datasets/build_ref_ego_objname.py

import json
from natsort import natsorted
import os 
import re

#json_path = "/home/yuqian_fu/Projects/PSALM/check_text_byname_600_objname_llavaname.json"
json_path = "/work/yuqian_fu/Data/datasets/DAVIS/2017/trainval_test_psalm_20gap.json"
# data_path = "/work/yuqian_fu/Ego/data_segswap_test"
with open(json_path, "r") as fp:
    datas = json.load(fp)
# split_path = "/home/yuqian_fu/Projects/ego-exo4d-relation/correspondence/split.json"
# with open(split_path, "r") as fp:
#     takes_all = json.load(fp)
# takes = takes_all["test"]
#names = ["basketball", "bike", "cooking", "health", "music", "soccer"]
#names = ["basketball", "bike", "music", "soccer"]


#统计这批数据中的take_id
# takes = []
# for name in names:
#     data = datas[name]
#     for d in data:
#         video_name = d["video_name"]
#         takes.append(video_name)
# takes = list(set(takes))
# print("takes:", len(takes))


#重新格式化数据，每个数据中增加新键：scene
# datas_new = []
# for name in names:
#     data_this_name = datas[name]
#     for data in data_this_name:
#         data["scene"] = name
#         datas_new.append(data)
# print("datas_new:", len(datas_new))


datas_final_save = []
# # 给datas_new中的每个数据添加一个新的键：obj_name
# for take in takes:
#     # datas_tmp = []
#     gt_path = f"{data_path}/{take}/annotation.json"
#     with open(gt_path, 'r') as fp:
#         gt = json.load(fp)
#     # 创建逆字典
#     objs = natsorted(list(gt["masks"].keys()))
#     coco_id_to_cont_id = {cont_id + 1: coco_id for cont_id, coco_id in enumerate(objs)}
  

#     # for data in datas_new:
#     #     if data["video_name"] == take:
#     #         datas_tmp.append(data)
    
for data in datas:
    # 只处理属于当前 take 的数据
    # if data.get("video_name") != take:
    #     continue
        
    annos = data["first_frame_anns"]
    annos_new = []
    for anno in annos:
        category_id = int(anno["category_id"])  # 转换为整数
        obj_name = "obj"
        anno["text"] = {}
        anno["text"] = obj_name
        annos_new.append(anno)
    data["first_frame_anns"] = annos_new
    datas_final_save.append(data)



# datas_final_save = []
# for data in datas:
#     annos = data["first_frame_anns"]
#     annos_new = []
#     for anno in annos:
#         obj_name = anno["obj_name"]
#         result = re.sub(r'_\d+$', '', obj_name) 
#         anno["obj_name"] = result
#         llava_text = anno["llava_text"]
#         # 如果llava_text中有表示a、an的量词，比如“a basketball”，则删除”a“ 变为”baskerball“
#         if "a " in llava_text:
#             llava_text = llava_text.replace("a ", "")
#         elif "an " in llava_text:
#             llava_text = llava_text.replace("an ", "")
#         anno["llava_text"] = llava_text
#         annos_new.append(anno)
#     data["first_frame_anns"] = annos_new
#     datas_final_save.append(data)



print("datas_final_save:", len(datas_final_save))
save_path = "/home/yuqian_fu/Projects/DAVIS_test_gap20.json"
with open(save_path, "w") as fp:
    json.dump(datas_final_save, fp)




            





        
# path = "/home/yuqian_fu/Projects/ego-exo4d-relation/correspondence/split.json"
# with open(path, "r") as fp:
#     data = json.load(fp)
# takes = data["test"]
# print("takes:", len(takes))


# takes_new = os.listdir("/work/yuqian_fu/Ego/data_segswap_test_20250508_new888_v2")
# print("takes_new:", len(takes_new))

# # 找出 takes 中有但 takes_new 中没有的元素
# diff_in_takes = set(takes) - set(takes_new)

# # 找出 takes_new 中有但 takes 中没有的元素
# diff_in_takes_new = set(takes_new) - set(takes)

# # 打印结果
# print("takes 中有但 takes_new 中没有的元素:", diff_in_takes)
# print("takes_new 中有但 takes 中没有的元素:", diff_in_takes_new)