coco数据集标准格式转voc格式

将 COCO 格式转换为 VOC 格式可以使用以下步骤：

1. 安装 Python COCO API

pip install pycocotools

2. 创建 voc 格式目标文件夹

mkdir VOCdevkit/VOC2007
mkdir VOCdevkit/VOC2007/Annotations
mkdir VOCdevkit/VOC2007/ImageSets
mkdir VOCdevkit/VOC2007/JPEGImages

3. 将 COCO 图片复制到 JPEGImages 文件夹中

cp path/to/coco/images/* VOCdevkit/VOC2007/JPEGImages/

4. 使用以下代码将 COCO 标注文件转换为 VOC 格式的 XML 文件

import json
import xml.etree.ElementTree as ET
import os

# COCO 类别名称到 VOC 类别名称的映射
coco_to_voc = {
    "person": "person",
    "bicycle": "bicycle",
    "car": "car",
    "motorcycle": "motorbike",
    "airplane": "aeroplane",
    "bus": "bus",
    "train": "train",
    "truck": "truck",
    "boat": "boat",
    "traffic light": "traffic_light",
    "fire hydrant": "fire_hydrant",
    "stop sign": "stop_sign",
    "parking meter": "parking_meter",
    "bench": "bench",
    "bird": "bird",
    "cat": "cat",
    "dog": "dog",
    "horse": "horse",
    "sheep": "sheep",
    "cow": "cow",
    "elephant": "elephant",
    "bear": "bear",
    "zebra": "zebra",
    "giraffe": "giraffe",
    "backpack": "backpack",
    "umbrella": "umbrella",
    "handbag": "handbag",
    "tie": "tie",
    "suitcase": "suitcase",
    "frisbee": "frisbee",
    "skis": "skis",
    "snowboard": "snowboard",
    "sports ball": "sports_ball",
    "kite": "kite",
    "baseball bat": "baseball_bat",
    "baseball glove": "baseball_glove",
    "skateboard": "skateboard",
    "surfboard": "surfboard",
    "tennis racket": "tennis_racket",
    "bottle": "bottle",
    "wine glass": "wine_glass",
    "cup": "cup",
    "fork": "fork",
    "knife": "knife",
    "spoon": "spoon",
    "bowl": "bowl",
    "banana": "banana",
    "apple": "apple",
    "sandwich": "sandwich",
    "orange": "orange",
    "broccoli": "broccoli",
    "carrot": "carrot",
    "hot dog": "hot_dog",
    "pizza": "pizza",
    "donut": "donut",
    "cake": "cake",
    "chair": "chair",
    "couch": "sofa",
    "potted plant": "potted_plant",
    "bed": "bed",
    "dining table": "dining_table",
    "toilet": "toilet",
    "tv": "tv_monitor",
    "laptop": "laptop",
    "mouse": "mouse",
    "remote": "remote",
    "keyboard": "keyboard",
    "cell phone": "cell_phone",
    "microwave": "microwave",
    "oven": "oven",
    "toaster": "toaster",
    "sink": "sink",
    "refrigerator": "fridge",
    "book": "book",
    "clock": "clock",
    "vase": "vase",
    "scissors": "scissors",
    "teddy bear": "teddy_bear",
    "hair drier": "hair_drier",
    "toothbrush": "toothbrush"
}

# COCO 标注文件路径
ann_file = 'path/to/coco/annotations.json'

# VOC 目标文件夹路径
voc_dir = 'VOCdevkit/VOC2007/'

# 加载 COCO 标注文件
with open(ann_file, 'r') as f:
    coco_data = json.load(f)

# 对每个图像处理
for img_data in coco_data['images']:
    # 创建 VOC XML 文件
    xml_file = ET.Element('annotation')
    ET.SubElement(xml_file, 'folder').text = 'VOC2007'
    ET.SubElement(xml_file, 'filename').text = img_data['file_name']
    source = ET.SubElement(xml_file, 'source')
    ET.SubElement(source, 'database').text = 'COCO 2017'
    ET.SubElement(source, 'annotation').text = 'COCO 2017'
    ET.SubElement(source, 'image').text = 'Flickr'
    size = ET.SubElement(xml_file, 'size')
    ET.SubElement(size, 'width').text = str(img_data['width'])
    ET.SubElement(size, 'height').text = str(img_data['height'])
    ET.SubElement(size, 'depth').text = '3'
    ET.SubElement(xml_file, 'segmented').text = '0'

    # 查找该图像的所有标注框
    bbox_list = []
    for ann_data in coco_data['annotations']:
        if ann_data['image_id'] == img_data['id']:
            bbox = ann_data['bbox']
            bbox_list.append(bbox)

    # 对每个标注框处理
    for bbox in bbox_list:
        # 转换 COCO 格式到 VOC 格式
        x_min = bbox[0]
        y_min = bbox[1]
        x_max = bbox[0] + bbox[2]
        y_max = bbox[1] + bbox[3]
        class_name = coco_to_voc[coco_data['categories'][ann_data['category_id']-1]['name']]
        # 创建 VOC XML 标注
        obj = ET.SubElement(xml_file, 'object')
        ET.SubElement(obj, 'name').text = class_name
        ET.SubElement(obj, 'pose').text = 'Unspecified'
        ET.SubElement(obj, 'truncated').text = '0'
        ET.SubElement(obj, 'difficult').text = '0'
        bndbox = ET.SubElement(obj, 'bndbox')
        ET.SubElement(bndbox, 'xmin').text = str(int(x_min))
        ET.SubElement(bndbox, 'ymin').text = str(int(y_min))
        ET.SubElement(bndbox, 'xmax').text = str(int(x_max))
        ET.SubElement(bndbox, 'ymax').text = str(int(y_max))

    # 将 XML 文件保存到 VOC 目标文件夹中
    xml_str = ET.tostring(xml_file)
    with open(os.path.join(voc_dir, 'Annotations', img_data['file_name'].replace('.jpg', '.xml')), 'wb') as f:
        f.write(xml_str)

5. 创建 VOC 格式的图像列表文件

import os

# VOC 目标文件夹路径
voc_dir = 'VOCdevkit/VOC2007/'

# 获取所有 VOC 图像文件名
img_files = os.listdir(os.path.join(voc_dir, 'JPEGImages'))

# 创建 VOC 图像列表文件
with open(os.path.join(voc_dir, 'ImageSets', 'Main', 'trainval.txt'), 'w') as f:
    for img_file in img_files:
        img_name = img_file.replace('.jpg', '')
        f.write(img_name + '\n')

完成以上步骤，您就可以将 COCO 数据集转换为 VOC 格式了。