陈鼎彦
I saw the example dataset just use 8 pictures for train and 8 for test. If i use more than 8 pictures as the dataset, will the model be better?
It seems like we cannot fit ndarray smaller than shape(-1,3). e.g. if we want to fit in some array of shape(-1,2):  And it will call a function in gen_meta_features,...
Does the code implementation contains inference step, especially source prototype part and bimodel test to check those unknown samples? It seems the RAINCOAT algorithm only implements alignment part and correction...