I have a large dataset of bottles. I want to train a model with this dataset. But before feeding the input images to the model I want to crop out the bottle from the background. Is there a way to do it using image processing. i.e without using machine learning?