microsoft/X-Decoder

Generic segmentation

trqminh opened this issue · 1 comments

Hi @MaureenZOU ,
I have quick question about where is in the code for the arrow from Text Encoder to Semantic Output when inference.

image

jwyang commented

THe arrow from text encoder to semantic output simply means the dot product between the text embeddings and semantic outputs decoded from our X-Decoder. Please refer to this function:

def compute_similarity(self, v_emb, name='default', fake=False):