当前位置：首页 > ops >正文

批量清洗与修改 YOLO 标签：删除与替换指定类别

ops 2025/9/4 18:12:50

在使用 YOLO 格式的数据进行训练或部署前，常常需要对标签文件进行清洗或修改。本文整理了两种常见场景的 Python 脚本：删除指定类别 和 修改某类为其他类，并支持自动打印检测到该类别的文件名，帮助你快速定位问题数据。

📁 标签格式说明（YOLO格式）

YOLO 标签通常为 .txt 文件，每行格式如下：

<类别ID> <x_center> <y_center> <width> <height>

场景一：删除类别为 `10` 的标注，并输出相关文件名

该脚本会扫描指定目录下的所有 .txt 标签文件，删除类别为 10 的标注行，并在发现该类别时输出文件名。

import osdef process_txt_file(file_path):with open(file_path, 'r') as file:lines = file.readlines()# 检查是否存在以 "10" 开头的行has_class_10 = any(line.strip().startswith("10") for line in lines)# 过滤掉以 "10" 开头的行filtered_lines = [line for line in lines if not line.strip().startswith("10")]# 输出含有类别10的文件名if has_class_10:print(f"⚠️ 检测到类别10: {file_path}")# 重新写入过滤后的内容with open(file_path, 'w') as file:file.writelines(filtered_lines)def process_folder(folder_path):for root, dirs, files in os.walk(folder_path):for file in files:if file.endswith(".txt"):file_path = os.path.join(root, file)process_txt_file(file_path)# 修改为你自己的标签路径
folder_to_process = "/path/to/labels"
process_folder(folder_to_process)

场景二：将类别 `11` 修改为 `10`

有时我们需要将错误标注的类别进行更正，例如将所有类别为 11 的改为 10。

import osdef modify_yolo_labels(directory, old_class=11, new_class=10):for filename in os.listdir(directory):if filename.endswith(".txt"):file_path = os.path.join(directory, filename)with open(file_path, 'r') as f:lines = f.readlines()new_lines = []modified = Falsefor line in lines:parts = line.strip().split()if len(parts) > 0 and parts[0] == str(old_class):parts[0] = str(new_class)modified = Truenew_lines.append(' '.join(parts) + '\n')if modified:print(f"修改类别11为10：{file_path}")with open(file_path, 'w') as f:f.writelines(new_lines)# 替换为你的目录
label_folder = "/path/to/labels"
modify_yolo_labels(label_folder)

以下是将“删除类别”和“替换类别”功能封装为一个 YOLO 标签清洗工具类 的完整实现，便于项目中调用或集成到数据清洗流程中。 yolo_label_cleaner.py 模块：

import osclass YOLOLabelCleaner:def __init__(self, folder_path):self.folder_path = folder_pathdef delete_class(self, target_class):"""删除所有标签文件中类别为 target_class 的标注行，并输出涉及文件名。"""print(f"\n🧹 删除类别：{target_class}")for root, _, files in os.walk(self.folder_path):for file in files:if file.endswith('.txt'):file_path = os.path.join(root, file)with open(file_path, 'r') as f:lines = f.readlines()has_target = any(line.strip().startswith(str(target_class)) for line in lines)filtered = [line for line in lines if not line.strip().startswith(str(target_class))]if has_target:print(f" 检测到类别{target_class}: {file_path}")with open(file_path, 'w') as f:f.writelines(filtered)def replace_class(self, old_class, new_class):"""替换所有标签文件中 old_class 为 new_class，并输出修改的文件名。"""print(f"\n 替换类别：{old_class} ➜ {new_class}")for root, _, files in os.walk(self.folder_path):for file in files:if file.endswith('.txt'):file_path = os.path.join(root, file)with open(file_path, 'r') as f:lines = f.readlines()new_lines = []modified = Falsefor line in lines:parts = line.strip().split()if len(parts) > 0 and parts[0] == str(old_class):parts[0] = str(new_class)modified = Truenew_lines.append(' '.join(parts) + '\n')if modified:print(f"修改类别{old_class}为{new_class}: {file_path}")with open(file_path, 'w') as f:f.writelines(new_lines)

使用示例

from yolo_label_cleaner import YOLOLabelCleaner# 指定标签路径
label_dir = "/mnt/data/code/szl/dataset/self-driving/SelfDriving/test/labels"
cleaner = YOLOLabelCleaner(label_dir)# 删除类别10
cleaner.delete_class(10)# 替换类别11为10
cleaner.replace_class(11, 10)

如果需要替换多类别，可以进行下面的修改：

import osclass YOLOLabelCleaner:def __init__(self, folder_path):self.folder_path = folder_pathdef delete_class(self, target_class):print(f"\n 删除类别：{target_class}")for root, _, files in os.walk(self.folder_path):for file in files:if file.endswith('.txt'):file_path = os.path.join(root, file)with open(file_path, 'r') as f:lines = f.readlines()has_target = any(line.strip().startswith(str(target_class)) for line in lines)filtered = [line for line in lines if not line.strip().startswith(str(target_class))]if has_target:print(f"检测到类别{target_class}: {file_path}")with open(file_path, 'w') as f:f.writelines(filtered)def replace_class(self, old_class, new_class):print(f"\n替换类别：{old_class} ➜ {new_class}")self.replace_classes_bulk({old_class: new_class})def replace_classes_bulk(self, class_map: dict):"""批量替换多个类别，class_map 格式为 {old_class: new_class}"""print(f"\n批量替换类别: {class_map}")for root, _, files in os.walk(self.folder_path):for file in files:if file.endswith('.txt'):file_path = os.path.join(root, file)with open(file_path, 'r') as f:lines = f.readlines()modified = Falsenew_lines = []for line in lines:parts = line.strip().split()if parts and int(parts[0]) in class_map:old = parts[0]parts[0] = str(class_map[int(old)])modified = Truenew_lines.append(' '.join(parts) + '\n')if modified:print(f"✅ 替换成功: {file_path}")with open(file_path, 'w') as f:f.writelines(new_lines)

使用示例（批量替换）

from yolo_label_cleaner import YOLOLabelCleaner# 初始化
label_dir = "/test/labels"
cleaner = YOLOLabelCleaner(label_dir)# 替换多个类别：11 ➜ 10, 12 ➜ 9, 13 ➜ 8
class_map = {11: 10,12: 9,13: 8
}
cleaner.replace_classes_bulk(class_map)