How to train RCNN object detector for multiple traffic signs categories?

Hi!

i would like to ask, how to train object detector ( trainRCNNObjectDetector function) for multiple category?

rcnn = trainRCNNObjectDetector(T, cifar10Net, options,'NegativeOverlapRange', [0 0.3], 'PositiveOverlapRange',[0.5 1])

Where:

'Cifar10Net' is learned network for traffic signs (200 categories).

'T': Table contains input data (1 column = image link), 2nd column contains coordinates of sign in training image. With this table, the object detector is able to recognize signs in image, but is not able to label signs (because all coordinates are in one category). I need how to prepare input table data for multiple category.

I think the table should be like this:

Example of Table T

COL1 COL2 COL3 COL4

ImagePath StopSign MainRoad GiveWay ...

IMG1.jpg [10 20 50 50] none none

IMG2.jpg none [200 50 100 100] none

IMG3.jpg none [300 200 150 150] [800 400 150 150]

Which data i should to fill to other rows (instead 'none') if only StopSign is on IMG1.

IMG2.jpg contains only 'MainRoad' traffic sign. Also one image should contains more traffis signs..

IMG3.jpg contains two signs (Main Road and GiveWay).

Could someone help how to train the RCNN object detector for multiple (200) categories which one image should contains one or more traffic signs?

Thank you for your response