augmentationSample.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# A Sample of Data Augmentation\n",
    "---\n",
    "Zhiang Chen, Nov 2018 @ DREAMS"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Generate Masks from XML\n",
    "- Download the xml annotations and original images from LabelMe\n",
    "- Create dataset folders under directory \"datasets\"  \n",
    "e.g. a tornado datasets\n",
    "```\n",
    "--data_augmentator\n",
    "|--datasets  \n",
    "    |--rocks\n",
    "        |--annotation: rocks1.xml, rocks2.xml, ...\n",
    "        |--image: rocks1.jpg, rocks2.jpg, ...\n",
    "```\n",
    "- Genrate masks from XML files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "177it [00:01, 158.58it/s]\n"
     ]
    }
   ],
   "source": [
    "from xml2mask import polygonReader\n",
    "objects = ['drone'] # give a list of the annotated objects\n",
    "pr = polygonReader(\"drone\", objects)\n",
    "\n",
    "#save compact masks on .jpg files\n",
    "#pr.saveMask(dim=(4000,4000))\n",
    "\n",
    "#generate and save multi-channel images on .npy files\n",
    "pr.generateMask2(dim=(1024, 768), resize=(768, 1024), saveOnline=True) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Augment data\n",
    "Before running the script, create a folder and move .npy files into the folder. And create a folder for augmentation data.\n",
    "```\n",
    "--data_augmentator\n",
    "|--datasets\n",
    "    |--rocks\n",
    "        |--annotation: rocks1.xml, rocks2.xml, ...\n",
    "        |--image: rocks1.jpg, rocks2.jpg, ...\n",
    "        |--npy: rocks1.npy, rocks2.npy, ...\n",
    "        |--aug: augmentation data\n",
    "```\n",
    "Resize the data before augmentation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from augmentation import augmentor\n",
    "import numpy as np\n",
    "from tqdm import tqdm\n",
    "import cv2\n",
    "import uuid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "0it [00:00, ?it/s]\n"
     ]
    },
    {
     "ename": "error",
     "evalue": "/build/opencv-L2vuMj/opencv-3.2.0+dfsg/modules/imgproc/src/imgwarp.cpp:5028: error: (-215) ifunc != 0 in function remap\n",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31merror\u001b[0m                                     Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-7-47a3da57a41d>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     15\u001b[0m \u001b[0maug\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0maugmentor\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimage_path\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mannotation_path\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     16\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 17\u001b[0;31m \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mm\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mf\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mtqdm\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0maug\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     18\u001b[0m     \u001b[0munid\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0muuid\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0muuid4\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhex\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     19\u001b[0m     \u001b[0mimg_name\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0maug_path\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0mf\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;34m\"_\"\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0munid\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;34m'.png'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tqdm/std.py\u001b[0m in \u001b[0;36m__iter__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   1125\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1126\u001b[0m         \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1127\u001b[0;31m             \u001b[0;32mfor\u001b[0m \u001b[0mobj\u001b[0m \u001b[0;32min\u001b[0m \u001b[0miterable\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   1128\u001b[0m                 \u001b[0;32myield\u001b[0m \u001b[0mobj\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1129\u001b[0m                 \u001b[0;31m# Update and possibly print the progressbar.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/data_augmentor/augmentation.py\u001b[0m in \u001b[0;36maugmentor\u001b[0;34m(image_path, annotation_path, mode, resize_dim, batch_number, rotation_min, rotation_max, fliplr, flipud, zoom_min, zoom_max)\u001b[0m\n\u001b[1;32m    143\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    144\u001b[0m                 img,mask = sample(img, mask, rotation_min, rotation_max,\n\u001b[0;32m--> 145\u001b[0;31m                                   fliplr, flipud, zoom_min, zoom_max)\n\u001b[0m\u001b[1;32m    146\u001b[0m                 \u001b[0;32myield\u001b[0m \u001b[0mimg\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmask\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mimage\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msplit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'.'\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[0m\n\u001b[1;32m    147\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/data_augmentor/augmentation.py\u001b[0m in \u001b[0;36msample\u001b[0;34m(image, mask, rotation_min, rotation_max, fliplr, flipud, zoom_min, zoom_max)\u001b[0m\n\u001b[1;32m     78\u001b[0m     \u001b[0mangle\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrandom\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0muniform\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrotation_min\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrotation_max\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     79\u001b[0m     \u001b[0mimage\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mrotateImage\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimage\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mangle\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 80\u001b[0;31m     \u001b[0mmask\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mrotateImage\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmask\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mangle\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     81\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     82\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mfliplr\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/data_augmentor/augmentation.py\u001b[0m in \u001b[0;36mrotateImage\u001b[0;34m(image, angle)\u001b[0m\n\u001b[1;32m     20\u001b[0m     \u001b[0mimage_center\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtuple\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimage\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m/\u001b[0m \u001b[0;36m2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     21\u001b[0m     \u001b[0mrot_mat\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcv2\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgetRotationMatrix2D\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimage_center\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mangle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m1.0\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 22\u001b[0;31m     \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcv2\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwarpAffine\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimage\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrot_mat\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mimage\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mflags\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mcv2\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mINTER_LINEAR\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     23\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mresult\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m<\u001b[0m\u001b[0ml\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     24\u001b[0m         \u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31merror\u001b[0m: /build/opencv-L2vuMj/opencv-3.2.0+dfsg/modules/imgproc/src/imgwarp.cpp:5028: error: (-215) ifunc != 0 in function remap\n"
     ]
    }
   ],
   "source": [
    "config = dict(\n",
    "            mode=2, \n",
    "            resize_dim=(500, 500),\n",
    "            batch_number=10,\n",
    "            rotation_min=-90,\n",
    "            rotation_max=90,\n",
    "            fliplr=True,\n",
    "            flipud=True,\n",
    "            zoom_min=0.8,\n",
    "            zoom_max=1.2)\n",
    "\n",
    "image_path = './datasets/Crater/image/'\n",
    "annotation_path = './datasets/Crater/npy/'\n",
    "aug_path = './datasets/Crater/aug/'\n",
    "aug = augmentor(image_path, annotation_path, **config)\n",
    "\n",
    "for i,m,f in tqdm(aug):\n",
    "    unid = uuid.uuid4().hex\n",
    "    img_name = aug_path + f + \"_\" + unid + '.png'\n",
    "    ann_name = aug_path + f + \"_\" + unid + '.npy'\n",
    "    cv2.imwrite(img_name, i)\n",
    "    np.save(ann_name, m)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "config = dict(\n",
    "            mode=2, \n",
    "            resize_dim=(1000, 1000),\n",
    "            batch_number=10,\n",
    "            rotation_min=-90,\n",
    "            rotation_max=90,\n",
    "            fliplr=True,\n",
    "            flipud=True,\n",
    "            zoom_min=0.8,\n",
    "            zoom_max=1.2)\n",
    "\n",
    "image_path = './datasets/drone/image/'\n",
    "annotation_path = './datasets/drone/npy/'\n",
    "aug_path = './datasets/drone/aug/'\n",
    "aug = augmentor(image_path, annotation_path, **config)\n",
    "\n",
    "for i,m,f in tqdm(aug):\n",
    "    unid = uuid.uuid4().hex\n",
    "    img_name = aug_path + f + \"_\" + unid + '.jpg'\n",
    "    ann_name = aug_path + f + \"_\" + unid + '.npy'\n",
    "    cv2.imwrite(img_name, i)\n",
    "    np.save(ann_name, m)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Rename images and annotations by numerical order"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It is recommended to back up the data before renaming."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from rename import renamer\n",
    "rn = renamer('./datasets/rocks/image/', './datasets/rocks/annotation/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "rn.rename(mode=1, image2png=True, annotation_suffix='.npy', annotation2suffix='.npy')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Alright, data augmentation done."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}