电影口碑与海报图像的相关性分析

1. 通过人脸检测获取海报中人脸个数及颜色均值。

2. 分析海报人脸个数，颜色均值和评分的关系。

main.py

# -*- coding: utf-8 -*-

import pandas as pd
from tools import get_n_face, get_color_mean
import matplotlib.pyplot as plt

# 数据集路径
dataset_path = './dataset/movie_metadata.csv'

# 是否第一次运行
is_first_run = False


def run_main():
    """
        主函数
    """
    if is_first_run:
        # 第一次运行程序
        data_df = pd.read_csv(dataset_path, nrows=100)

        # 记录海报中人脸个数
        print('海报人脸检测：')
        data_df['n_face'] = data_df['movie_imdb_link'].apply(get_n_face)

        # 保存结果
        data_df.to_csv('./imdb_face.csv', index=False, encoding='utf-8')

        # 记录海报的平均像素值（这里代表图像的主要颜色，可通过颜色的表达方式替代）
        print('海报像素均值计算：')
        data_df['color_mean'] = data_df['movie_imdb_link'].apply(get_color_mean)

        # 保存结果
        data_df.to_csv('./imdb_face_color.csv', index=False, encoding='utf-8')

    data_df = pd.read_csv('./imdb_face_color.csv')
    # 分析结果
    # 人脸个数和评分的关系
    # 去除无效人脸
    data_df = data_df[data_df['n_face'] != -1]
    face_score = data_df['imdb_score'].groupby(data_df['n_face']).mean()
    face_score.name = 'Score'
    face_score.index.name = 'Number of Face'
    face_score.plot(kind='bar')
    plt.show()

    # 像素均值和评分的关系
    color_score = data_df['imdb_score'].groupby(data_df['color_mean']).mean()
    color_score.name = 'Score'
    color_score.index.name = 'Mean of color'
    color_score.plot(kind='bar')
    plt.show()


if __name__ == '__main__':
    run_main()

tools.py

# -*- coding: utf-8 -*-

from bs4 import BeautifulSoup
import urllib.request
import cognitive_face as CF
from skimage import io
import numpy as np


def get_img_link(movie_link):
    """
        通过电影的链接爬取海报的链接
    """
    movie_html = urllib.request.urlopen(movie_link)
    movie_html_obj = BeautifulSoup(movie_html, 'html.parser', from_encoding='utf-8')
    # 获取海报小图的链接
    small_poster_img_link = movie_html_obj.find('div', class_='poster').find('img')['src']

    # 获取海报大图的链接
    big_poster_img_link = small_poster_img_link[:small_poster_img_link.find('._V1_') + 4] + '.jpg'

    return big_poster_img_link


def get_n_face(movie_link):
    """
        通过图像链接获取包含的人脸个数
    """
    print('正在处理链接：', movie_link)
    img_link = get_img_link(movie_link)
    Key = 'xxxxxxxxxx '  # 这里请替换成自己申请的key
    CF.Key.set(Key)
    n_face = -1
    try:
        face_list = CF.face.detect(img_link)
        n_face = len(face_list)
        print('人脸个数：', n_face)
    except CF.util.CognitiveFaceException:
        print('无效图片')
    return n_face


def round_to_int(x, base=10):
    """
        将数字转换到最近的整数
    """
    return int(base * round(float(x)/base))


def get_color_mean(movie_link):
    """
        通过图像链接获取其平均像素值
    """
    print('正在处理链接：', movie_link)
    img_link = get_img_link(movie_link)
    image = io.imread(img_link)
    int_mean_color = round_to_int(np.mean(image))
    print('像素均值：', int_mean_color)
    return int_mean_color