Generate text or segment objects from an image
Compare SigLIP1 and SigLIP2 on zero shot classification