How Realistic Should Synthetic Images Be for Training Crowd Counting Models?

Emanuele Ledda, Lorenzo Putzu, Rita Delussu, Andrea Loddo, Giorgio Fumera