This paper introduces EfficientNetV2, a new family of convolutional networks that have faster training speed and better parameter efficiency than previous models. To develop these models, we use a combination of training-aware neural architecture search and scaling, to jointly optimize training speed and parameter efficiency. The models were searched from the search space enriched with new ops such as Fused-MBConv. Our experiments show that EfficientNetV2 models train much faster than state-of-the-art models while being up to 6.8x smaller.Our training can be further sped up by progressively increasing the image size during training, but it often causes a drop in accuracy. To compensate for this accuracy drop, we propose an improved method of progressive learning, which adaptively adjusts regularization (e.g. data augmentation) along with image size.With progressive learning, our EfficientNetV2 significantly outperforms previous models on Im-ageNet and CIFAR/Cars/Flowers datasets. By pretraining on the same ImageNet21k, our Effi-cientNetV2 achieves 87.3% top-1 accuracy on ImageNet ILSVRC2012, outperforming the recent ViT by 2.0% accuracy while training 5x-11x faster using the same computing resources. Code is available at https://github.com/google/ automl/tree/master/efficientnetv2.

Convolutional Neural Networks (ConvNets) are commonly developed at a fixed resource budget, and then scaled up for better accuracy if more resources are available. In this paper, we systematically study model scaling and identify that carefully balancing network depth, width, and resolution can lead to better performance. Based on this observation, we propose a new scaling method that uniformly scales all dimensions of depth/width/resolution using a simple yet highly effective compound coefficient. We demonstrate the effectiveness of this method on scaling up MobileNets and ResNet.To go even further, we use neural architecture search to design a new baseline network and scale it up to obtain a family of models, called EfficientNets, which achieve much better accuracy and efficiency than previous ConvNets. In particular, our EfficientNet-B7 achieves state-of-the-art 84.3% top-1 accuracy on ImageNet, while being 8.4x smaller and 6.1x faster on inference than the best existing ConvNet. Our EfficientNets also transfer well and achieve state-of-the-art accuracy on CIFAR-100 (91.7%), Flowers (98.8%), and 3 other transfer learning datasets, with an order of magnitude fewer parameters. Source code is at https: //github.com/tensorflow/tpu/tree/ master/models/official/efficientnet.

由于存储器和计算资源有限，部署在移动设备上的卷积神经网络（CNNS）是困难的。我们的目标是通过利用特征图中的冗余来设计包括CPU和GPU的异构设备的高效神经网络，这很少在神经结构设计中进行了研究。对于类似CPU的设备，我们提出了一种新颖的CPU高效的Ghost（C-Ghost）模块，以生成从廉价操作的更多特征映射。基于一组内在的特征映射，我们使用廉价的成本应用一系列线性变换，以生成许多幽灵特征图，可以完全揭示内在特征的信息。所提出的C-Ghost模块可以作为即插即用组件，以升级现有的卷积神经网络。 C-Ghost瓶颈旨在堆叠C-Ghost模块，然后可以轻松建立轻量级的C-Ghostnet。我们进一步考虑GPU设备的有效网络。在建筑阶段的情况下，不涉及太多的GPU效率（例如，深度明智的卷积），我们建议利用阶段明智的特征冗余来制定GPU高效的幽灵（G-GHOST）阶段结构。舞台中的特征被分成两个部分，其中使用具有较少输出通道的原始块处理第一部分，用于生成内在特征，另一个通过利用阶段明智的冗余来生成廉价的操作。在基准测试上进行的实验证明了所提出的C-Ghost模块和G-Ghost阶段的有效性。 C-Ghostnet和G-Ghostnet分别可以分别实现CPU和GPU的准确性和延迟的最佳权衡。代码可在https://github.com/huawei-noah/cv-backbones获得。

Model efficiency has become increasingly important in computer vision. In this paper, we systematically study neural network architecture design choices for object detection and propose several key optimizations to improve efficiency. First, we propose a weighted bi-directional feature pyramid network (BiFPN), which allows easy and fast multiscale feature fusion; Second, we propose a compound scaling method that uniformly scales the resolution, depth, and width for all backbone, feature network, and box/class prediction networks at the same time. Based on these optimizations and better backbones, we have developed a new family of object detectors, called EfficientDet, which consistently achieve much better efficiency than prior art across a wide spectrum of resource constraints. In particular, with singlemodel and single-scale, our EfficientDet-D7 achieves stateof-the-art 55.1 AP on COCO test-dev with 77M parameters and 410B FLOPs 1 , being 4x -9x smaller and using 13x -42x fewer FLOPs than previous detectors. Code is available at https://github.com/google/automl/tree/ master/efficientdet.

Vision transformer (ViT) models exhibit substandard optimizability. In particular, they are sensitive to the choice of optimizer (AdamW vs. SGD), optimizer hyperparameters, and training schedule length. In comparison, modern convolutional neural networks are easier to optimize. Why is this the case? In this work, we conjecture that the issue lies with the patchify stem of ViT models, which is implemented by a stride-p p×p convolution (p = 16 by default) applied to the input image. This large-kernel plus large-stride convolution runs counter to typical design choices of convolutional layers in neural networks. To test whether this atypical design choice causes an issue, we analyze the optimization behavior of ViT models with their original patchify stem versus a simple counterpart where we replace the ViT stem by a small number of stacked stride-two 3×3 convolutions. While the vast majority of computation in the two ViT designs is identical, we find that this small change in early visual processing results in markedly different training behavior in terms of the sensitivity to optimization settings as well as the final model accuracy. Using a convolutional stem in ViT dramatically increases optimization stability and also improves peak performance (by ∼1-2% top-1 accuracy on ImageNet-1k), while maintaining flops and runtime. The improvement can be observed across the wide spectrum of model complexities (from 1G to 36G flops) and dataset scales (from ImageNet-1k to ImageNet-21k). These findings lead us to recommend using a standard, lightweight convolutional stem for ViT models in this regime as a more robust architectural choice compared to the original ViT model design.

In this paper, we present a modified Xception architecture, the NEXcepTion network. Our network has significantly better performance than the original Xception, achieving top-1 accuracy of 81.5% on the ImageNet validation dataset (an improvement of 2.5%) as well as a 28% higher throughput. Another variant of our model, NEXcepTion-TP, reaches 81.8% top-1 accuracy, similar to ConvNeXt (82.1%), while having a 27% higher throughput. Our model is the result of applying improved training procedures and new design decisions combined with an application of Neural Architecture Search (NAS) on a smaller dataset. These findings call for revisiting older architectures and reassessing their potential when combined with the latest enhancements.

最近，变压器和多层感知器（MLP）体系结构在各种视觉任务上取得了令人印象深刻的结果。但是，如何有效地结合这些操作员形成高性能混合视觉体系结构仍然是一个挑战。在这项工作中，我们通过提出一种新型的统一体系结构搜索方法来研究卷积，变压器和MLP的可学习组合。我们的方法包含两个关键设计，以实现高性能网络的搜索。首先，我们以统一的形式对截然不同的可搜索运算符进行建模，从而使操作员能够用相同的配置参数进行表征。这样，总体搜索空间规模大大减少，总搜索成本变得负担得起。其次，我们提出上下文感知的倒数采样模块（DSM），以减轻不同类型的操作员之间的差距。我们提出的DSM能够更好地适应不同类型的操作员的功能，这对于识别高性能混合体系结构很重要。最后，我们将可配置的运算符和DSM集成到统一的搜索空间中，并使用基于增强学习的搜索算法进行搜索，以充分探索操作员的最佳组合。为此，我们搜索一个基线网络并扩大规模，以获得一个名为UNINET的模型系列，该模型的准确性和效率比以前的Convnets和Transformers更好。特别是，我们的UNET-B5在ImageNet上获得了84.9％的TOP-1精度，比效应网络-B7和Botnet-T7分别少了44％和55％。通过在Imagenet-21K上进行预处理，我们的UNET-B6获得了87.4％，表现优于SWIN-L，拖鞋少51％，参数减少了41％。代码可在https://github.com/sense-x/uninet上找到。

Transformers have attracted increasing interests in computer vision, but they still fall behind state-of-the-art convolutional networks. In this work, we show that while Transformers tend to have larger model capacity, their generalization can be worse than convolutional networks due to the lack of the right inductive bias. To effectively combine the strengths from both architectures, we present CoAtNets (pronounced "coat" nets), a family of hybrid models built from two key insights:(1) depthwise Convolution and self-Attention can be naturally unified via simple relative attention; (2) vertically stacking convolution layers and attention layers in a principled way is surprisingly effective in improving generalization, capacity and efficiency. Experiments show that our CoAtNets achieve state-of-the-art performance under different resource constraints across various datasets: Without extra data, CoAtNet achieves 86.0% ImageNet top-1 accuracy; When pre-trained with 13M images from ImageNet-21K, our CoAtNet achieves 88.56% top-1 accuracy, matching ViT-huge pre-trained with 300M images from JFT-300M while using 23x less data; Notably, when we further scale up CoAtNet with JFT-3B, it achieves 90.88% top-1 accuracy on ImageNet, establishing a new state-of-the-art result.1 The initial projection stage can be seen as an aggressive down-sampling convolutional stem.

We introduce submodel co-training, a regularization method related to co-training, self-distillation and stochastic depth. Given a neural network to be trained, for each sample we implicitly instantiate two altered networks, ``submodels'', with stochastic depth: we activate only a subset of the layers. Each network serves as a soft teacher to the other, by providing a loss that complements the regular loss provided by the one-hot label. Our approach, dubbed cosub, uses a single set of weights, and does not involve a pre-trained external model or temporal averaging. Experimentally, we show that submodel co-training is effective to train backbones for recognition tasks such as image classification and semantic segmentation. Our approach is compatible with multiple architectures, including RegNet, ViT, PiT, XCiT, Swin and ConvNext. Our training strategy improves their results in comparable settings. For instance, a ViT-B pretrained with cosub on ImageNet-21k obtains 87.4% top-1 acc. @448 on ImageNet-val.

人们普遍认为，对于准确的语义细分，必须使用昂贵的操作（例如，非常卷积）结合使用昂贵的操作（例如非常卷积），从而导致缓慢的速度和大量的内存使用。在本文中，我们质疑这种信念，并证明既不需要高度的内部决议也不是必需的卷积。我们的直觉是，尽管分割是一个每像素的密集预测任务，但每个像素的语义通常都取决于附近的邻居和遥远的环境。因此，更强大的多尺度功能融合网络起着至关重要的作用。在此直觉之后，我们重新访问常规的多尺度特征空间（通常限制为P5），并将其扩展到更丰富的空间，最小的P9，其中最小的功能仅为输入大小的1/512，因此具有很大的功能接受场。为了处理如此丰富的功能空间，我们利用最近的BIFPN融合了多尺度功能。基于这些见解，我们开发了一个简化的分割模型，称为ESEG，该模型既没有内部分辨率高，也没有昂贵的严重卷积。也许令人惊讶的是，与多个数据集相比，我们的简单方法可以以比以前的艺术更快地实现更高的准确性。在实时设置中，ESEG-Lite-S在189 fps的CityScapes [12]上达到76.0％MIOU，表现优于更快的[9]（73.1％MIOU时为170 fps）。我们的ESEG-LITE-L以79 fps的速度运行，达到80.1％MIOU，在很大程度上缩小了实时和高性能分割模型之间的差距。

Designing accurate and efficient ConvNets for mobile devices is challenging because the design space is combinatorially large. Due to this, previous neural architecture search (NAS) methods are computationally expensive. ConvNet architecture optimality depends on factors such as input resolution and target devices. However, existing approaches are too resource demanding for case-by-case redesigns. Also, previous work focuses primarily on reducing FLOPs, but FLOP count does not always reflect actual latency. To address these, we propose a differentiable neural architecture search (DNAS) framework that uses gradient-based methods to optimize Con-vNet architectures, avoiding enumerating and training individual architectures separately as in previous methods. FBNets (Facebook-Berkeley-Nets), a family of models discovered by DNAS surpass state-of-the-art models both designed manually and generated automatically. FBNet-B achieves 74.1% top-1 accuracy on ImageNet with 295M FLOPs and 23.1 ms latency on a Samsung S8 phone, 2.4x smaller and 1.5x faster than MobileNetV2-1.3[17] with similar accuracy. Despite higher accuracy and lower latency than MnasNet[20], we estimate FBNet-B's search cost is 420x smaller than MnasNet's, at only 216 GPUhours. Searched for different resolutions and channel sizes, FBNets achieve 1.5% to 6.4% higher accuracy than Mo-bileNetV2. The smallest FBNet achieves 50.2% accuracy and 2.9 ms latency (345 frames per second) on a Samsung S8. Over a Samsung-optimized FBNet, the iPhone-Xoptimized model achieves a 1.4x speedup on an iPhone X. FBNet models are open-sourced at https://github. com/facebookresearch/mobile-vision. * Work done while interning at Facebook.… Figure 1. Differentiable neural architecture search (DNAS) for ConvNet design. DNAS explores a layer-wise space that each layer of a ConvNet can choose a different block. The search space is represented by a stochastic super net. The search process trains the stochastic super net using SGD to optimize the architecture distribution. Optimal architectures are sampled from the trained distribution. The latency of each operator is measured on target devices and used to compute the loss for the super net.

Designing convolutional neural networks (CNN) for mobile devices is challenging because mobile models need to be small and fast, yet still accurate. Although significant efforts have been dedicated to design and improve mobile CNNs on all dimensions, it is very difficult to manually balance these trade-offs when there are so many architectural possibilities to consider. In this paper, we propose an automated mobile neural architecture search (MNAS) approach, which explicitly incorporate model latency into the main objective so that the search can identify a model that achieves a good trade-off between accuracy and latency. Unlike previous work, where latency is considered via another, often inaccurate proxy (e.g., FLOPS), our approach directly measures real-world inference latency by executing the model on mobile phones. To further strike the right balance between flexibility and search space size, we propose a novel factorized hierarchical search space that encourages layer diversity throughout the network. Experimental results show that our approach consistently outperforms state-of-the-art mobile CNN models across multiple vision tasks. On the ImageNet classification task, our MnasNet achieves 75.2% top-1 accuracy with 78ms latency on a Pixel phone, which is 1.8× faster than MobileNetV2 [29] with 0.5% higher accuracy and 2.3× faster than NASNet [36] with 1.2% higher accuracy. Our MnasNet also achieves better mAP quality than MobileNets for COCO object detection. Code is at https://github.com/tensorflow/tpu/ tree/master/models/official/mnasnet.

在本文中，我们询问视觉变形金刚（VIT）是否可以作为改善机器学习模型对抗逃避攻击的对抗性鲁棒性的基础结构。尽管较早的作品集中在改善卷积神经网络上，但我们表明VIT也非常适合对抗训练以实现竞争性能。我们使用自定义的对抗训练配方实现了这一目标，该配方是在Imagenet数据集的一部分上使用严格的消融研究发现的。与卷积相比，VIT的规范培训配方建议强大的数据增强，部分是为了补偿注意力模块的视力归纳偏置。我们表明，该食谱在用于对抗训练时可实现次优性能。相比之下，我们发现省略所有重型数据增强，并添加一些额外的零件（$ \ varepsilon $ -Warmup和更大的重量衰减），从而大大提高了健壮的Vits的性能。我们表明，我们的配方在完整的Imagenet-1k上概括了不同类别的VIT体系结构和大规模模型。此外，调查了模型鲁棒性的原因，我们表明，在使用我们的食谱时，在训练过程中产生强烈的攻击更加容易，这会在测试时提高鲁棒性。最后，我们通过提出一种量化对抗性扰动的语义性质并强调其与模型的鲁棒性的相关性来进一步研究对抗训练的结果。总体而言，我们建议社区应避免将VIT的规范培训食谱转换为在对抗培训的背景下进行强大的培训和重新思考常见的培训选择。

神经网络的宽度很重要，因为增加了宽度，这必然会增加模型容量。但是，网络的性能不会随宽度而线性地提高，并且很快就会饱和。在这种情况下，我们认为，增加网络数量（合奏）的数量比纯粹增加宽度可以实现更好的准确性效率折衷。为了证明这一点，一个大型网络就其参数和正则化组件分为几个小网络。这些小型网络中的每一个都有原始参数的一小部分。然后，我们一起训练这些小型网络，使他们看到相同数据的各种观点，以增加它们的多样性。在此共同培训过程中，网络也可以相互学习。结果，小型网络可以比几乎没有或没有额外参数或拖船的大型网络获得更好的合奏性能，即实现更好的准确性效率折衷。通过并发运行，小型网络还可以比大型推理速度更快。以上所有内容都表明，网络的数量是模型缩放的新维度。我们通过广泛的实验在共同基准上使用8种不同的神经体系结构来验证我们的论点。该代码可在\ url {https://github.com/freeformrobotics/divide-and-co-training}中获得。

知识蒸馏（KD）最近成为压缩神经网络的一种流行方法。在最近的研究中，已经提出了同时找到学生模型的参数和体系结构的广义蒸馏方法。尽管如此，这种搜索方法仍需要大量的计算来搜索体系结构，并且缺点是仅考虑其搜索空间中的卷积块。本文介绍了一种新的算法，认为是信任区域意识架构搜索以有效提炼知识（贸易），该算法迅速找到了使用信任区域贝叶斯优化方法从几种最先进的架构中找到有效的学生体系结构。实验结果表明，我们提出的贸易算法始终优于KD培训下的常规NAS方法和预定义的架构。

With the success of Vision Transformers (ViTs) in computer vision tasks, recent arts try to optimize the performance and complexity of ViTs to enable efficient deployment on mobile devices. Multiple approaches are proposed to accelerate attention mechanism, improve inefficient designs, or incorporate mobile-friendly lightweight convolutions to form hybrid architectures. However, ViT and its variants still have higher latency or considerably more parameters than lightweight CNNs, even true for the years-old MobileNet. In practice, latency and size are both crucial for efficient deployment on resource-constraint hardware. In this work, we investigate a central question, can transformer models run as fast as MobileNet and maintain a similar size? We revisit the design choices of ViTs and propose an improved supernet with low latency and high parameter efficiency. We further introduce a fine-grained joint search strategy that can find efficient architectures by optimizing latency and number of parameters simultaneously. The proposed models, EfficientFormerV2, achieve about $4\%$ higher top-1 accuracy than MobileNetV2 and MobileNetV2$\times1.4$ on ImageNet-1K with similar latency and parameters. We demonstrate that properly designed and optimized vision transformers can achieve high performance with MobileNet-level size and speed.

我们提出了三种新型的修剪技术，以提高推理意识到的可区分神经结构搜索（DNAS）的成本和结果。首先，我们介绍了DNA的随机双路构建块，它可以通过内存和计算复杂性在内部隐藏尺寸上进行搜索。其次，我们在搜索过程中提出了一种在超级网的随机层中修剪块的算法。第三，我们描述了一种在搜索过程中修剪不必要的随机层的新技术。由搜索产生的优化模型称为Prunet，并在Imagenet Top-1图像分类精度的推理潜伏期中为NVIDIA V100建立了新的最先进的Pareto边界。将Prunet作为骨架还优于COCO对象检测任务的GPUNET和EFIDENENET，相对于平均平均精度（MAP）。

我们介绍了贴片采样时间表（PSS）的概念，该概念在训练过程中每批次使用的视觉变压器（VIT）贴片的数量变化。由于对于大多数视觉目标（例如，分类），所有补丁都不同样重要，因此我们认为，不太重要的补丁可以用于较少的训练迭代中，从而导致较短的训练时间，对性能的影响最小。此外，我们观察到，使用PSS的训练可以使VIT在推理过程中对更宽的贴片采样范围更强。这允许在推理过程中进行吞吐量和准确性之间的细粒度，动态的权衡。我们使用PSSS在VIT上评估Imagenet的VIT，均通过从头开始训练并使用重建损耗函数进行了预训练。对于预训练的模型，与使用所有斑块相比，我们的分类准确性降低了0.26％（从25小时到17小时）降低了0.26％。代码，模型检查点和日志可在https://github.com/bradmcdanel/pss上找到。

Transformers have been recently adapted for large scale image classification, achieving high scores shaking up the long supremacy of convolutional neural networks. However the optimization of image transformers has been little studied so far. In this work, we build and optimize deeper transformer networks for image classification. In particular, we investigate the interplay of architecture and optimization of such dedicated transformers. We make two transformers architecture changes that significantly improve the accuracy of deep transformers. This leads us to produce models whose performance does not saturate early with more depth, for instance we obtain 86.5% top-1 accuracy on Imagenet when training with no external data, we thus attain the current SOTA with less FLOPs and parameters. Moreover, our best model establishes the new state of the art on Imagenet with Reassessed labels and Imagenet-V2 / match frequency, in the setting with no additional training data. We share our code and models 1 .

我们展示了如何通过基于关注的全球地图扩充任何卷积网络，以实现非本地推理。我们通过基于关注的聚合层替换为单个变压器块的最终平均池，重量贴片如何参与分类决策。我们使用2个参数（宽度和深度）使用简单的补丁卷积网络，使用简单的补丁的卷积网络插入学习的聚合层。与金字塔设计相比，该架构系列在所有层上维护输入补丁分辨率。它在准确性和复杂性之间产生了令人惊讶的竞争权衡，特别是在记忆消耗方面，如我们在各种计算机视觉任务所示：对象分类，图像分割和检测的实验所示。