      "(None,GuIfSRW): 273.0769230769231\n",
      "(None,GuIfSRW): 249.54751131221718\n",
      "(Aug-08,GuIfSRW): 253.16742081447964\n",
      "(None,GuIfSRW): 251.3574660633484\n",
      "(None,GuIfSRW): 253.16742081447964\n",
      "(None,GuIfSRW): 264.02714932126696\n",
      "(None,GuIfSRW): 282.1266968325792\n",
      "(None,GuIfSRW): 289.3665158371041\n",
      "(None,GuIfSRW): 276.6968325791855\n",
      "(None,GuIfSRW): 282.1266968325792\n",
      "(Jun-06,GuIfSRW): 256.7873303167421\n",
      "(Jun-06,GuIfSRW): 251.3574660633484\n",
      "(None,GuIfSRW): 249.54751131221718\n",
      "(None,GuIfSRW): 260.40723981900453\n",
      "(None,GuIfSRW): 285.74660633484166\n",
      "(None,GuIfSRW): 294.79638009049773\n",
      "(None,GuIfSRW): 307.4660633484163\n",
      "(None,GuIfSRW): 341.8552036199095\n",
      "(None,GuIfSRW): 349.09502262443436\n",
      "(Apr-06,GuIfSRW): 381.67420814479635\n",
      "(Apr-06,GuIfSRW): 390.7239819004525\n",
      "(None,GuIfSRW): 390.7239819004525\n",
      "(None,GuIfSRW): 459.5022624434389\n",
      "(None,GuIfSRW): 439.59276018099547\n",
      "(None,GuIfSRW): 432.3529411764706\n",
      "(None,GuIfSRW): 412.4434389140271\n",
      "(None,GuIfSRW): 407.0135746606335\n",
      "(None,GuIfSRW): 432.3529411764706\n",
      "(Feb-08,GuIfSRW): 370.81447963800906\n"
     ]
    }
   ],
   "source": [
    "display(Image(filename=sample_img_path))\n",
    "for result in sample_annotation:\n",
    "    for c, v in zip(result['column_value'], result['values']):\n",
    "        print(f\"({c},{result['row_value']}): {v}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "525ea515",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "chart2table",
   "language": "python",
   "name": "chart2table"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}