以下教程用于将使用Labelme软件标注生成的json格式文件转成YOLO使用的txt格式，适用场景：矩形框

        使用方法：将json文件夹路径填到json_dir后， 将保存转化后txt的路径填入txt_dir后，运行即可转化。

        运行完成后会生成labelclassess.txt文件，其中保存了标签名称，这里的名称与转换后的txt对应。

import os
import cv2
import glob
import json
import numpy as np
from tqdm import tqdm# 参考链接 https://zhuanlan.zhihu.com/p/622246547# 将labelme_json标注转yolo_txt
def convert(size, box):"""convert [xmin, xmax, ymin, ymax] to [x_centre, y_centre, w, h]"""dw = 1. / size[0]dh = 1. / size[1]x = (box[0] + box[1]) / 2.0y = (box[2] + box[3]) / 2.0w = box[1] - box[0]h = box[3] - box[2]x = x * dww = w * dwy = y * dhh = h * dhreturn [x, y, w, h]if __name__ == "__main__":class_names = []# json文件夹路径json_dir = r"O:\DeepLearningTool\01_handle_dataset\dataset\object\image"# 转化后txt的保存路径txt_dir = r"O:\DeepLearningTool\01_handle_dataset\dataset\object\label"if not os.path.exists(txt_dir):os.makedirs(txt_dir, exist_ok=True)json_pths = glob.glob(json_dir + "/*.json")for json_pth in tqdm(json_pths,  desc='Processing'):f1 = open(json_pth, "r")json_data = json.load(f1)img_pth = os.path.join(json_dir, json_pth.replace("json", "jpg"))img = cv2.imread(img_pth)h, w = img.shape[:2]tag = os.path.basename(json_pth)out_file = open(os.path.join(txt_dir, tag.replace("json", "txt")), "w")label_infos = json_data["shapes"]for label_info in label_infos:label = label_info["label"]points = label_info["points"]if len(points) >= 3:points = np.array(points)xmin, xmax = max(0, min(np.unique(points[:, 0]))), min(w, max(np.unique(points[:, 0])))ymin, ymax = max(0, min(np.unique(points[:, 1]))), min(h, max(np.unique(points[:, 1])))elif len(points) == 2:x1, y1 = points[0]x2, y2 = points[1]xmin, xmax = min(x1, x2), max(x1, x2)ymin, ymax = min(y1, y2), max(y1, y2)else:continuebbox = [xmin, xmax, ymin, ymax]bbox_ = convert((w,h), bbox)if label not in class_names:class_names.append(label)cls_id = class_names.index(label)out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bbox_]) + '\n')with open(txt_dir + 'classes.txt', 'w') as f:for i in class_names:f.write(i + '\n')