| import json |
| import os |
| import matplotlib.pyplot as plt |
| import numpy as np |
| from PIL import Image |
|
|
|
|
| num_objects = 0 |
| num_frames = 0 |
| num_masks = 0 |
| category_item = [] |
| json_path = "/Users/gongsitong/Desktop/dataset_processing/Ref_LVOS/video_exp.json" |
| json_dict = json.load(open(json_path, 'r'))["videos"] |
| |
| |
|
|
| |
| |
| |
| category_list = [] |
| word_number = [] |
| num_objects = 0 |
| for video_name, video_dict in json_dict.items(): |
| new_obj_id = -1 |
| num_frames += len(video_dict["frames"]) |
| for exp_id, exp_dict in video_dict["expressions"].items(): |
| if type(exp_dict["category"]) != list: |
| category_list.append(exp_dict["category"]) |
| word_number.append(len(exp_dict["human_exp"].strip().split())) |
| |
| for obj_id in exp_dict["obj_id"]: |
| if int(obj_id) != new_obj_id: |
| new_obj_id = int(obj_id) |
| num_objects += 1 |
| num_masks += len(video_dict["frames"]) * (new_obj_id + 1) |
| |
| print("the categories: ", set(category_list)) |
| print("total number of objects: ", num_objects) |
| print("mean word number of references", sum(word_number) / len(word_number)) |
| print("num masks", num_masks) |
|
|
| |
| |
| |
| |
| |
| |
| |
| |
|
|
| meta_video_dict = dict() |
| vid_id = 0 |
| new_video_dict = dict() |
| for video_name, video_dict in json_dict.items(): |
| new_exp_dict = dict() |
| for exp_id, exp_dict in video_dict["expressions"].items(): |
| new_exp_dict[exp_id] = {"exp": exp_dict["human_exp"], |
| "obj_id": exp_dict["obj_id"], |
| "category": exp_dict["category"]} |
| new_video_dict[video_name] = {"expressions": new_exp_dict, |
| "vid_id": vid_id, |
| "height": video_dict["height"], |
| "width": video_dict["width"], |
| "frames": video_dict["frames"]} |
| output_dict = {"videos": new_video_dict} |
| output_path = "/Users/gongsitong/Desktop/dataset_processing/Ref_LVOS/final_meta_expressions.json" |
| with open(output_path, 'w') as f: |
| json_string = json.dumps(output_dict, indent=4) |
| f.write(json_string) |
