My dataset doesnot contains annotations for other hierarchies likes Word and Paragraph.
Also, my ground is polygons coordinates around the text-line and pixel-wise ground truth is not available?
Can you direct me towards the steps to follow to train such a custom dataset on Hi-SAM?