Hello, very nice work. Thanks for sharing your implementation! I find that the argument 'save_processed_im' in the _train_ScaleNet.py_ script is not used. Best regards, Zhenjun