Convolutional neural networks (CNNs) have so far been the de-facto model for visual data. Recent work has shown that (Vision) Transformer models (ViT) can achieve comparable or even superior performance on image classification tasks. This raises a central question: how are Vision Transformers solving these tasks? Are they acting like convolutional networks, or learning entirely different visual representations? Analyzing the internal representation structure of ViTs and CNNs on image classification benchmarks, we find striking differences between the two architectures, such as ViT having more uniform representations across all layers. We explore how these differences arise, finding crucial roles played by self-attention, which enables early aggregation of global information, and ViT residual connections, which strongly propagate features from lower to higher layers. We study the ramifications for spatial localization, demonstrating ViTs successfully preserve input spatial information, with noticeable effects from different classification methods. Finally, we study the effect of (pretraining) dataset scale on intermediate features and transfer learning, and conclude with a discussion on connections to new architectures such as the MLP-Mixer. This breakthrough highlights a fundamental question: how are Vision Transformers solving these image based tasks? Do they act like convolutions, learning the same inductive biases from scratch? Or are they developing novel task representations? What is the role of scale in learning these representations? And are there ramifications for downstream tasks? In this paper, we study these questions, uncovering key representational differences between ViTs and CNNs, the ways in which these difference arise, and effects on classification and transfer learning. Specifically, our contributions are:35th Conference on Neural Information Processing Systems (NeurIPS 2021).

Vision Transformers (ViTs) have gained significant popularity in recent years and have proliferated into many applications. However, it is not well explored how varied their behavior is under different learning paradigms. We compare ViTs trained through different methods of supervision, and show that they learn a diverse range of behaviors in terms of their attention, representations, and downstream performance. We also discover ViT behaviors that are consistent across supervision, including the emergence of Offset Local Attention Heads. These are self-attention heads that attend to a token adjacent to the current token with a fixed directional offset, a phenomenon that to the best of our knowledge has not been highlighted in any prior work. Our analysis shows that ViTs are highly flexible and learn to process local and global information in different orders depending on their training method. We find that contrastive self-supervised methods learn features that are competitive with explicitly supervised features, and they can even be superior for part-level tasks. We also find that the representations of reconstruction-based models show non-trivial similarity to contrastive self-supervised models. Finally, we show how the "best" layer for a given task varies by both supervision method and task, further demonstrating the differing order of information processing in ViTs.

While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train. 1

Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e. "mixing" the per-location features), and one with MLPs applied across patches (i.e. "mixing" spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers. 1

最近的工作表明，视觉变压器（VTS）的注意力图在接受自学训练时，可以包含一种语义分割结构，在监督训练时不会自发出现。在本文中，我们明确鼓励这种空间聚类的出现作为一种培训正规化的形式，这种方式包括在标准监督学习中进行自我监督的借口任务。更详细地，我们根据信息熵的空间公式提出了一种VT正则化方法。通过最大程度地减少提议的空间熵，我们明确要求VT生成空间有序的注意图，这是在训练过程中包括基于对象的先验。使用广泛的实验，我们表明，在不同的培训方案，数据集，下游任务和VT体系结构中，提出的正则化方法是有益的。该代码将在接受后可用。

视觉变压器（VIT）在各种机器视觉问题上表现出令人印象深刻的性能。这些模型基于多头自我关注机制，可以灵活地参加一系列图像修补程序以编码上下文提示。一个重要问题是在给定贴片上参加图像范围内的上下文的这种灵活性是如何促进在自然图像中处理滋扰，例如，严重的闭塞，域移位，空间置换，对抗和天然扰动。我们通过广泛的一组实验来系统地研究了这个问题，包括三个vit家族和具有高性能卷积神经网络（CNN）的比较。我们展示和分析了vit的以下迷恋性质：（a）变压器对严重闭塞，扰动和域移位高度稳健，例如，即使在随机堵塞80％的图像之后，也可以在想象中保持高达60％的前1个精度。内容。 （b）与局部纹理的偏置有抗闭锁的强大性能，与CNN相比，VITS对纹理的偏置显着偏差。当受到适当训练以编码基于形状的特征时，VITS展示与人类视觉系统相当的形状识别能力，以前在文献中无与伦比。 （c）使用VIT来编码形状表示导致准确的语义分割而没有像素级监控的有趣后果。 （d）可以组合从单VIT模型的现成功能，以创建一个功能集合，导致传统和几枪学习范例的一系列分类数据集中的高精度率。我们显示VIT的有效特征是由于自我关注机制可以实现灵活和动态的接受领域。

转移学习是一种标准技术，可以将知识从一个领域转移到另一个领域。对于医学成像中的应用，尽管域之间的任务和图像特征差异，但从Imagenet转移已成为事实上的方法。但是，尚不清楚哪些因素决定了哪些因素以及在何种程度上转移学习到医疗领域是有用的。最近，人们对源域重复使用的特征的长期假设最近受到质疑。通过在几个医学图像基准数据集上进行的一系列实验，我们探讨了传输学习，数据大小，模型的容量和电感偏置以及源域和目标域之间的距离之间的关系。我们的发现表明，在大多数情况下，转移学习是有益的，我们表征了重要的角色重复使用在其成功方面。

预训练在高级计算机视觉中标志着众多艺术状态，但曾经有很少的尝试调查图像处理系统中的预训练方式。在本文中，我们对图像预培训进行了深入研究。在实用价值考虑到实际价值的实际基础进行本研究，我们首先提出了一种通用，经济高效的变压器的图像处理框架。它在一系列低级任务中产生了高度竞争的性能，但在约束参数和计算复杂性下。然后，基于此框架，我们设计了一整套原则性的评估工具，认真对待和全面地诊断不同任务的图像预训练，并揭示其对内部网络表示的影响。我们发现预训练在低级任务中发挥着惊人的不同角色。例如，预训练将更多本地信息引入超级分辨率（SR）的更高层数，产生显着的性能增益，而预培训几乎不会影响去噪的内部特征表示，导致稍微收益。此外，我们探索了不同的预训练方法，揭示了多任务预训练更有效和数据效率。所有代码和模型将在https://github.com/fenglinglwb/edt发布。

将深度学习模型部署在具有有限计算资源的时间关键性应用程序中，例如在边缘计算系统和IoT网络中，是一项具有挑战性的任务，通常依赖于动态推理方法（例如早期退出）。在本文中，我们介绍了一种基于视觉变压器体系结构的新型架构，用于早期退出，以及一种微调策略，该策略与传统方法相比，在引入较少的开销的同时，显着提高了早期出口分支的准确性。通过有关图像和音频分类以及视听人群计数的广泛实验，我们表明我们的方法在分类和回归问题以及单模式设置中都适用于分类和回归问题。此外，我们引入了一种新颖的方法，用于在视听数据分析的早期出口中整合音频和视觉方式，这可能导致更细粒度的动态推断。

Vision transformers (ViTs) are quickly becoming the de-facto architecture for computer vision, yet we understand very little about why they work and what they learn. While existing studies visually analyze the mechanisms of convolutional neural networks, an analogous exploration of ViTs remains challenging. In this paper, we first address the obstacles to performing visualizations on ViTs. Assisted by these solutions, we observe that neurons in ViTs trained with language model supervision (e.g., CLIP) are activated by semantic concepts rather than visual features. We also explore the underlying differences between ViTs and CNNs, and we find that transformers detect image background features, just like their convolutional counterparts, but their predictions depend far less on high-frequency information. On the other hand, both architecture types behave similarly in the way features progress from abstract patterns in early layers to concrete objects in late layers. In addition, we show that ViTs maintain spatial information in all layers except the final layer. In contrast to previous works, we show that the last layer most likely discards the spatial information and behaves as a learned global pooling operation. Finally, we conduct large-scale visualizations on a wide range of ViT variants, including DeiT, CoaT, ConViT, PiT, Swin, and Twin, to validate the effectiveness of our method.

最近的几项研究表明，基于关注的网络，如视觉变压器（VIV），可以在几个计算机视觉任务上倾斜卷积神经网络（CNNS）而不使用卷积层。这自然导致以下问题：可以自我关注的Vit表达任何卷积操作吗？在这项工作中，我们证明了一种具有图像贴片的单个VIT层，因为输入可以建设性地执行任何卷积操作，其中多主题注意机制和相对位置编码起到基本角色。我们进一步提供了视觉变压器的头部数量的下限，以表达CNN。对应于我们的分析，实验结果表明，我们证据的建设可以帮助将卷积偏差注入变压器，并显着提高vit的低数据制度的性能。

随着变压器作为语言处理的标准及其在计算机视觉方面的进步，参数大小和培训数据的数量相应地增长。许多人开始相信，因此，变形金刚不适合少量数据。这种趋势引起了人们的关注，例如：某些科学领域中数据的可用性有限，并且排除了该领域研究资源有限的人。在本文中，我们旨在通过引入紧凑型变压器来提出一种小规模学习的方法。我们首次表明，具有正确的尺寸，卷积令牌化，变压器可以避免在小数据集上过度拟合和优于最先进的CNN。我们的模型在模型大小方面具有灵活性，并且在获得竞争成果的同时，参数可能仅为0.28亿。当在CIFAR-10上训练Cifar-10，只有370万参数训练时，我们的最佳模型可以达到98％的准确性，这是与以前的基于变形金刚的模型相比，数据效率的显着提高，比其他变压器小于10倍，并且是15％的大小。在实现类似性能的同时，重新NET50。 CCT还表现优于许多基于CNN的现代方法，甚至超过一些基于NAS的方法。此外，我们在Flowers-102上获得了新的SOTA，具有99.76％的TOP-1准确性，并改善了Imagenet上现有基线（82.71％精度，具有29％的VIT参数）以及NLP任务。我们针对变压器的简单而紧凑的设计使它们更可行，可以为那些计算资源和/或处理小型数据集的人学习，同时扩展了在数据高效变压器中的现有研究工作。我们的代码和预培训模型可在https://github.com/shi-labs/compact-transformers上公开获得。

由于具有强大的代表性，变形金刚在包括自然语言处理（NLP），计算机视觉和语音识别在内的广泛应用中越来越受欢迎。但是，利用这种代表性的能力有效地需要大量的数据，强大的正则化或两者兼而有之以减轻过度拟合。最近，基于掩盖的自动编码器的自我监督预处理策略已解锁了变压器的功能，这些策略依赖于直接或从未掩盖的内容对比的掩蔽输入进行重建。这种预训练的策略已在NLP中的BERT模型，Speak2VEC模型中使用，最近在Vision中的MAE模型中，该模型迫使该模型使用自动编码相关的目标来了解输入不同部分中的内容之间的关系。在本文中，我们提出了一种小说但令人惊讶的简单替代内容，以预测内容的位置，而无需为其提供位置信息。这样做需要变压器仅凭内容就可以理解输入不同部分之间的位置关系。这相当于有效的实现，其中借口任务是每个输入令牌所有可能位置之间的分类问题。我们在视觉和语音基准上进行了实验，我们的方法对强有力的监督训练基准进行了改进，并且与现代的无监督/自我监督预审方法相媲美。我们的方法还可以使经过训练的变压器在没有位置嵌入的情况下胜过训练有完整位置信息的训练的变压器。

We introduce dense vision transformers, an architecture that leverages vision transformers in place of convolutional networks as a backbone for dense prediction tasks. We assemble tokens from various stages of the vision transformer into image-like representations at various resolutions and progressively combine them into full-resolution predictions using a convolutional decoder. The transformer backbone processes representations at a constant and relatively high resolution and has a global receptive field at every stage. These properties allow the dense vision transformer to provide finer-grained and more globally coherent predictions when compared to fully-convolutional networks. Our experiments show that this architecture yields substantial improvements on dense prediction tasks, especially when a large amount of training data is available. For monocular depth estimation, we observe an improvement of up to 28% in relative performance when compared to a state-of-theart fully-convolutional network. When applied to semantic segmentation, dense vision transformers set a new state of the art on ADE20K with 49.02% mIoU. We further show that the architecture can be fine-tuned on smaller datasets such as NYUv2, KITTI, and Pascal Context where it also sets the new state of the art. Our models are available at https://github.com/intel-isl/DPT.

There still remains an extreme performance gap between Vision Transformers (ViTs) and Convolutional Neural Networks (CNNs) when training from scratch on small datasets, which is concluded to the lack of inductive bias. In this paper, we further consider this problem and point out two weaknesses of ViTs in inductive biases, that is, the spatial relevance and diverse channel representation. First, on spatial aspect, objects are locally compact and relevant, thus fine-grained feature needs to be extracted from a token and its neighbors. While the lack of data hinders ViTs to attend the spatial relevance. Second, on channel aspect, representation exhibits diversity on different channels. But the scarce data can not enable ViTs to learn strong enough representation for accurate recognition. To this end, we propose Dynamic Hybrid Vision Transformer (DHVT) as the solution to enhance the two inductive biases. On spatial aspect, we adopt a hybrid structure, in which convolution is integrated into patch embedding and multi-layer perceptron module, forcing the model to capture the token features as well as their neighboring features. On channel aspect, we introduce a dynamic feature aggregation module in MLP and a brand new "head token" design in multi-head self-attention module to help re-calibrate channel representation and make different channel group representation interacts with each other. The fusion of weak channel representation forms a strong enough representation for classification. With this design, we successfully eliminate the performance gap between CNNs and ViTs, and our DHVT achieves a series of state-of-the-art performance with a lightweight model, 85.68% on CIFAR-100 with 22.8M parameters, 82.3% on ImageNet-1K with 24.0M parameters. Code is available at https://github.com/ArieSeirack/DHVT.

视觉变压器（VIT）用作强大的视觉模型。与卷积神经网络不同，在前几年主导视觉研究，视觉变压器享有捕获数据中的远程依赖性的能力。尽管如此，任何变压器架构的组成部分，自我关注机制都存在高延迟和低效的内存利用，使其不太适合高分辨率输入图像。为了缓解这些缺点，分层视觉模型在非交错的窗口上局部使用自我关注。这种放松会降低输入尺寸的复杂性;但是，它限制了横窗相互作用，损害了模型性能。在本文中，我们提出了一种新的班次不变的本地注意层，称为查询和参加（QNA），其以重叠的方式聚集在本地输入，非常类似于卷积。 QNA背后的关键想法是介绍学习的查询，这允许快速高效地实现。我们通过将其纳入分层视觉变压器模型来验证我们的层的有效性。我们展示了速度和内存复杂性的改进，同时实现了与最先进的模型的可比准确性。最后，我们的图层尺寸尤其良好，窗口大小，需要高于X10的内存，而不是比现有方法更快。

多层erceptron（MLP），作为出现的第一个神经网络结构，是一个大的击中。但是由硬件计算能力和数据集的大小限制，它一旦沉没了数十年。在此期间，我们目睹了从手动特征提取到带有局部接收领域的CNN的范式转变，以及基于自我关注机制的全球接收领域的变换。今年（2021年），随着MLP混合器的推出，MLP已重新进入敏捷，并吸引了计算机视觉界的广泛研究。与传统的MLP进行比较，它变得更深，但改变了完全扁平化以补丁平整的输入。鉴于其高性能和较少的需求对视觉特定的感应偏见，但社区无法帮助奇迹，将MLP，最简单的结构与全球接受领域，但没有关注，成为一个新的电脑视觉范式吗？为了回答这个问题，本调查旨在全面概述视觉深层MLP模型的最新发展。具体而言，我们从微妙的子模块设计到全局网络结构，我们审查了这些视觉深度MLP。我们比较了不同网络设计的接收领域，计算复杂性和其他特性，以便清楚地了解MLP的开发路径。调查表明，MLPS的分辨率灵敏度和计算密度仍未得到解决，纯MLP逐渐发展朝向CNN样。我们建议，目前的数据量和计算能力尚未准备好接受纯的MLP，并且人工视觉指导仍然很重要。最后，我们提供了开放的研究方向和可能的未来作品的分析。我们希望这项努力能够点燃社区的进一步兴趣，并鼓励目前为神经网络进行更好的视觉量身定制设计。

探讨了语言建模流行的变形金刚，用于近期解决视觉任务，例如，用于图像分类的视觉变压器（VIT）。 VIT模型将每个图像分成具有固定长度的令牌序列，然后应用多个变压器层以模拟它们的全局关系以进行分类。然而，当从像想象中的中型数据集上从头开始训练时，VIT对CNNS达到较差的性能。我们发现它是因为：1）输入图像的简单标记未能模拟相邻像素之间的重要局部结构，例如边缘和线路，导致训练采样效率低。 2）冗余注意骨干骨干设计对固定计算预算和有限的训练样本有限的具有限制性。为了克服这些限制，我们提出了一种新的令牌到令牌视觉变压器（T2T-VIT），它包含1）层 - 明智的代币（T2T）转换，通过递归聚合相邻来逐步地结构于令牌到令牌。代币进入一个令牌（令牌到令牌），这样可以建模由周围令牌所代表的本地结构，并且可以减少令牌长度; 2）一种高效的骨干，具有深度狭窄的结构，用于在实证研究后CNN建筑设计的激励变压器结构。值得注意的是，T2T-VIT将Vanilla Vit的参数计数和Mac减少了一半，同时从想象中从头开始训练时，改善了超过3.0 \％。它还优于Endnets并通过直接培训Imagenet训练来实现与MobileNets相当的性能。例如，T2T-VTO与Reset50（21.5M参数）的可比大小（21.5M参数）可以在图像分辨率384 $ \ Times 384上实现83.3 \％TOP1精度。 （代码：https://github.com/yitu-opensource/t2t-vit）

We present in this paper a new architecture, named Convolutional vision Transformer (CvT), that improves Vision Transformer (ViT) in performance and efficiency by introducing convolutions into ViT to yield the best of both designs. This is accomplished through two primary modifications: a hierarchy of Transformers containing a new convolutional token embedding, and a convolutional Transformer block leveraging a convolutional projection. These changes introduce desirable properties of convolutional neural networks (CNNs) to the ViT architecture (i.e. shift, scale, and distortion invariance) while maintaining the merits of Transformers (i.e. dynamic attention, global context, and better generalization). We validate CvT by conducting extensive experiments, showing that this approach achieves state-of-the-art performance over other Vision Transformers and ResNets on ImageNet-1k, with fewer parameters and lower FLOPs. In addition, performance gains are maintained when pretrained on larger datasets (e.g. ImageNet-22k) and fine-tuned to downstream tasks. Pretrained on ImageNet-22k, our CvT-W24 obtains a top-1 accuracy of 87.7% on the ImageNet-1k val set. Finally, our results show that the positional encoding, a crucial component in existing Vision Transformers, can be safely removed in our model, simplifying the design for higher resolution vision tasks. Code will be released at https: //github.com/leoxiaobin/CvT.

We design a family of image classification architectures that optimize the trade-off between accuracy and efficiency in a high-speed regime. Our work exploits recent findings in attention-based architectures, which are competitive on highly parallel processing hardware. We revisit principles from the extensive literature on convolutional neural networks to apply them to transformers, in particular activation maps with decreasing resolutions. We also introduce the attention bias, a new way to integrate positional information in vision transformers.As a result, we propose LeVIT: a hybrid neural network for fast inference image classification. We consider different measures of efficiency on different hardware platforms, so as to best reflect a wide range of application scenarios. Our extensive experiments empirically validate our technical choices and show they are suitable to most architectures. Overall, LeViT significantly outperforms existing convnets and vision transformers with respect to the speed/accuracy tradeoff. For example, at 80% ImageNet top-1 accuracy, LeViT is 5 times faster than EfficientNet on CPU. We release the code at https: //github.com/facebookresearch/LeViT.