mbench/make_ref-ytvos_json.py

import sys
import os
from os import path as osp
sys.path.append(osp.abspath(osp.join(osp.dirname(__file__), '..')))

from datasets import build_dataset
import argparse
import opts


from pathlib import Path
import io

import numpy as np
import pandas as pd
import regex as re
import json

import cv2
from PIL import Image, ImageDraw
import torch
from torchvision.transforms import functional as F

from skimage import measure                        # (pip install scikit-image)
from shapely.geometry import Polygon, MultiPolygon # (pip install Shapely)

import matplotlib.pyplot as plt
import matplotlib.patches as patches
from matplotlib.collections import PatchCollection
from matplotlib.patches import Rectangle


import ipywidgets as widgets
from IPython.display import display, clear_output

#==================json 만들기===================
def createJson(train_dataset, metas):
    entire_json = {}

    #초기화
    vid_idx = 0
    
    while vid_idx < len(train_dataset):

        #하나의 비디오에 대해
        video_data = {}
        video_train_frames, video_train_info = train_dataset[vid_idx]
        video_meta = metas[vid_idx]

        video_id = video_meta['video']
        video_data['bins'] = video_meta['bins']
        bin_nums = len(video_meta['bins'])
        obj_nums = max([int(k) for k in list(video_meta['obj_id_cat'].keys())])
        
        annotation_data = []
        frame_names = []

        for i in range(bin_nums):
            bin_data = {}
            for j in range(obj_nums):
                obj_id = str(j+1)
                try:
                    obj_data = {
                        "category_name":video_meta['obj_id_cat'][obj_id],
                        "bbox":video_train_info['boxes'][i*obj_nums+j, :].tolist(),
                        "valid":video_train_info['valid'][i*obj_nums+j].item()
                    }
                except:
                    obj_data = {}
                bin_data[obj_id] = obj_data
            annotation_data.append(bin_data)
        
        video_data['annotations'] = annotation_data
        

        sample_indx = metas[vid_idx]['sample_indx']
        frames = metas[vid_idx]['frames']
        for i in sample_indx:
            frame_name = frames[i]
            frame_names.append(frame_name)

        video_data['frame_names'] = frame_names
        video_data['video_path'] = os.path.join(str(train_dataset.img_folder), 'JPEGImages', video_id)               
        entire_json[video_id] = video_data

        vid_idx += 1

    return entire_json


if __name__ == '__main__':
    parser = argparse.ArgumentParser('ReferFormer training and evaluation script', parents=[opts.get_args_parser()])
    args = parser.parse_args()

    #==================데이터 불러오기===================
    # 전체 데이터셋
    train_dataset = build_dataset('ytvos_ref', image_set = 'train', args = args)

    # 전체 데이터셋 메타데이터
    metas = train_dataset.metas
    
    #==================json 만들기===================
    entire_json_dict = createJson(train_dataset, metas)
    print(type(entire_json_dict))
    entire_json = json.dumps(entire_json_dict, indent=4)

    with open('mbench/sampled_frame3.json', mode='w') as file:
        file.write(entire_json)