File size: 2,173 Bytes
1c980b1 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 |
import json
from pathlib import Path
def process_json(input_path, output_path, file_stem):
# 读取原始JSON文件
with open(input_path, 'r', encoding='utf-8') as f:
original_data = json.load(f)
processed = []
# 处理每个条目
for index, (key, item) in enumerate(original_data.items()):
# 生成媒体路径
video_id = key.split('_')[0]
media_path = "./" + (Path("data") / file_stem / video_id).as_posix()
# 处理选项
options = []
for opt_id in ['A', 'B', 'C', 'D']:
if text := item.get(f'({opt_id})', ''):
options.append({"id": opt_id, "text": text.strip()})
# 处理答案
try:
answer_num = int(item['Answer index'])
answer_ids = [options[answer_num]['id']] if 0 <= answer_num < len(options) else []
except (ValueError, IndexError, KeyError):
answer_ids = []
# 构建数据结构
processed.append({
"index": index,
"media_type": "Video",
"media_paths": media_path,
"description": item.get("Category", ""),
"task_type": "Vision-Question-Answer",
"question": [item.get("Question", "")],
"question_type": "multi-choice",
"annotations": {},
"options": options,
"answer": answer_ids,
"source": "4D-Bench",
"domain": "Embodied_ai"
})
# 保存结果
with open(output_path, 'w', encoding='utf-8') as f:
json.dump(processed, f, indent=2, ensure_ascii=False)
if __name__ == "__main__":
# 静态参数配置
input_path = "/mnt/data/users/zys/proj/vlm_reasoning/unprocessed_data/emb_ai/4d/4D_Object_Question_Answering/data/4d_qa.json" # 默认输入文件
output_path = "/mnt/data/users/zys/proj/vlm_reasoning/dataset/4D_Object_Question_Answering.json" # 默认输出文件
file_stem = "4D_Object_Question_Answering" # 专用数据集标识
# 执行处理流程
process_json(
input_path=input_path,
output_path=output_path,
file_stem=file_stem
) |