Can we get the weights for this or text dataset along with the annotations? And some info on how to use the text detection?