CIFAR10 - Preprocessing.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import os.path\n",
    "\n",
    "from collections import OrderedDict\n",
    "\n",
    "import torch\n",
    "from torch import nn, optim\n",
    "\n",
    "import numpy as np\n",
    "\n",
    "from torch.utils.data import Dataset, DataLoader, sampler\n",
    "from skimage import io\n",
    "\n",
    "import pickle\n",
    "\n",
    "from utils import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "class CIFAR10(Dataset):\n",
    "    \n",
    "    img_shape = (3, 32, 32)\n",
    "    \n",
    "    def __init__(self, folder, batch_size=32, num_workers=0, val_size=.2, seed=123):\n",
    "        self.folder = folder\n",
    "        \n",
    "        with open(os.path.join(folder, 'batches.meta'), 'rb') as f:\n",
    "            d = pickle.load(f, encoding='bytes')\n",
    "            self.classes = list(map(lambda x: x.decode(), d[b'label_names']))\n",
    "            \n",
    "        self.collections = [\n",
    "            io.imread_collection(os.path.join(folder, cls, '*.png'))\n",
    "            for cls in self.classes\n",
    "        ]\n",
    "        \n",
    "        data = []\n",
    "        labels = []\n",
    "        \n",
    "        for filename in ['data_batch_%d' % i for i in range(1, 6)] + ['test_batch']:\n",
    "            with open(os.path.join(folder, filename), 'rb') as f:\n",
    "                d = pickle.load(f, encoding='bytes')\n",
    "                imgs = d[b'data'].astype(float) / 255\n",
    "                data.append(imgs.reshape([len(imgs)] + list(self.img_shape)))\n",
    "                labels.append(np.array(d[b'labels']))\n",
    "                \n",
    "        del d, f, imgs, filename\n",
    "        \n",
    "        train_val_size = sum(map(len, data[:-1]))\n",
    "        data = np.concatenate(data)\n",
    "        labels = np.concatenate(labels)\n",
    "        test_size = len(data) - train_val_size\n",
    "        \n",
    "        # Transpose data so its n_samples, n_channels, n_rows, n_cols\n",
    "        # Also, transform it to float Tensor\n",
    "        data = torch.Tensor(data).float()\n",
    "        \n",
    "        self.seed = seed\n",
    "        rs = np.random.RandomState(seed)\n",
    "        train_val = np.arange(train_val_size)\n",
    "        rs.shuffle(train_val)\n",
    "        \n",
    "        test = np.arange(train_val_size, len(data))\n",
    "        \n",
    "        dataset_size = len(data)\n",
    "        idx = np.arange(dataset_size)\n",
    "\n",
    "        train_val_size = len(train_val)\n",
    "\n",
    "        train_size = int(round(train_val_size * (1 - val_size)))\n",
    "        val_size = train_val_size - train_size\n",
    "        train, val = train_val[:train_size], train_val[train_size:]\n",
    "\n",
    "        train_sampler = sampler.SubsetRandomSampler(train)\n",
    "        val_sampler = sampler.SubsetRandomSampler(val)\n",
    "        train_val_sampler = sampler.SubsetRandomSampler(train_val)\n",
    "        test_sampler = sampler.SubsetRandomSampler(test)\n",
    "\n",
    "        train_loader = DataLoader(self, batch_size=batch_size, num_workers=num_workers, sampler=train_sampler)\n",
    "        val_loader = DataLoader(self, batch_size=batch_size, num_workers=num_workers, sampler=val_sampler)\n",
    "        train_val_loader = DataLoader(self, batch_size=batch_size, num_workers=num_workers, sampler=train_val_sampler)\n",
    "        test_loader = DataLoader(self, batch_size=batch_size, num_workers=num_workers, sampler=test_sampler)\n",
    "\n",
    "        num_classes = len(self.classes)\n",
    "        \n",
    "        for k, v in locals().items():\n",
    "            setattr(self, k, v)\n",
    "        \n",
    "    def __len__(self):\n",
    "        return self.dataset_size\n",
    "    \n",
    "    def __getitem__(self, idx):\n",
    "        X = self.data[idx]\n",
    "        y = self.labels[idx]\n",
    "        cls = self.classes[y]\n",
    "        \n",
    "        return X, y, cls"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = CIFAR10(os.path.join('data', 'cifar-10-batches-py'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b6e43ea20>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "axes = subplots(2, 5).flatten()\n",
    "\n",
    "for ax in axes:\n",
    "    idx = np.random.choice(dataset.train_size)\n",
    "    im, _, label = dataset[dataset.train[idx]]\n",
    "    imshow(im, ax, title=label)"
   ]
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}