arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1807.09664cs.AIcs.CV
收藏

行动前先注意:利用人类视觉注意力进行持续学习

Attend Before you Act: Leveraging human visual attention for continual learning

Mila - McGill University(米拉-麦吉尔大学);Google Deepmind(谷歌DeepMind)

Khimya Khetarpal, Doina Precup

摘要:

当人类执行任务时,例如玩游戏,他们选择性地关注视觉输入的某些部分,收集相关信息并将其顺序组合以从感官数据构建表示。在这项工作中,我们探索利用人类在图像中的位置作为决策的显着指示。我们在DeepMind Lab的3D导航迷宫环境中构建了UNREAL架构。我们训练代理与原始图像和foveated图像,这是通过重叠的原始图像与显着图使用实时光谱残差技术产生的。我们调查的有效性,这种方法在迁移学习的环境中的噪声的背景下测量性能。

英文摘要:

When humans perform a task, such as playing a game, they selectively pay attention to certain parts of the visual input, gathering relevant information and sequentially combining it to build a representation from the sensory data. In this work, we explore leveraging where humans look in an image as an implicit indication of what is salient for decision making. We build on top of the UNREAL architecture in DeepMind Lab's 3D navigation maze environment. We train the agent both with original images and foveated images, which were generated by overlaying the original images with saliency maps generated using a real-time spectral residual technique. We investigate the effectiveness of this approach in transfer learning by measuring performance in the context of noise in the environment.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑