前言

在《基于人脸识别和图像分割技术制作证件照》博文中介绍了如何使用deeplab图像分割模型制作证件照，在《基于PASCAL VOC 2012数据集训练deeplab图像分割模型》博文中介绍了如何调优模型，在《使用labelme工具标注人像数据集》博文中介绍了如何标注人像数据集，今天教大家获取肖像数据。

代码

详情参考github

python">import requests
import urllib
from lxml import etree, html
import time
import os
def crawl_person_image_url(page: int = 1, header=header):
    url = f'https://www.vcg.com/creative-image/xiaoxiang/?creativeRace=1&page={page}'
    req: requests.Response = requests.get(url=url, headers=header)
    src = req.text
    root = etree.HTML(src)
    figures = root.xpath(xpath)  # 获取图像列表
    figure_url_list = []
    for fg in figures:
        try:
            img = fg.xpath('a/img')[0]  # 获取图像元素
            data = ('https:' + img.get('data-src'), 'https:' + img.get('data-min'))
            figure_url_list.append(data)
        except e:
            continue
    return figure_url_list


def download_image(url: str, save_path='./vcg/'):
    os.makedirs(save_path, exist_ok=True)
    image_name = url.split('/')[-1]
    full_save_file_name = save_path + image_name
    data = requests.get(url).content
    with open(full_save_file_name, 'wb') as f:
        f.write(data)
    return full_save_file_name


def main():
    for i in range(1, MAX_PAGE + 1):
        urls = crawl_person_image_url(i)
        for u in urls:
            download_image(u[0])
            time.sleep(0.1)  # 防止过快