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 2401.00463cs.CV
收藏

分析自监督Vision Transformers的局部表示

Analyzing Local Representations of Self-supervised Vision Transformers

Yerevan State University (YSU)(埃里温国立大学);ETH Zurich(苏黎世联邦理工学院);YerevaNN(耶里瓦NN)

Ani Vanyan, Alvard Barseghyan, Hakob Tamazyan, Vahan Huroyan, Hrant Khachatrian, Martin Danelljan

AI总结:

本文比较分析了自监督ViTs的局部表示能力,发现对比学习方法DINO生成的patch表示比掩蔽图像建模更通用,且移除高方差特征能提升MAE的k-NN性能。

摘要:

本文对各种自监督Vision Transformers(ViTs)进行了比较分析,重点关注其局部表示能力。受大语言模型启发,我们研究了ViTs在极少或无需微调的情况下执行各种计算机视觉任务的能力。我们设计了评估框架,在few-shot语义分割、实例识别、目标检索和跟踪的背景下分析局部(即patch级别)表示的质量。我们发现,与masked image modeling相比,基于对比学习的方法(如DINO)能产生更通用的patch表示,无需参数调整即可直接应用于下游任务。使用后者方法(例如masked autoencoders)学习的嵌入具有高方差特征,会损害基于距离的算法(如k-NN),并且对于大多数下游任务不包含有用信息。此外,我们证明移除这些高方差特征可以增强MAE及其近期扩展Scale-MAE的k-NN性能。最后,我们发现在一种目标实例检索设置中,在数据量多两个数量级上预训练的模型DINOv2,表现不如其计算量较小的对应模型DINO。

英文摘要:

In this paper, we present a comparative analysis of various self-supervised Vision Transformers (ViTs), focusing on their local representative power. Inspired by large language models, we examine the abilities of ViTs to perform various computer vision tasks with little to no fine-tuning. We design evaluation framework to analyze the quality of local, i.e.\ patch-level, representations in the context of few-shot semantic segmentation, instance identification, object retrieval and tracking. We discover that contrastive learning based methods like DINO produce more universal patch representations that can be immediately applied for downstream tasks with no parameter tuning, compared to masked image modeling. The embeddings learned using the latter approach, e.g. in masked autoencoders, have high variance features that harm distance-based algorithms, such as k-NN, and do not contain useful information for most downstream tasks. Furthermore, we demonstrate that removing these high-variance features enhances k-NN for MAE, as well as for its recent extension Scale-MAE. Finally, we find an object instance retrieval setting where DINOv2, a model pretrained on two orders of magnitude more data, falls short of its less compute intensive counterpart DINO.

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。

↑