search

HLR / Cross_Modality_Relevance

The source code of ACL 2020 paper: "Cross-Modality Relevance for Reasoning on Language and Vision"
26 stars 8 forks source link

Some questions about visual question answering #2

Open haoopan opened 3 years ago

haoopan commented 3 years ago

Hi, It is great work and I'm following your work. I have some questions and hope you give me some solution. First, how to code "Entity Relevance" and "Relational Relevance" in VQA task, just use the "logit4 = self.logit_fc4(relate_fc1)" and "logit2 = self.logit_fc2(entity_fc1)" in the NLVR code? Second, I want to konw the ablation experiments about the VQA, specifically, the result about "Entity Relevance" and "Relational Relevance" added in VQA task.

Thank you very much, I'm looking forward you reply!

Leavelk commented 2 years ago

Hello, have you reproduced the experimental results of VQA and can you give me some advice?