最近在尝试做一个人脸识别项目，在对比几款主流人脸识别SDK后，采用了虹软的Arcface SDK，因为它提供了免费版本，并且可以离线使用，接入难度也比较低。项目中有一个需求就是显示检测到的人脸，但是如何从一张图片中抠取合适大小的人脸呢？本文将从以下步骤来介绍如何实现：
1. 如何获得人脸框
f __name__ == "__main__":
"""给定输入文件（每行一个文档，格式：id\tjson内容），请求时效性标签服务，返回时效性结果"""
in_file = 'in_sample.txt'
out_file = 'out_sample.txt'
with open(out_file, 'w') as fout:
with open(in_file, 'r') as fin:
for line in fin:
id_, info = line.strip().split('\t')
info = json.loads(info)
title = info['title']
summary = info['summary']
content = info['content']
publish_time = info['publish_time']
publish_time = time.strftime("%Y-%m-%d", time.localtime(publish_time))
fresh_res = request_fresh_level(id_, title, content, summary)
fout.write('\t'.join([id_, title, str(fresh_level), matched_words, \
audit_url, to_line(summary), to_line(content), publish_time]))
