{ "cells": [ { "cell_type": "markdown", "metadata": { "_cell_guid": "3505f1d1-44c4-1bf2-e108-a4d08901e4bc" }, "source": [ " " ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "f53a90fb-2076-63d1-9118-53d507da373d" }, "outputs": [], "source": [ "from sklearn.ensemble import ExtraTreesRegressor\n", "from sklearn.metrics import mean_squared_error\n", "from multiprocessing import Pool, cpu_count\n", "from PIL import Image, ImageStat\n", "import matplotlib.pyplot as plt\n", "%matplotlib inline\n", "import pandas as pd\n", "import numpy as np\n", "import glob, os\n", "\n", "def im_multi(path):\n", " im_stats_im_ = Image.open(path)\n", " return [path, {'Stats': ImageStat.Stat(im_stats_im_), 'Size': im_stats_im_.size}]\n", "\n", "def im_stats(im_stats_df):\n", " im_stats_d = {}\n", " p = Pool(cpu_count())\n", " ret = p.map(im_multi, im_stats_df['path'])\n", " for i in range(len(ret)):\n", " im_stats_d[ret[i][0]] = ret[i][1]\n", " im_stats_df['size_0'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Size'][0])\n", " im_stats_df['size_1'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Size'][1])\n", " print('size...')\n", " im_stats_df['sum_0'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].sum[0])\n", " im_stats_df['sum_1'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].sum[1])\n", " im_stats_df['sum_2'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].sum[2])\n", " print('sum...')\n", " im_stats_df['mean_0'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].mean[0])\n", " im_stats_df['mean_1'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].mean[1])\n", " im_stats_df['mean_2'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].mean[2])\n", " print('mean...')\n", " im_stats_df['rms_0'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].rms[0])\n", " im_stats_df['rms_1'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].rms[1])\n", " im_stats_df['rms_2'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].rms[2])\n", " print('var...')\n", " im_stats_df['var_0'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].var[0])\n", " im_stats_df['var_1'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].var[1])\n", " im_stats_df['var_2'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].var[2])\n", " print('stddev...')\n", " im_stats_df['stddev_0'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].stddev[0])\n", " im_stats_df['stddev_1'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].stddev[1])\n", " im_stats_df['stddev_2'] = im_stats_df['path'].map(lambda x: im_stats_d[x]['Stats'].stddev[2])\n", " print(len(im_stats_df))\n", " #add fun features here\n", " return im_stats_df\n", "\n", "train = pd.read_csv('../input/Train/train.csv')\n", "train['path'] = train.train_id.map(lambda x: '../input/Train/' + str(x) + '.jpg')\n", "train['dpath'] = train.train_id.map(lambda x: '../input/TrainDotted/' + str(x) + '.jpg')\n", "train = train[train.path.isin(glob.glob('../input/Train/*.jpg'))] #restrict\n", "test = pd.read_csv('../input/sample_submission.csv')\n", "test['path'] = test.test_id.map(lambda x: '../input/Test/' + str(x) + '.jpg')\n", "\n", "print(len(train),len(test))\n", "\n", "train = im_stats(train)\n", "#test = im_stats(test)\n", "\n", "colx = [c for c in train.columns if c not in ['train_id', 'path', 'dpath','adult_males', 'subadult_males', 'adult_females', 'juveniles', 'pups']]\n", "coly = ['adult_males', 'subadult_males', 'adult_females', 'juveniles', 'pups']\n", "\n", "dmean = train[colx].mean(axis=0)\n", "dmedian = train[colx].mean(axis=0)\n", "for c in colx:\n", " test[c] = int((dmean[c] + dmedian[c]) / 2)\n", " \n", "rfr = ExtraTreesRegressor(n_estimators=100, max_depth=6, n_jobs=-1, random_state=0, verbose=0)\n", "rfr.fit(train[colx], train[coly])\n", "print(mean_squared_error(train[coly], rfr.predict(train[colx]))**0.5)\n", "\n", "pred = pd.DataFrame(rfr.predict(test[colx])).astype(int)\n", "test[coly[0]] = pred[0]\n", "test[coly[1]] = pred[1]\n", "test[coly[2]] = pred[2]\n", "test[coly[3]] = pred[3]\n", "test[coly[4]] = pred[4]\n", "\n", "test[['test_id','adult_males', 'subadult_males', 'adult_females', 'juveniles', 'pups']].to_csv('z01_submission01.csv', index=False)\n", "\n", "img = glob.glob('../input/Train/*.jpg') + glob.glob('../input/TrainDotted/*.jpg')\n", "plt.rcParams['figure.figsize'] = (12.0, 10.0)\n", "for i in range(20):\n", " plt.subplot(5, 4, i+1)\n", " im = Image.open(img[i])\n", " plt.imshow(im); plt.axis('off')\n", " \n", " \n", " " ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "3400a271-c270-4b70-111b-c6c90263d352" }, "outputs": [], "source": [ "" ] } ], "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 }