Empty cells in trainYOLOv​2ObjectDet​ector trainingData

Can the trainingData cointain empty cells in the coulumns two and up, or do I have to use training data that has images containing all the possible classes in them, thus providing a complete table (without empty cells in any of the columns).

I am trying to train a network to detect different traffic signs. Obviously I cannot have pictures containing several dozen signs on each.

How would I go about defining the trainingData?