nips 208 visual question answering 導讀

Overcoming Language Priors in Visual Question Answering with Adversarial Regularization Modern Visual Question Answering (VQA) models have been shown to rely heavily on superficial correlations betwee

>>阅读原文<<