This code snippet is to remove labels (bounding boxes) marked as difficult in RTTS dataset. Please replace label_dir with original labels directory and out_dir is where to save newly modified labels.

remove_difficult_objects.py

import os
import xml.etree.ElementTree as ET

label_dir = '/home/mahmood/Downloads/RTTS/labels_original'
out_dir = '/home/mahmood/Downloads/RTTS/labels'

label_file_paths = os.listdir(label_dir)

print(f'Processing ...')
total_difficult = 0
for label_file in label_file_paths:
    s = os.path.join(label_dir, label_file)
    d = os.path.join(out_dir, label_file)
    tree = ET.parse(s)
    root = tree.getroot()
    for object in root.findall('object'):
        is_difficult = int(object.find('difficult').text)
        if is_difficult:
            total_difficult += 1
            root.remove(object)

    tree.write(d)
print(f'Finished. Removed total {total_difficult} difficult labels saved to {out_dir}')