search

microsoft / GLIP

Grounded Language-Image Pre-training
MIT License
2.18k stars 191 forks source link

Query about word-region alignment score #135

Closed tfc-0 closed 11 months ago

tfc-0 commented 1 year ago

Hi,

Thanks for the exciting work. I have a mall query: Does this line produce the region-word alignment scores S_ground mentioned in the paper? Line: https://github.com/microsoft/GLIP/blob/main/maskrcnn_benchmark/modeling/rpn/vldyhead.py#L843

inhaowu commented 11 months ago

我觉得应该是的，前者代表的是图像的特征，后者代表文本的特征。

liunian-harold-li commented 11 months ago

Thank you for your interest in our work. Sorry for the delay in response!

Yes, this line produces the region-word alignment scores.