PulseAugur
实时 15:43:47
English(EN) Foveation-Guided Dynamic Token Selection for Robust and Efficient Vision Transformers

注视动态Transformer模仿人眼视觉,实现高效与鲁棒性

研究人员开发了一种受人眼视觉系统注视采样和眼动启发的注视动态Transformer(FDT)。该架构动态选择要处理的Token,在无需针对这些挑战进行专门训练的情况下,即可提高效率和对噪声及对抗性攻击的鲁棒性。在50%的注视预算下,FDT展现出比DeiT-S更高的准确率,同时显著减少了计算量,展示了准确率与效率之间有希望的权衡。 AI

影响 该模型的效率和鲁棒性有望在计算机视觉任务中带来更具适应性和韧性的AI系统。

排序理由 该集群包含一篇详细介绍新模型架构的学术论文。

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →

注视动态Transformer模仿人眼视觉,实现高效与鲁棒性

报道来源 [2]

  1. arXiv cs.LG TIER_1 English(EN) · Ibrahim Batuhan Akkaya, Kishaan Jeeveswaran, Bahram Zonooz, Elahe Arani ·

    Foveation-Guided Dynamic Token Selection for Robust and Efficient Vision Transformers

    arXiv:2607.09480v1 Announce Type: cross Abstract: The human visual system (HVS) employs foveated sampling and eye movements to achieve efficient perception, conserving both metabolic energy and computational resources. Drawing inspiration from this robustness and adaptability, we…

  2. arXiv cs.LG TIER_1 English(EN) · Elahe Arani ·

    面向鲁棒高效Vision Transformer的注视引导动态Token选择

    The human visual system (HVS) employs foveated sampling and eye movements to achieve efficient perception, conserving both metabolic energy and computational resources. Drawing inspiration from this robustness and adaptability, we introduce the Foveated Dynamic Transformer (FDT),…