学习笔记——一些数据转换脚本（Python）

注：json 文件是 X-Anylabeling 多边形矩阵的标注。

json2YOLO（txt）

自定义name2id
自定义json_floder_path、txt_outer_path
保证存放 txt 的文件夹存在

# 处理 X-Anylabeling 多边形矩阵的标注 json 转化 txt，提取点
import json
import osname2id = {'crack_concrete': 4}  # 修改你的类别并且赋与 indexdef decode_json(json_floder_path, txt_outer_path, json_name):txt_name = os.path.join(txt_outer_path,json_name[:-5]) + '.txt'with open(txt_name, 'a') as f:json_path = os.path.join(json_floder_path, json_name)data = json.load(open(json_path, 'r', encoding='gb2312', errors='ignore'))img_w = data['imageWidth']img_h = data['imageHeight']isshape_type = data['shapes'][0]['shape_type']print(isshape_type)dw = 1. / (img_w)dh = 1. / (img_h)for i in data['shapes']:label_name = i['label']if (i['shape_type'] == 'polygon'):point = []for lk in range(len(i['points'])):x = float(i['points'][lk][0])y = float(i['points'][lk][1])point_x = x * dwpoint_y = y * dhpoint.append(point_x)point.append(point_y)f.write(str(name2id[label_name]) + " " + " ".join([str(a) for a in point]) + '\n')f.close()if __name__ == "__main__":json_floder_path = r'D:\JetBrains\PyCharm 2023.2\PycharmProjects\YOLOv8\datasets\crack_concrete_hybrid_augmentation\jsons'  # 存放 json 的文件夹的绝对路径txt_outer_path = r'D:\JetBrains\PyCharm 2023.2\PycharmProjects\YOLOv8\datasets\crack_concrete_hybrid_augmentation\labels'  # 存放 txt 的文件夹绝对路径json_names = os.listdir(json_floder_path)flagcount = 0for json_name in json_names:decode_json(json_floder_path, txt_outer_path, json_name)flagcount += 1print('-----------转化完毕------------')

VOC（xml）2YOLO（txt）

自定义class_mapping
自定义input_folder、output_folder
不必确保output_folder文件夹存在，程序会自动创建

import xml.etree.ElementTree as ET
import os# 标签映射，将类别名称映射到类别 ID
class_mapping = {"person": 0,"helmet": 1,"life jacket": 2,"truck": 3,"excavator": 4,"car crane": 5,"crawler crane": 6,"rotary drill rig": 7,"concrete tanker": 8,# 添加更多类别映射
}# 输入和输出文件夹路径
input_folder = "../datasets/drone/labels/convert"
output_folder = "../datasets/drone/labels/train"# 遍历 XML 文件
for xml_file in os.listdir(input_folder):if xml_file.endswith(".xml"):xml_path = os.path.join(input_folder, xml_file)tree = ET.parse(xml_path)root = tree.getroot()# 获取图像宽度和高度width = int(root.find("size/width").text)height = int(root.find("size/height").text)# 创建输出文件的路径和名称txt_file = os.path.splitext(xml_file)[0] + ".txt"txt_path = os.path.join(output_folder, txt_file)# 打开输出文件并写入 YOLO 格式数据with open(txt_path, "w") as f:for obj in root.findall("object"):class_name = obj.find("name").textclass_id = class_mapping.get(class_name)if class_id is not None:bbox = obj.find("bndbox")x_center = (float(bbox.find("xmin").text) + float(bbox.find("xmax").text)) / 2.0 / widthy_center = (float(bbox.find("ymin").text) + float(bbox.find("ymax").text)) / 2.0 / heightwidth_norm = (float(bbox.find("xmax").text) - float(bbox.find("xmin").text)) / widthheight_norm = (float(bbox.find("ymax").text) - float(bbox.find("ymin").text)) / heightf.write(f"{class_id} {x_center:.6f} {y_center:.6f} {width_norm:.6f} {height_norm:.6f}\n")

image2h5

自定义image_folder、h5_folder
不必确保h5_folder文件夹存在，程序会自动创建

import h5py
from PIL import Image
import numpy as np
import osdef image_to_h5(image_folder, h5_folder):"""图片转 h5"""# 确保输出文件夹存在if not os.path.exists(h5_folder):os.makedirs(h5_folder)# 获取输入文件夹中的所有文件files = os.listdir(image_folder)total_files = len(files)for i, file_name in enumerate(files, 1):# 检查文件扩展名，如果是图像文件则进行处理if file_name.endswith('.png') or file_name.endswith('.jpg'):input_image_path = os.path.join(image_folder, file_name)output_h5_path = os.path.join(h5_folder, file_name.replace('.png', '.h5'))  # 根据文件名生成输出路径# 打开图像并转换为 NumPy 数组image = Image.open(input_image_path)image_array = np.array(image)# 创建 HDF5 文件并写入数据with h5py.File(output_h5_path, 'w') as hf:hf.create_dataset('image', data=image_array)# 输出处理进度print(f"Processed {i}/{total_files} images. Current image: {file_name}")if __name__ == "__main__":image_folder = '../data/crack_concrete/images'  # 输入文件夹路径h5_folder = '../data/crack_concrete/h5s'  # 输出文件夹路径image_to_h5(image_folder, h5_folder)

json2npz

自定义json_folder、npz_folder
不必确保npz_folder文件夹存在，程序会自动创建

import os
import json
import numpy as np
from PIL import Image, ImageDrawdef load_json(json_path):"""加载 json 文件"""with open(json_path, 'r') as f:return json.load(f)def create_image(data):"""读取图像数据"""image = Image.open(data['imagePath'])return np.array(image, dtype=np.float32)def create_label(data):"""创建与图像相同尺寸的空白 mask"""points = data['shapes'][0]['points']points_tuple = [(float(point[0]), float(point[1])) for point in points]label = Image.new('L', (data['imageWidth'], data['imageHeight']), 0)draw = ImageDraw.Draw(label)draw.polygon(points_tuple, fill=1)return np.array(label, dtype=np.float32)def json_to_npz(json_folder, npz_folder):"""将文件夹中的所有 json 文件转换为 npz 文件"""if not os.path.exists(npz_folder):os.makedirs(npz_folder)json_files = [file for file in os.listdir(json_folder) if file.endswith('.json')]total_files = len(json_files)for i, json_file in enumerate(json_files, 1):json_path = os.path.join(json_path, json_file)json_data = load_json(json_path)image_array = create_image(json_data)label_array = create_label(json_data)npz_file_path = os.path.join(npz_folder, json_file.split('.')[0] + '.npz')np.savez(npz_file_path, image=image_array, label=label_array)# 输出处理进度print(f"Processed {i}/{total_files} JSON files. Current file: {json_file}")if __name__ == "__main__":json_folder = '../data/crack_concrete/jsons'npz_folder = '../data/crack_concrete/npzs'json_to_npz(json_folder, npz_folder)