Can this algorithm be used for document image layout segmentation？ about sne-roadseg HOT 5 CLOSED

Sorry that I am not familiar with this task. Considering that our network is designed for taking two different modalities of data as input, I think you can try our approach if your task also has such demand.

from sne-roadseg.

It's dividing the document into columns,
The column effect is as follows: red on the left represents label 1, and yellow on the right represents label 2：

from sne-roadseg.

According to your description, I think it is more related to an object detection problem than a semantic segmentation problem.

from sne-roadseg.

from sne-roadseg.

I think you can try existing instance segmentation approaches such as Mask-RCNN, because it is more appropriate to divide different columns with different ids than different labels.

from sne-roadseg.