|
| 1 | +{ |
| 2 | + "cells": [ |
| 3 | + { |
| 4 | + "cell_type": "markdown", |
| 5 | + "id": "b59304be", |
| 6 | + "metadata": {}, |
| 7 | + "source": [ |
| 8 | + "# Installing Dependencies" |
| 9 | + ] |
| 10 | + }, |
| 11 | + { |
| 12 | + "cell_type": "code", |
| 13 | + "execution_count": null, |
| 14 | + "id": "77dcbaf5", |
| 15 | + "metadata": {}, |
| 16 | + "outputs": [], |
| 17 | + "source": [ |
| 18 | + "!pip3 install hub numpy pandas --quiet" |
| 19 | + ] |
| 20 | + }, |
| 21 | + { |
| 22 | + "cell_type": "markdown", |
| 23 | + "id": "b5ac6c37", |
| 24 | + "metadata": {}, |
| 25 | + "source": [ |
| 26 | + "# Loading Packages" |
| 27 | + ] |
| 28 | + }, |
| 29 | + { |
| 30 | + "cell_type": "code", |
| 31 | + "execution_count": null, |
| 32 | + "id": "97dec19b", |
| 33 | + "metadata": {}, |
| 34 | + "outputs": [], |
| 35 | + "source": [ |
| 36 | + "import hub\n", |
| 37 | + "import numpy as np\n", |
| 38 | + "import pandas as pd" |
| 39 | + ] |
| 40 | + }, |
| 41 | + { |
| 42 | + "cell_type": "markdown", |
| 43 | + "id": "40100d39", |
| 44 | + "metadata": {}, |
| 45 | + "source": [ |
| 46 | + "# Downloading Raw Data" |
| 47 | + ] |
| 48 | + }, |
| 49 | + { |
| 50 | + "cell_type": "code", |
| 51 | + "execution_count": null, |
| 52 | + "id": "f50cde54", |
| 53 | + "metadata": {}, |
| 54 | + "outputs": [], |
| 55 | + "source": [ |
| 56 | + "source_url = \"http://codh.rois.ac.jp/kmnist/dataset/kmnist/\"\n", |
| 57 | + "\n", |
| 58 | + "train_images_filepath = \"kmnist-train-imgs.npz\"\n", |
| 59 | + "train_labels_filepath = \"kmnist-train-labels.npz\"\n", |
| 60 | + "\n", |
| 61 | + "test_images_filepath = \"kmnist-test-imgs.npz\"\n", |
| 62 | + "test_labels_filepath = \"kmnist-test-labels.npz\"\n", |
| 63 | + "\n", |
| 64 | + "class_map_filepath = \"kmnist_classmap.csv\"" |
| 65 | + ] |
| 66 | + }, |
| 67 | + { |
| 68 | + "cell_type": "code", |
| 69 | + "execution_count": null, |
| 70 | + "id": "e5bc309d", |
| 71 | + "metadata": {}, |
| 72 | + "outputs": [], |
| 73 | + "source": [ |
| 74 | + "!curl -O {source_url}/{train_images_filepath} # Can also use `wget` if available\n", |
| 75 | + "!curl -O {source_url}/{train_labels_filepath}\n", |
| 76 | + "\n", |
| 77 | + "!curl -O {source_url}/{test_images_filepath}\n", |
| 78 | + "!curl -O {source_url}/{test_labels_filepath}\n", |
| 79 | + "\n", |
| 80 | + "!curl -O {source_url}/{class_map_filepath}" |
| 81 | + ] |
| 82 | + }, |
| 83 | + { |
| 84 | + "cell_type": "markdown", |
| 85 | + "id": "d57879f2", |
| 86 | + "metadata": {}, |
| 87 | + "source": [ |
| 88 | + "# Loading Class Labels" |
| 89 | + ] |
| 90 | + }, |
| 91 | + { |
| 92 | + "cell_type": "code", |
| 93 | + "execution_count": null, |
| 94 | + "id": "4a7dce68", |
| 95 | + "metadata": {}, |
| 96 | + "outputs": [], |
| 97 | + "source": [ |
| 98 | + "class_map_table = pd.read_csv(\n", |
| 99 | + " class_map_filepath, \n", |
| 100 | + " encoding='utf-8', \n", |
| 101 | + " index_col=0\n", |
| 102 | + ")\n", |
| 103 | + "\n", |
| 104 | + "class_names = class_map_table.codepoint.tolist()" |
| 105 | + ] |
| 106 | + }, |
| 107 | + { |
| 108 | + "cell_type": "markdown", |
| 109 | + "id": "a5257846", |
| 110 | + "metadata": {}, |
| 111 | + "source": [ |
| 112 | + "# Creating Dataset and Uploading to `hub`" |
| 113 | + ] |
| 114 | + }, |
| 115 | + { |
| 116 | + "cell_type": "markdown", |
| 117 | + "id": "7485fcd7", |
| 118 | + "metadata": {}, |
| 119 | + "source": [ |
| 120 | + "## Login\n", |
| 121 | + "\n", |
| 122 | + "This is needed if using Activeloop storage." |
| 123 | + ] |
| 124 | + }, |
| 125 | + { |
| 126 | + "cell_type": "code", |
| 127 | + "execution_count": null, |
| 128 | + "id": "641c1fc7", |
| 129 | + "metadata": {}, |
| 130 | + "outputs": [], |
| 131 | + "source": [ |
| 132 | + "username = \"<USERNAME>\"\n", |
| 133 | + "password = \"<PASSWORD>\"\n", |
| 134 | + "\n", |
| 135 | + "!activeloop login -u '{username}' -p '{password}'" |
| 136 | + ] |
| 137 | + }, |
| 138 | + { |
| 139 | + "cell_type": "code", |
| 140 | + "execution_count": null, |
| 141 | + "id": "9f1f6482", |
| 142 | + "metadata": {}, |
| 143 | + "outputs": [], |
| 144 | + "source": [ |
| 145 | + "workspace_path = f\"hub://{username}\" # Or `\".\"` if local" |
| 146 | + ] |
| 147 | + }, |
| 148 | + { |
| 149 | + "cell_type": "markdown", |
| 150 | + "id": "e136b1e3", |
| 151 | + "metadata": {}, |
| 152 | + "source": [ |
| 153 | + "## Train Set" |
| 154 | + ] |
| 155 | + }, |
| 156 | + { |
| 157 | + "cell_type": "code", |
| 158 | + "execution_count": null, |
| 159 | + "id": "ce50ee1f", |
| 160 | + "metadata": {}, |
| 161 | + "outputs": [], |
| 162 | + "source": [ |
| 163 | + "dataset_name = \"kmnist-train\"\n", |
| 164 | + "dataset_path = f\"{workspace_path}/{dataset_name}\"" |
| 165 | + ] |
| 166 | + }, |
| 167 | + { |
| 168 | + "cell_type": "code", |
| 169 | + "execution_count": null, |
| 170 | + "id": "919f3198", |
| 171 | + "metadata": {}, |
| 172 | + "outputs": [], |
| 173 | + "source": [ |
| 174 | + "ds = hub.empty(dataset_path, overwrite=True) # Set `overwrite=True` to overwrite any existing data under the same path\n", |
| 175 | + "\n", |
| 176 | + "with ds:\n", |
| 177 | + " ds.create_tensor('images', htype = 'image', sample_compression = \"jpg\")\n", |
| 178 | + " ds.create_tensor('labels', htype = 'class_label', class_names = class_names)\n", |
| 179 | + "\n", |
| 180 | + " ds.info.update(\n", |
| 181 | + " description = \"Kuzushiji-MNIST is a drop-in replacement for the MNIST dataset. It contains 70,000 28x28 grayscale images spanning 10 classes (one from each column of hiragana), and is perfectly balanced like the original MNIST dataset (6k/1k train/test for each class).\", \n", |
| 182 | + " citation=\"@online{clanuwat2018deep, author={Tarin Clanuwat and Mikel Bober-Irizar and Asanobu Kitamoto and Alex Lamb and Kazuaki Yamamoto and David Ha}, title={Deep Learning for Classical Japanese Literature}, date={2018-12-03}, year={2018}, eprintclass={cs.CV}, eprinttype={arXiv}, eprint={cs.CV/1812.01718}}\"\n", |
| 183 | + " )\n", |
| 184 | + "\n", |
| 185 | + "\n", |
| 186 | + "with ds:\n", |
| 187 | + " for image, label in zip(np.load(train_images_filepath)['arr_0'], np.load(train_labels_filepath)['arr_0']):\n", |
| 188 | + " ds.append({'images': image, 'labels': np.uint32(label)})" |
| 189 | + ] |
| 190 | + }, |
| 191 | + { |
| 192 | + "cell_type": "markdown", |
| 193 | + "id": "683fa2bd", |
| 194 | + "metadata": {}, |
| 195 | + "source": [ |
| 196 | + "# Test Set" |
| 197 | + ] |
| 198 | + }, |
| 199 | + { |
| 200 | + "cell_type": "code", |
| 201 | + "execution_count": null, |
| 202 | + "id": "6d0bafdc", |
| 203 | + "metadata": {}, |
| 204 | + "outputs": [], |
| 205 | + "source": [ |
| 206 | + "dataset_name = \"kmnist-test\"\n", |
| 207 | + "dataset_path = f\"{workspace_path}/{dataset_name}\"" |
| 208 | + ] |
| 209 | + }, |
| 210 | + { |
| 211 | + "cell_type": "code", |
| 212 | + "execution_count": null, |
| 213 | + "id": "f2eac537", |
| 214 | + "metadata": {}, |
| 215 | + "outputs": [], |
| 216 | + "source": [ |
| 217 | + "ds = hub.empty(dataset_path, overwrite=True)\n", |
| 218 | + "\n", |
| 219 | + "with ds:\n", |
| 220 | + " ds.create_tensor('images', htype = 'image', sample_compression = \"jpg\")\n", |
| 221 | + " ds.create_tensor('labels', htype = 'class_label', class_names = class_names)\n", |
| 222 | + "\n", |
| 223 | + " ds.info.update(\n", |
| 224 | + " description = \"Kuzushiji-MNIST is a drop-in replacement for the MNIST dataset. It contains 70,000 28x28 grayscale images spanning 10 classes (one from each column of hiragana), and is perfectly balanced like the original MNIST dataset (6k/1k train/test for each class).\", \n", |
| 225 | + " citation=\"@online{clanuwat2018deep, author={Tarin Clanuwat and Mikel Bober-Irizar and Asanobu Kitamoto and Alex Lamb and Kazuaki Yamamoto and David Ha}, title={Deep Learning for Classical Japanese Literature}, date={2018-12-03}, year={2018}, eprintclass={cs.CV}, eprinttype={arXiv}, eprint={cs.CV/1812.01718}}\"\n", |
| 226 | + " )\n", |
| 227 | + "\n", |
| 228 | + "\n", |
| 229 | + "with ds:\n", |
| 230 | + " for image, label in zip(np.load(test_images_filepath)['arr_0'], np.load(test_labels_filepath)['arr_0']):\n", |
| 231 | + " ds.append({'images': image, 'labels': np.uint32(label)})" |
| 232 | + ] |
| 233 | + }, |
| 234 | + { |
| 235 | + "cell_type": "markdown", |
| 236 | + "id": "82339916", |
| 237 | + "metadata": {}, |
| 238 | + "source": [ |
| 239 | + "Dataset documentation: https://docs.activeloop.ai/datasets/kmnist" |
| 240 | + ] |
| 241 | + }, |
| 242 | + { |
| 243 | + "cell_type": "code", |
| 244 | + "execution_count": null, |
| 245 | + "id": "0d4013aa", |
| 246 | + "metadata": {}, |
| 247 | + "outputs": [], |
| 248 | + "source": [] |
| 249 | + } |
| 250 | + ], |
| 251 | + "metadata": { |
| 252 | + "kernelspec": { |
| 253 | + "display_name": "Python 3 (ipykernel)", |
| 254 | + "language": "python", |
| 255 | + "name": "python3" |
| 256 | + }, |
| 257 | + "language_info": { |
| 258 | + "codemirror_mode": { |
| 259 | + "name": "ipython", |
| 260 | + "version": 3 |
| 261 | + }, |
| 262 | + "file_extension": ".py", |
| 263 | + "mimetype": "text/x-python", |
| 264 | + "name": "python", |
| 265 | + "nbconvert_exporter": "python", |
| 266 | + "pygments_lexer": "ipython3", |
| 267 | + "version": "3.9.5" |
| 268 | + } |
| 269 | + }, |
| 270 | + "nbformat": 4, |
| 271 | + "nbformat_minor": 5 |
| 272 | +} |
0 commit comments