I am following the "Towards Data Science" guide https://towardsdatascience.com/custom-object-detection-using-tensorflow-from-scratch-e61da2e10087
I ran into an issue on step 6 where I need to create a tf record. The create_tf_record.py script requires a trimaps folder full of png's but that is confusing. Shouldn't I not need a trimap's folder if I have already given it the xml annotations created by labelme? Isn't the required information to train on already inscribed into the xml files for each image?
