Excuse me, is it true that the method for training my own dataset involves preparing a folder to store images for training, write a new dataloader by imitating wireframe adding it to dataset_util.py, and then following the provided five steps for training?