Did anyone try to use ArcFace Loss on object other than faces?

I was wondering if the ArcFace loss is as effective in other domains as it is in the Face domain.

For example could one use ArcFace instead of Triplet Loss//Contrastive Learning/... losses to solve metric learning for other domains such as object recognition? and if it's doable, is it really stronger on that front too or is it specifically made for alignable data (like faces)?