mnist_mlp_pipeline.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# MNIST Multi Layer Perceptron Classifier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Content"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Data Preprocessing\n",
    "<br/>Random Baseline \n",
    "<br/>Model Training\n",
    "<br/>Model Inference\n",
    "<br/>DIY MNIST Image Inference"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from data_preprocessing import unzip_file\n",
    "from data_preprocessing import to_csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "savepath = \"./data\"\n",
    "baseurl = \"http://yann.lecun.com/exdb/mnist\"\n",
    "files = [\n",
    "\t\"train-images-idx3-ubyte.gz\",\n",
    "\t\"train-labels-idx1-ubyte.gz\",\n",
    "\t\"t10k-images-idx3-ubyte.gz\",\n",
    "\t\"t10k-labels-idx1-ubyte.gz\"]\n",
    "\n",
    "unzip_file(savepath, baseurl, files)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "to_csv(\"train\", 60000)\n",
    "to_csv(\"t10k\", 10000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Random Baseline "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 9.68 %\n"
     ]
    }
   ],
   "source": [
    "!python -m score \\\n",
    "    --train_dataset 'data/train.csv' \\\n",
    "    --test_dataset 'data/t10k.csv' \\\n",
    "    --label [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Accuracy: 99.43 %\n"
     ]
    }
   ],
   "source": [
    "!python -m train \\\n",
    "    --train_dataset 'data/train.csv' \\\n",
    "    --save_model_path 'model/mnist_model.sav' \\\n",
    "    --log_path 'log/training_log.xlsx' \\\n",
    "    --label [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model Inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 97.90 %\n"
     ]
    }
   ],
   "source": [
    "!python -m inference \\\n",
    "    --test_dataset 'data/t10k.csv' \\\n",
    "    --load_model_path 'model/mnist_model.sav' \\\n",
    "    --log_path 'log/inference_log.xlsx' \\\n",
    "    --label [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### DIY MNIST Image Inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of samples: 10\n",
      "Accuracy: 40.00 %\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1100x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "!python -m diy_mnist_image_inference \\\n",
    "    --load_model_path 'model/mnist_model.sav' \\\n",
    "    --image_path './diy/original/' \\\n",
    "    --save_plot_path './diy/plot/diy_inference.png' \\\n",
    "    --log_path 'log/diy_inference_log.xlsx' \\\n",
    "    --label [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]\n",
    "\n",
    "import cv2\n",
    "import matplotlib.pyplot as plt\n",
    "plt.figure(figsize=(11, 4))\n",
    "img_path = './diy/plot/diy_inference.png'\n",
    "img = cv2.imread(img_path, cv2.IMREAD_GRAYSCALE)\n",
    "plt.imshow(img, cmap='gray')\n",
    "plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Reference"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>Data</b>\n",
    "<br>[MNIST (Original)](http://yann.lecun.com/exdb/mnist/)\n",
    "<br>[MNIST (Hugginface)](https://huggingface.co/datasets/ylecun/mnist)\n",
    "\n",
    "<br><b>Github</b>\n",
    "<br>[eesgetdegrees/format-your-own-images-to-match-MNIST-dataset](https://github.com/eesgetdegrees/format-your-own-images-to-match-MNIST-dataset)\n",
    "\n",
    "<br><b>Kaggle</b>\n",
    "<br>[MNIST SciKit-Learn Tutorial](https://www.kaggle.com/code/gainknowledge/mnist-scikit-learn-tutorial)\n",
    "<br>[MNIST | Data Augmentation | 97% Accuracy](https://www.kaggle.com/code/pranjalsriv/mnist-data-augmentation-97-accuracy)\n",
    "\n",
    "<br><b>Blog</b>\n",
    "<br>[[Machine Learning] 이미지 내부 문자 인식하기](https://teeeeeeemo.tistory.com/42)\n",
    "<br>[Save and Load Machine Learning Models in Python with scikit-learn](https://machinelearningmastery.com/save-load-machine-learning-models-python-scikit-learn/)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "exercise",
   "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": "3.11.5"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}