BriefGPT.xyz
Aug, 2022
鲁棒视觉问答的生成偏差
Generative Bias for Visual Question Answering
HTML
PDF
Jae Won Cho, Dong-jin Kim, Hyeonggon Ryu, In So Kweon
TL;DR
采用基于生成网络、对抗性目标函数和知识蒸馏相结合的方法直接训练VQA模型的偏见模型，有效减轻VQA模型中的数据集偏差问题。
Abstract
The task of
visual question answering
(VQA) is known to be plagued by the issue of
vqa models
exploiting biases within the dataset to make its final prediction. Many previous ensemble based
→