import cv2 import json import os import numpy as np class AnnotsGTGetter: def __init__(self, cfg_obj): self.cfg_obj = cfg_obj self.img_folder_path = cfg_obj['dataset']['img_folder_path'] self.json_folder_path = cfg_obj['dataset']['annotations_folder_path'] self.annot_json_fname = cfg_obj['dataset']['annotations_fname'] self.labels_dict = cfg_obj['error_analysis']['labels_dict'] json_file = open(self.json_folder_path + self.annot_json_fname) self.annot_data = json.load(json_file) self.img_ids_in_json = [annot['image_id'] for annot in self.annot_data['annotations']] self.all_imgs = os.listdir(self.img_folder_path) return def get_imgs(self): """method to get the mutually -inclusive- images between the img_ids in json and those in the folder path not needed because all images in folder were accounted for in the json... """ all_img_ids_in_folder = [int(i[:-4]) for i in self.all_imgs] all_imgs_found = [i for i in all_img_ids_in_folder if i in self.img_ids_in_json] print (len(all_imgs_found)) def get_annots(self, img_fname = '000000576052.jpg'): """retrieve annotation given a filename Args: img_fname (_type_): image file name Returns: np array: all annotations of an image """ # change img_fname for extraction purpose # assumes jpg, png, but not jpeg... # TODO - what if jpeg? annots = [] img_id = int(img_fname[:-4]) for annot in self.annot_data['annotations']: if img_id == annot['image_id']: if annot['category_id'] in list(self.labels_dict.values()): annots.append([annot['category_id'],annot['bbox'][0],annot['bbox'][1],annot['bbox'][2],annot['bbox'][3]]) return np.array(annots) def get_gt_annots(self): """goes into the image folder, calls get_annots to extract image annotation Returns: dict: all annotations """ # create dictionary of gt annots # for img in os.listdir(self.img_folder_path): # self.get_annots(img) all_gt_annots = {img: self.get_annots(img) for img in os.listdir(self.img_folder_path)} return all_gt_annots if __name__ == '__main__': # get_annots() annots_obj = AnnotsGTGetter() gt_dict = annots_obj.get_gt_annots() print (gt_dict) # annots_obj.get_imgs()