flag.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-

# --- 1.必要なライブラリのインポート
import urllib
import cv2
import os
import numpy as np
import matplotlib.pyplot as plt
from bs4 import BeautifulSoup

# --- 2. 関数
def maybe_download(url, name):
    """ ファイルが存在しない場合のみダウンロードしてくる """  
    if not os.path.exists(name):
        urllib.urlretrieve(url, name)
        print "File " + name + " downloaded!"
    else:
        print "File " + name + " already exists!"
    return name

def parse_html(name):
    """ Wikipedia「国旗の一覧」から国旗画像のURLを取り出す """
    soup = BeautifulSoup(open(name,'r').read(), "lxml")
    img = soup.find_all("img")[:206]
    return ["https:"+i.attrs["src"] for i in img]

if __name__ == '__main__':
    # --- 3. 国旗の画像をダウンロードする
    # URL of 国旗の一覧
    url = "https://ja.wikipedia.org/wiki/%E5%9B%BD%E6%97%97%E3%81%AE%E4%B8%80%E8%A6%A7"
    # download html
    html_name = maybe_download(url, "flag_list.html")
    # parse html
    flag_urls = parse_html(html_name)
    # download figs
    for u in flag_urls:
        png_name = u.split("px-")[-1].split(".")[0] + ".png"
        maybe_download(u, os.path.join("./figs", png_name))

    # --- 4. 国旗画像をNumPyのArrayとして取り込む(OpenCV)
    flag_list = []
    for fig_name in os.listdir("./figs"):
        img = cv2.imread(os.path.join("./figs", fig_name))
        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) # default: BGR
        flag_list.append(img)
    flag_list = np.array(flag_list)

    # --- 5. 国旗を全て表示してみる
    fig = plt.figure(figsize=(12, 15))
    fig.subplots_adjust(left=0, right=1, bottom=0, top=0.5, hspace=0.05, wspace=0.05)
    for i in xrange(206):
        ax = fig.add_subplot(15, 15, i + 1, xticks=[], yticks=[])
        ax.imshow(flag_list[i], interpolation="none")

    # --- 6. 各色の出現頻度を調べる
    hue_list = np.array([cv2.cvtColor(f, cv2.COLOR_RGB2HSV)[:,:,0] for f in flag_list])

    hist = np.zeros(180)
    for h in hue_list:
        hist += np.histogram(h, bins=np.arange(181))[0]

    fig = plt.figure(figsize=(12, 5))        
    plt.bar(np.arange(180)+1, hist)
    plt.show()

    # --- 7. JSONで色相の頻度を保存
    nhist = hist / np.max(hist) # normalize
    f = open("hue.js", "w")
    f.write("data=")
    json.dump(dict(zip(np.arange(180), nhist)), f)
    f.close()