In Defense of Grid Features for Visual Question Answering实现流程

网格特征预训练代码

这是该论文的特征预训练代码发布：

@InProceedings{jiang2020defense,
  title={In Defense of Grid Features for Visual Question Answering},
  author={Jiang, Huaizu and Misra, Ishan and Rohrbach, Marcus and Learned-Miller, Erik and Chen, Xinlei},
  booktitle={IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2020}
}

为了更持久的维护，我们使用Detectron2发布了代码，而不是基于mask-rcnn-benchmark的原始代码。当前的代码库应该能够复现论文中报告的结果，例如，对于与