如题，假设需要删除的标签名称为"area", "exit"。沿用前一篇文章中提到的文件结构。则批量删除labelme标签可使用如下代码：

批量删除labelme标签：

import os
import json
 
# 基本路径
base_path = "E:/yolov8/20240624/escala/"
label_dirs = ["labelmes/test", "labelmes/train", "labelmes/val"]
new_label_dirs = ["labelmes_new/test", "labelmes_new/train", "labelmes_new/val"]
 
# 创建新的标注目录
for new_label_dir in new_label_dirs:
    os.makedirs(os.path.join(base_path, new_label_dir), exist_ok=True)
 
# 需要删除的标签
labels_to_remove = {"area", "exit"}
 
# 遍历所有标注目录
for label_dir, new_label_dir in zip(label_dirs, new_label_dirs):
    label_dir_path = os.path.join(base_path, label_dir)
    new_label_dir_path = os.path.join(base_path, new_label_dir)
 
    for label_file in os.listdir(label_dir_path):
        if label_file.endswith('.json'):
            labelme_anno_path = os.path.join(label_dir_path, label_file)
            new_labelme_anno_path = os.path.join(new_label_dir_path, label_file)
 
            with open(labelme_anno_path, 'r') as f:
                labelme_annotations = json.load(f)
 
            # 过滤掉指定标签
            new_shapes = [shape for shape in labelme_annotations["shapes"] if shape["label"] not in labels_to_remove]
            labelme_annotations["shapes"] = new_shapes
 
            # 将处理后的标注文件保存到新的目录
            with open(new_labelme_anno_path, 'w') as f:
                json.dump(labelme_annotations, f, indent=2)
 
            print(f"Processed {label_file} and saved to {new_labelme_anno_path}")