羽毛球数据集下载

import json import requests from pathlib import Path from tqdm import tqdm import concurrent.futures # 配置 NDJSON_FILE dataset.ndjson # 你的ndjson文本保存文件名 SAVE_ROOT Path(shuttlecock_dataset) THREAD_COUNT 10 # 并发下载线程 # # 创建目录 for part in [train, val, test]: (SAVE_ROOT / images / part).mkdir(parentsTrue, exist_okTrue) (SAVE_ROOT / labels / part).mkdir(parentsTrue, exist_okTrue) def task_download(line): try: data json.loads(line.strip()) if data[type] ! image: return img_url data[url] img_name data[file] split data[split] boxes data[annotations][boxes] img_save_path SAVE_ROOT / images / split / img_name label_save_path SAVE_ROOT / labels / split / Path(img_name).with_suffix(.txt) # 下载图片 resp requests.get(img_url, timeout20) if resp.status_code 200: with open(img_save_path, wb) as f: f.write(resp.content) else: print(f下载失败 {img_name} code:{resp.status_code}) return # 写入YOLO标签 label_lines [] for box in boxes: cls, cx, cy, w, h box label_lines.append(f{cls} {cx:.6f} {cy:.6f} {w:.6f} {h:.6f}) with open(label_save_path, w, encodingutf-8) as f: f.write(\n.join(label_lines)) except Exception as err: print(f异常: {err}) if __name__ __main__: with open(NDJSON_FILE, r, encodingutf-8) as f: lines [ln for ln in f.readlines() if ln.strip()] print(f总共待处理图片{len(lines)}) with concurrent.futures.ThreadPoolExecutor(max_workersTHREAD_COUNT) as pool: list(tqdm(pool.map(task_download, lines), totallen(lines))) print(✅ 所有任务执行完毕)