Max Pooling with Vision Transformers Reconciles Class and Shape in Weakly Supervised Semantic Segmentation

Simone Rossetti, Damiano Zappia, Marta Sanzari, Marco Schaerf, Fiora Pirri