将COCO格式的目标检测框标注文件转成VOC格式

文件目录如下：

如图，目的是把COCO文件夹下的子文件夹内的标注文件（图片和标注文件都在这些子文件夹下）转换成VOC格式，然后统一重命名放到VOC文件夹下。
main代码：

import os
import numpy as np
import codecs
import json
from glob import glob
import cv2
import shutil

BASEDIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))

# src_dir #coco格式标注的数据
src_dir = os.path.join(BASEDIR, "COCO")

# target_dir # VOC格式标注数据
xml_dir = os.path.join(BASEDIR, "VOC", "labels")
target_image_dir = os.path.join(BASEDIR, "VOC", "images")

if not os.path.exists(xml_dir):
    os.makedirs(xml_dir)
if not os.path.exists(target_image_dir):
    os.makedirs(target_image_dir)

# 将src_dir中COCO标注数据转成VOC标注数据，将VOC格式标注文件以及对应图片存储到target_dir,图片名字被修改
class_dirs = os.listdir(src_dir)  # smoke, fall, phone
count = 0  # 用于计数和图片的命名
for class_dir in class_dirs:
    json_names = list(filter(lambda x: x.endswith('.json'), os.listdir(os.path.join(src_dir, class_dir))))
    for json_name in json_names:
        json_path = os.path.join(src_dir, class_dir, json_name)
        json_info = json.load(open(json_path, "r", encoding="utf-8"))
        # print(json_info)
        # print(json_info["shapes"][0]["points"][0][0])
        # points = np.array(json_info["shapes"][0]["points"])
        # label = json_info["shapes"][0]["label"]
        # xmin = min(points[:, 0])
        # xmax = max(points[:, 0])
        # ymin = min(points[:, 1])
        # ymax = max(points[:, 1])
        # print(xmin,ymin,xmax,ymax)
        src_image_name = json_info["imagePath"]  # src_dir中图片的名字
        src_image_path = os.path.join(src_dir, class_dir, src_image_name)
        # print(os.path.join(src_dir, class_dir, src_image_name))
        height, width, channel = cv2.imread(src_image_path).shape

        #zfill用于把数字变成如00001的形式
        target_image_name ="{}.jpg".format(str(count).zfill(5))  # target_dir中图片的名字
        target_image_path = os.path.join(target_image_dir, target_image_name)
        print("将图片{}复制到{}".format(src_image_path, target_image_path))
        shutil.copy(src_image_path, target_image_path)
        xml_name = "{}.xml".format(str(count).zfill(5))          # target_dir中图片对应的xml文件的名字
        # print(target_image_name, xml_name)
        xml_path = os.path.join(xml_dir, xml_name)
        print("将图片{}的标注信息写入到{}".format(target_image_path, xml_path))
        with codecs.open(xml_path, "w", "utf-8") as xml:
            xml.write('\n')
            xml.write('\t' + 'human_action' + '\n')
            xml.write('\t' + target_image_name + '\n')
            xml.write('\t\n')
            xml.write('\t\t' + str(width) + '\n')
            xml.write('\t\t' + str(height) + '\n')
            xml.write('\t\t' + str(channel) + '\n')
            xml.write('\t\n')
            for bbox_shape in json_info["shapes"]: #有多个框
                points = np.array(bbox_shape["points"])
                label = bbox_shape["label"]
                xmin = min(points[:, 0])
                xmax = max(points[:, 0])
                ymin = min(points[:, 1])
                ymax = max(points[:, 1])
                if xmax <= xmin:
                    pass
                elif ymax <= ymin:
                    pass
                else:
                    xml.write('\t\n')
            xml.write('')
        print("已完成第{}张图片".format(count))
        count += 1