labelme_to_coco(多个json文件转换成一个）

Question

我是用labelme标注的动物数据集生成多个身体关键点json文件，然後我想转换成coco格式并合并成一個类似animal_keypoints_train.json文件（类似下面的配置信息）

{"image_id": 1448, "bbox": [234, 11, 344, 248], "keypoints": [[298, 45, 1], [274, 49, 1], [300, 59, 1], [294, 29, 1], [259, 32, 1], [293, 150, 1], [256, 156, 1], [330, 166, 1], [307, 173, 1], [296, 207, 1], [267, 206, 1], [325, 212, 1], [0, 0, 0], [292, 239, 1], [261, 239, 1], [320, 232, 1], [292, 234, 1], [274, 106, 1], [0, 0, 0], [0, 0, 0]], "num_keypoints": 20, "category_id": 1}, {"image_id": 1449, "bbox": [140, 26, 500, 333], "keypoints": [[257, 130, 1], [167, 113, 1], [177, 176, 1], [338, 113, 1], [173, 76, 1], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [262, 242, 1], [376, 168, 1], [0, 0, 0]], "num_keypoints": 20, "category_id": 1}, {"image_id": 1450, "bbox": [79, 1, 435, 334], "keypoints": [[336, 144, 1], [208, 124, 1], [298, 183, 1], [357, 101, 1], [147, 71, 1], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0], [249, 318, 1], [0, 0, 0], [0, 0, 0]], "num_keypoints": 20, "category_id": 2}, {"image_id": 1451, "bbox": [110, 158, 299, 267], "keypoints": [[0, 0, 0], [279, 210, 1], [286, 220, 1], [266, 169, 1], [266, 187, 1], [0, 0, 0], [235, 253, 1], [173, 198, 1], [116, 242, 1], [0, 0, 0], [273, 246, 1], [0, 0, 0], [0, 0, 0], [0, 0, 0], [291, 246, 1], [0, 0, 0], [0, 0, 0], [250, 226, 1], [213, 214, 1], [128, 195, 1]], "num_keypoints": 20, "category_id": 2},
{"supercategory": "animal", "id": 4, "name": "horse", "keypoints": ["left_eye", "right_eye", "nose", "left_ear", "right_ear", "left_front_elbow", "right_front_elbow", "left_back_elbow", "right_back_elbow", "left_front_knee", "right_front_knee", "left_back_knee", "right_back_knee", "left_front_paw", "right_front_paw", "left_back_paw", "right_back_paw", "throat", "withers", "tailbase"], "skeleton": [[0, 1], [0, 2], [1, 2], [0, 3], [1, 4], [2, 17], [18, 19], [5, 9], [6, 10], [7, 11], [8, 12], [9, 13], [10, 14], [11, 15], [12, 16]]},

Answer 1

结全ChatGPT和我自己的想法：

将从LabelMe标注生成的多个JSON文件转换为单个COCO格式的JSON文件，您可以按照以下步骤使用Python进行操作。我将为您提供一个脚本，该脚本会读取多个LabelMe JSON文件，提取相关信息，然后创建一个单独的COCO格式的JSON文件。

假设您有一个包含LabelMe JSON文件的目录，以及您提供的COCO格式JSON模式，您可以使用以下脚本：

import json
import os

# LabelMe JSON文件列表
labelme_json_files = [
    'path/to/labelme_file1.json',
    'path/to/labelme_file2.json',
    # 根据需要添加更多文件路径
]

# COCO格式骨架和关键点信息
coco_keypoints_info = {
    # 在此处填写您的COCO关键点信息
}

# 初始化COCO注释格式
coco_data = {
    "images": [],
    "annotations": [],
    "categories": [coco_keypoints_info],
}

# 加载并处理每个LabelMe JSON文件
image_id_counter = 1
annotation_id_counter = 1

for labelme_file_path in labelme_json_files:
    with open(labelme_file_path, 'r') as f:
        labelme_data = json.load(f)

    # 提取图像信息
    image_info = {
        "id": image_id_counter,
        "width": labelme_data['imageWidth'],
        "height": labelme_data['imageHeight'],
        "file_name": os.path.basename(labelme_file_path).replace('.json', '.jpg'),
    }
    coco_data['images'].append(image_info)

    # 处理注释
    for annotation in labelme_data['shapes']:
        # 提取注释信息
        bbox = annotation['points']
        keypoints = []  # 从LabelMe格式提取关键点
        # 将关键点转换为COCO格式

        # 创建COCO注释
        coco_annotation = {
            "id": annotation_id_counter,
            "image_id": image_id_counter,
            "category_id": 1,  # 根据需要替换为适当的category_id
            "bbox": [bbox[0][0], bbox[0][1], bbox[1][0] - bbox[0][0], bbox[1][1] - bbox[0][1]],
            "keypoints": keypoints,
            "num_keypoints": len(keypoints) // 3,
        }
        coco_data['annotations'].append(coco_annotation)

        annotation_id_counter += 1

    image_id_counter += 1

# 保存COCO格式JSON
output_json_path = 'path/to/animal_keypoints_train.json'
with open(output_json_path, 'w') as f:
    json.dump(coco_data, f, indent=4)

请注意，您需要根据实际的目录结构、路径和数据格式调整此脚本。此外，请确保您根据LabelMe数据正确填充脚本中的keypoints列表。