convert_brush_to_mask.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# conda create -n pcd -c conda-forge python=3.9 numpy opencv scikit-image scipy pillow matplotlib jupyter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import cv2\n",
    "import json\n",
    "import os "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "JSON_PATH = 'project-5-at-2023-11-29-13-41-b52afe32.json'\n",
    "IMAGES_PATH = 'images5'\n",
    "LABELS_PATH = 'labels5'\n",
    "LABELS_COLOR_PATH = 'labels5_color'\n",
    "\n",
    "# TODO buat list di text aja\n",
    "SAPI = 1\n",
    "KAMBING = 2\n",
    "AYAM = 3\n",
    "BEBEK = 4\n",
    "\n",
    "SAPI_COLOR = (255,0,0)\n",
    "KAMBING_COLOR = (0,255,0)\n",
    "AYAM_COLOR = (0,0,255)\n",
    "BEBEK_COLOR = (255,255,0)\n",
    "\n",
    "\n",
    "if not os.path.exists(LABELS_PATH): \n",
    "    os.makedirs(LABELS_PATH) \n",
    "\n",
    "if not os.path.exists(LABELS_COLOR_PATH): \n",
    "    os.makedirs(LABELS_COLOR_PATH) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# https://github.com/HumanSignal/label-studio-converter/blob/master/label_studio_converter/brush.py#L45\n",
    "class InputStream:\n",
    "    def __init__(self, data):\n",
    "        self.data = data\n",
    "        self.i = 0\n",
    "\n",
    "    def read(self, size):\n",
    "        out = self.data[self.i : self.i + size]\n",
    "        self.i += size\n",
    "        return int(out, 2)\n",
    "    \n",
    "def access_bit(data, num):\n",
    "    \"\"\"from bytes array to bits by num position\"\"\"\n",
    "    base = int(num // 8)\n",
    "    shift = 7 - int(num % 8)\n",
    "    return (data[base] & (1 << shift)) >> shift\n",
    "\n",
    "\n",
    "def bytes2bit(data):\n",
    "    \"\"\"get bit string from bytes data\"\"\"\n",
    "    return ''.join([str(access_bit(data, i)) for i in range(len(data) * 8)])\n",
    "def decode_rle(rle, print_params: bool = False):\n",
    "    \"\"\"from LS RLE to numpy uint8 3d image [width, height, channel]\n",
    "\n",
    "    Args:\n",
    "        print_params (bool, optional): If true, a RLE parameters print statement is suppressed\n",
    "    \"\"\"\n",
    "    input = InputStream(bytes2bit(rle))\n",
    "    num = input.read(32)\n",
    "    word_size = input.read(5) + 1\n",
    "    rle_sizes = [input.read(4) + 1 for _ in range(4)]\n",
    "\n",
    "    if print_params:\n",
    "        print(\n",
    "            'RLE params:', num, 'values', word_size, 'word_size', rle_sizes, 'rle_sizes'\n",
    "        )\n",
    "\n",
    "    i = 0\n",
    "    out = np.zeros(num, dtype=np.uint8)\n",
    "    while i < num:\n",
    "        x = input.read(1)\n",
    "        j = i + 1 + input.read(rle_sizes[input.read(2)])\n",
    "        if x:\n",
    "            val = input.read(word_size)\n",
    "            out[i:j] = val\n",
    "            i = j\n",
    "        else:\n",
    "            while i < j:\n",
    "                val = input.read(word_size)\n",
    "                out[i] = val\n",
    "                i += 1\n",
    "    return out\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "f = open(JSON_PATH)\n",
    "data = json.load(f)\n",
    "for dt in data:\n",
    "    file_upload = dt['file_upload']\n",
    "    print(file_upload)\n",
    "    img = cv2.imread(IMAGES_PATH+'\\\\'+file_upload)\n",
    "    height, width, channels = img.shape\n",
    "    label_image  = np.zeros([height, width], dtype=np.uint8)\n",
    "    label_color  = np.zeros([height, width, 3], dtype=np.uint8)\n",
    "\n",
    "    for ano in dt['annotations']:\n",
    "        for res in ano['result']:\n",
    "            ori_height = res['original_height']\n",
    "            ori_width = res['original_width']\n",
    "            if height!=ori_height and width!=ori_width:\n",
    "              raise Exception(\"Size not match!\")\n",
    "\n",
    "            label = res['value']['brushlabels'][0]\n",
    "            labelnum = 0\n",
    "            labelcol = (255,255,255)\n",
    "            if label=='sapi':\n",
    "                labelnum = SAPI\n",
    "                labelcol = SAPI_COLOR\n",
    "            elif label=='kambing':\n",
    "                labelnum = KAMBING\n",
    "                labelcol = KAMBING_COLOR\n",
    "            elif label=='ayam':\n",
    "                labelnum = AYAM\n",
    "                labelcol = AYAM_COLOR\n",
    "            elif label=='bebek':\n",
    "                labelnum = BEBEK\n",
    "                labelcol = BEBEK_COLOR\n",
    "            print(label)\n",
    "            \n",
    "            rle = res['value']['rle']\n",
    "            lbl_rle = np.reshape(decode_rle(rle), [height, width, 4])[:, :, 3]\n",
    "            label_num  = np.full([height, width], labelnum, dtype=np.uint8) \n",
    "            label_ano = cv2.bitwise_and(lbl_rle, label_num)\n",
    "            label_image = np.where(label_ano == 0, label_image, label_ano)\n",
    "\n",
    "            lbl_rle_col = cv2.merge((lbl_rle,lbl_rle,lbl_rle))\n",
    "            label_num_col  = np.full([height, width, 3], labelcol, dtype=np.uint8)\n",
    "            label_ano_col = cv2.bitwise_and(lbl_rle_col, label_num_col)\n",
    "            label_color = np.where(label_ano_col == [0,0,0], label_color, label_ano_col) # overlap\n",
    "\n",
    "    cv2.imwrite(LABELS_PATH+'\\\\'+file_upload+'.png', label_image) \n",
    "    cv2.imwrite(LABELS_COLOR_PATH+'\\\\'+file_upload+'.png', label_color)\n",
    "\n",
    "\n",
    "        "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "labelimg",
   "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": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}