1 paper
Ander Salaberria, Gorka Azkune, Oier Lopez de Lacalle +3
Existing work has observed that current text-to-image systems do not accurately reflect explicit spatial relations between objects such as 'left of' or 'below'. We hypothesize that…