Zero-shot result of GLIPv2 on phrasecut

[594zyc]

Thanks for this great work! It is amazing that GLIPv2 can perform various tasks in a zero-shot manner as shown in section 4.2 of the paper. However, I see the PhraseCut results are not reported in Table 2, is there a particular reason for that? I wonder what the zero-shot performance would be like for GLIPv2 on PhraseCut.