{ "cells": [ { "cell_type": "markdown", "metadata": { "_cell_guid": "b5fb0ca9-b8a4-02ef-f7d5-5458f927500f" }, "source": [ "# Checking conversion of numbers from 255 greyscale to black & white with different brightness thresholds" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "58166aa1-b46f-1868-adb5-a9654e0281e7" }, "outputs": [], "source": [ "# This Python 3 environment comes with many helpful analytics libraries installed\n", "# It is defined by the kaggle/python docker image: https://github.com/kaggle/docker-python\n", "# For example, here's several helpful packages to load in \n", "\n", "import numpy as np # linear algebra\n", "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n", "\n", "import matplotlib.pyplot as plt\n", "%matplotlib inline\n", "\n", "# Input data files are available in the \"../input/\" directory.\n", "# For example, running this (by clicking run or pressing Shift+Enter) will list the files in the input directory\n", "\n", "from subprocess import check_output\n", "print(check_output([\"ls\", \"../input\"]).decode(\"utf8\"))\n", "\n", "# Any results you write to the current directory are saved as output." ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "a2d209b4-e4de-c4cb-dca3-afb348636f9e" }, "outputs": [], "source": [ "# load train set\n", "trn = pd.read_csv(\"../input/train.csv\")\n", "\n", "# Split labels and images\n", "trn_img = (trn.ix[:,1:].values).astype('float32')\n", "trn_lbl = trn.ix[:,0].values.astype('int32')\n", "\n", "# Load test set\n", "tst_img = (pd.read_csv(\"../input/test.csv\").values).astype('float32')\n", "\n", "# Show a few labels to pick a subset\n", "trn_lbl[1:100]" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "afa14463-dd55-f023-1725-f7f8a8f2b58c" }, "outputs": [], "source": [ "# Take subset and reshape\n", "vis_img = trn_img[30006:30014,:]\n", "vis_lbl = trn_lbl[30006:30014]\n", "vis_img = vis_img.reshape(vis_img.shape[0], 28, 28)\n", "vis_img.shape" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "e0646680-c407-78d9-b65f-cc0e4129c4b4" }, "outputs": [], "source": [ "# Visualize greyscale numbers\n", "plt.subplots(1, 8, figsize=(12,2))\n", "for i in range(0, 8):\n", " plt.subplot(180 + (i+1))\n", " plt.imshow(vis_img[i], cmap=plt.get_cmap('gray'))\n", " plt.title(vis_lbl[i]);\n", " plt.axis('off')" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "834a4748-9442-a61c-1eeb-569a4b7926e9" }, "source": [ "## PIcked this special slice because 4th digit failed to me on prediction for 112 threshold, now I can see why, predicted was 3 ... no wonder" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "7d2e63b2-5664-9f42-077b-4eb308b9e3fa" }, "outputs": [], "source": [ "# Threshold will be from 64 brightness to 192 in steps of 16\n", "plt.subplots(9, 9, figsize=(13, 14))\n", "for k in range(0, 9):\n", "\tthreshold = k * 16 + 64\n", "\tbaw_img = vis_img.__ge__(threshold)\n", "\tplt.subplot(9,9,k*9+1)\n", "\tplt.axis('off')\n", "\tplt.title(threshold)\n", "\tfor i in range(0, 8):\n", "\t\tplt.subplot(9,9,k*9+i+2)\n", "\t\tplt.imshow(baw_img[i], cmap=plt.get_cmap('gray'))\n", "\t\tplt.axis('off')\n", "\t\tif k == 0:\n", "\t\t\tplt.title(vis_lbl[i])" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "2ce085b3-ee14-42af-016f-0fcabec37026" }, "outputs": [], "source": [ "# Picking 112 as threshold\n", "#threshold = 112\n", "#csv_trn_lbl = trn_lbl\n", "#csv_trn_img = trn_img.__ge__(threshold)\n", "#csv_tst_img = tst_img.__ge__(threshold)\n", "\n", "#np.savetxt('csv_trn_lbl.csv', csv_trn_lbl, fmt='%1d', delimiter=',')\n", "#np.savetxt('csv_trn_img.csv', csv_trn_img, fmt='%1d', delimiter=',')\n", "#np.savetxt('csv_tst_img.csv', csv_tst_img, fmt='%1d', delimiter=',')" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "17bf73df-a0bc-b43b-482c-2414bc0f36b6" }, "source": [ "Used Octave and logistic regression and splitted the training set in 30k for training and 12k for validation\n", "\n", "## 112 threshold got a 90.591667 accuracy" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "f2cf8b34-9ab2-6c1a-2915-b1e317f8958b" }, "outputs": [], "source": [ "# Picking 176 as threshold (skipping test set this time)\n", "#threshold = 176\n", "#csv_trn_lbl = trn_lbl\n", "#csv_trn_img = trn_img.__ge__(threshold)\n", "#csv_tst_img = tst_img.__ge__(threshold)\n", "\n", "#np.savetxt('csv_trn_lbl_176.csv', csv_trn_lbl, fmt='%1d', delimiter=',')\n", "#np.savetxt('csv_trn_img_176.csv', csv_trn_img, fmt='%1d', delimiter=',')\n", "#np.savetxt('csv_tst_img_176.csv', csv_tst_img, fmt='%1d', delimiter=',')" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "e7267929-5f00-8ab7-43f8-31d64f9cc7ee" }, "source": [ "## 176 threshold accuracy went down to: 89.650000" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "efeb03fd-4ed6-1c24-b1be-86c56669acde" }, "source": [ "## greyscale accuracy with same conditions is: 91.150000\n", "## thus 112 threshold was really a good aproximation" ] } ], "metadata": { "_change_revision": 0, "_is_fork": false, "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": "3.6.0" } }, "nbformat": 4, "nbformat_minor": 0 }