arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2403.07883cs.CVcs.AI
收藏

基于文本相关图像块选择的高效视觉与语言预训练

Efficient Vision-and-Language Pre-training with Text-Relevant Image Patch Selection

National Engineering Research Center for Software Engineering, Peking University(北京大学国家软件工程研究中心);DAMO Academy, Alibaba Group(阿里巴巴集团达摩院)

Wei Ye, Chaoya Jiang, Haiyang Xu, Chenhao Ye, Chenliang Li, Ming Yan, Shikun Zhang, Songhang Huang, Fei Huang

AI总结:

本文提出TRIPS方法,通过文本引导的图像块选择逐步缩减视觉序列,在不增加参数的情况下加速ViT-based视觉语言预训练,实现40%提速并保持或提升下游任务性能。

摘要:

视觉Transformer(ViTs)在大规模视觉与语言预训练(VLP)模型中越来越受欢迎。尽管先前的VLP研究已证明ViTs的有效性,但这些工作仍面临由冗长视觉序列导致的计算效率低下的问题。为应对这一挑战,我们提出了一种名为TRIPS的高效VLP方法,即文本相关图像块选择(Text-Relevant Image Patch Selection)。TRIPS通过在视觉骨干网络中引入文本引导的块选择层,逐步缩减视觉序列,从而加速训练和推理过程。该块选择层动态计算文本相关的视觉注意力,使其能够在文本引导下识别注意力图像标记,并以端到端方式融合非注意力标记。重要的是,TRIPS不增加任何额外参数,且可推广至大多数基于ViT的VLP模型。我们将TRIPS集成到三个代表性VLP模型中,涵盖单流、双流和生成范式,并在五个广泛使用的多模态基准数据集上进行了大量实验。实验结果表明,TRIPS实现了40%的加速,同时在下游任务上保持竞争性或更优的性能。

英文摘要:

Vision Transformers (ViTs) have become increasingly popular in large-scale Vision and Language Pre-training (VLP) models. Although previous VLP research has demonstrated the efficacy of ViTs, these efforts still struggle with computational inefficiencies caused by lengthy visual sequences. To address this challenge, we introduce an efficient VLP approach called TRIPS, which stands for Text-Relevant Image Patch Selection. TRIPS progressively reduces the visual sequence using a text-guided patch-selection layer in the visual backbone, thereby accelerating both training and inference processes. This patch-selection layer dynamically computes text-dependent visual attention, enabling it to identify attentive image tokens with text guidance and fuse inattentive ones in an end-to-end fashion. Importantly, TRIPS does not add any extra parameters and generalizes to most ViT-based VLP models. We incorporate TRIPS into three representative VLP models covering single-stream, dual-stream, and generative paradigms, and conduct extensive experiments on five widely-used multi-modal benchmark datasets. Our experimental results reveal that TRIPS delivers a 40% speedup, while maintaining competitive or superior performance on downstream tasks.

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。

↑