Aligning machine and human visual representations across abstraction levels.

Because of these factors, the relationships between image representations change during alignment according to their semantic relationship; representations of images from the same basic category tend to move closer together, those of images from the same superordinate category tend to move somewhat closer, and those from different superordinate categories tend to move apart (all effects are highly significant; t > 3.93, P < 0.001).

also in 132,142 other papers