> 文章列表 > 论文精读1:(网格特征)In Defense of Grid Features for Visual Question Answering(CVPR2020)

论文精读1:(网格特征)In Defense of Grid Features for Visual Question Answering(CVPR2020)

论文精读1:(网格特征)In Defense of Grid Features for Visual Question Answering(CVPR2020)

在这里插入图片描述

  1. 马萨诸塞州立大学阿默斯特分校
  2. Facebook 人工智能研究

目录

    • 1. Introduction
    • 2. Related Work
      • Visual features for vision and language tasks
      • Pre-training for VQA
      • Regions vs. grids.
    • 3. From Regions to Grids
      • 3.1. Bottom-Up Attention with Regions
        • Region selection
        • Region feature computation