In order to handle modern convolutional neural networks (CNNs) efficiently, a hardware architecture of CNN inference accelerator is proposed to handle depthwise convolutions and regular convolutions, which are both essential building blocks for embedded-computer-vision algorithms. Different from related works, the proposed architecture can support filter kernels with different sizes with high flexibility since it does not require extra costs for intra-kernel parallelism, and it can generate convolution results faster than the architecture of the related works. The experimental results show the importance of supporting depthwise convolutions and dilated convolutions with the proposed hardware architecture. In addition to depthwise convolutions with large-kernels, a new structure called DDC layer, which includes the combination of depthwise convolutions and dilated convolutions, is also analyzed in this paper. For face detection, the computational costs decrease by 30%, and the model size decreases by 20% when the DDC layers are applied to the network. For image classification, the accuracy is increased by 1% by simply replacing $3 \times 3$ filters with $5 \times 5$ filters in depthwise convolutions.
Chonnam National University（全南国立大学）；Xidian University（西安电子科技大学）
Although deep convolutional neural networks (CNNs) have obtained outstanding performance in image superresolution (SR), their computational cost increases geometrically as CNN models get deeper and wider. Meanwhile, the features of intermediate layers are treated equally across the channel, thus hindering the representational capability of CNNs. In this paper, we propose an attention-aware linear depthwise network to address the problems for single image SR, named ALDNet. Specifically, linear depthwise convolution allows CNN-based SR models to preserve useful information for reconstructing a s…
Indian Institute of Technology Gandhinagar（印度理工学院甘地纳格尔分校）
In this paper, we propose a new convolutional layer called Depthwise-STFT Separable layer that can serve as an alternative to the standard depthwise separable convolutional layer. The construction of the proposed layer is inspired by the fact that the Fourier coefficients can accurately represent important features such as edges in an image. It utilizes the Fourier coefficients computed (channelwise) in the 2D local neighborhood (e.g., 3x3) of each position of the input map to obtain the feature maps. The Fourier coefficients are computed using 2D Short Term Fourier Transform (STFT) at multipl…
Alibaba Group（阿里巴巴集团）；Shandong University（山东大学）；The Hebrew University of Jerusalem（耶路撒冷希伯来大学）；Tel Aviv University（特拉维夫大学）；Peking University（北京大学）
Convolutional layers are the core building blocks of Convolutional Neural Networks (CNNs). In this paper, we propose to augment a convolutional layer with an additional depthwise convolution, where each input channel is convolved with a different 2D kernel. The composition of the two convolutions constitutes an over-parameterization, since it adds learnable parameters, while the resulting linear operation can be expressed by a single convolution layer. We refer to this depthwise over-parameterized convolutional layer as DO-Conv. We show with extensive experiments that the mere replacement of c…
Beijing Normal University（北京师范大学）；University of Waterloo（滑铁卢大学）；EPFL（洛桑联邦理工学院）
Standard 3D convolution operations require much larger amounts of memory and computation cost than 2D convolution operations. The fact has hindered the development of deep neural nets in many 3D vision tasks. In this paper, we investigate the possibility of applying depthwise separable convolutions in 3D scenario and introduce the use of 3D depthwise convolution. A 3D depthwise convolution splits a single standard 3D convolution into two separate steps, which would drastically reduce the number of parameters in 3D convolutions with more than one order of magnitude. We experiment with 3D depthw…
Depthwise convolutions provide significant performance benefits owing to the reduction in both parameters and mult-adds. However, training depthwise convolution layers with GPUs is slow in current deep learning frameworks because their implementations cannot fully utilize the GPU capacity. To address this problem, in this paper we present an efficient method (called diagonalwise refactorization) for accelerating the training of depthwise convolution layers. Our key idea is to rearrange the weight vectors of a depthwise convolution into a large diagonal weight matrix so as to convert the depthw…
We present an interpretation of Inception modules in convolutional neural networks as being an intermediate step in-between regular convolution and the depthwise separable convolution operation (a depthwise convolution followed by a pointwise convolution). In this light, a depthwise separable convolution can be understood as an Inception module with a maximally large number of towers. This observation leads us to propose a novel deep convolutional neural network architecture inspired by Inception, where Inception modules have been replaced with depthwise separable convolutions. We show that th…
Seoul National University（首尔国立大学）；CCB Fintech（CCB金融科技）；Columbia University（哥伦比亚大学）
How can we efficiently compress Convolutional Neural Network (CNN) while retaining their accuracy on classification tasks? Depthwise Separable Convolution (DSConv), which replaces a standard convolution with a depthwise convolution and a pointwise convolution, has been used for building lightweight architectures. However, previous works based on depthwise separable convolution are limited when compressing a trained CNN model since 1) they are mostly heuristic approaches without a precise understanding of their relations to standard convolution, and 2) their accuracies do not match that of the …
Tsinghua University（清华大学）；University College London（伦敦大学学院）
Depthwise convolution has gradually become an indispensable operation for modern efficient neural networks and larger kernel sizes ($\ge5$) have been applied to it recently. In this paper, we propose a novel extremely separated convolutional block (XSepConv), which fuses spatially separable convolutions into depthwise convolution to further reduce both the computational cost and parameter size of large kernels. Furthermore, an extra $2\times2$ depthwise convolution coupled with improved symmetric padding strategy is employed to compensate for the side effect brought by spatially separable conv…
Depthwise separable convolution has shown great efficiency in network design, but requires time-consuming training procedure with full training-set available. This paper first analyzes the mathematical relationship between regular convolutions and depthwise separable convolutions, and proves that the former one could be approximated with the latter one in closed form. We show depthwise separable convolutions are principal components of regular convolutions. And then we propose network decoupling (ND), a training-free method to accelerate convolutional neural networks (CNNs) by transferring pre…
Concordia Institute for Information Systems Engineering（康考迪亚信息系统工程研究院）；Concordia University（康考迪亚大学）
The U-Net architecture, built upon the fully convolutional network, has proven to be effective in biomedical image segmentation. However, U-Net applies skip connections to merge semantically different low- and high-level convolutional features, resulting in not only blurred feature maps, but also over- and under-segmented target regions. To address these limitations, we propose a simple, yet effective end-to-end depthwise encoder-decoder fully convolutional network architecture, called Sharp U-Net, for binary and multi-class biomedical image segmentation. The key rationale of Sharp U-Net is th…
University of Electronic Science and Technology of China（电子科技大学）
In this paper, we propose efficient and effective methods for 2D human pose estimation. A new ResBlock is proposed based on depthwise separable convolution and is utilized instead of the original one in Hourglass network. It can be further enhanced by replacing the vanilla depthwise convolution with a mixed depthwise convolution. Based on it, we propose a bottom-up multi-person pose estimation method. A rooted tree is used to represent human pose by introducing person centroid as the root which connects to all body joints directly or hierarchically. Two branches of sub-networks are used to pre…
University of Chinese Academy of Sciences（中国科学院大学）；Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences（中国科学院深圳先进技术研究院保罗·C·劳特伯生物医学成像研究中心）；Guangdong General Hospital, Guangdong Academy of Medical Sciences（广东省医学科学院广东省人民医院）；Henan Provincial People’s Hospital（河南省人民医院）；Nanchang University（南昌大学）
The morbidity of brain stroke increased rapidly in the past few years. To help specialists in lesion measurements and treatment planning, automatic segmentation methods are critically required for clinical practices. Recently, approaches based on deep learning and methods for contextual information extraction have served in many image segmentation tasks. However, their performances are limited due to the insufficient training of a large number of parameters, which sometimes fail in capturing long-range dependencies. To address these issues, we propose a depthwise separable convolution based X-…
University of Siegen（锡根大学）；University of Cambridge（剑桥大学）
An increasing number of models require the control of the spectral norm of convolutional layers of a neural network. While there is an abundance of methods for estimating and enforcing upper bounds on those during training, they are typically costly in either memory or time. In this work, we introduce a very simple method for spectral normalization of depthwise separable convolutions, which introduces negligible computational and memory overhead. We demonstrate the effectiveness of our method on image classification tasks using standard architectures like MobileNetV2.
Depthwise convolution is becoming increasingly popular in modern efficient ConvNets, but its kernel size is often overlooked. In this paper, we systematically study the impact of different kernel sizes, and observe that combining the benefits of multiple kernel sizes can lead to better accuracy and efficiency. Based on this observation, we propose a new mixed depthwise convolution (MixConv), which naturally mixes up multiple kernel sizes in a single convolution. As a simple drop-in replacement of vanilla depthwise convolution, our MixConv improves the accuracy and efficiency for existing Mobil…
Deep convolutional neural networks have been proven successful in multiple benchmark challenges in recent years. However, the performance improvements are heavily reliant on increasingly complex network architecture and a high number of parameters, which require ever increasing amounts of storage and memory capacity. Depthwise separable convolution (DSConv) can effectively reduce the number of required parameters through decoupling standard convolution into spatial and cross-channel convolution steps. However, the method causes a degradation of accuracy. To address this problem, we present dep…
Towards Lightweight Hyperspectral Image Super-Resolution with Depthwise Separable Dilated Convolutional Network
Department of Computer Science, Aalto University, Finland（艾尔沃斯大学计算机科学系）；School of Electrical and Electronic Engineering, University of Sheffield, United Kingdom（谢菲尔德大学电子与电气工程学院）
National ASIC Engineering Technology Research Center, School of Electronics Science and Engineering, Southeast University（东南大学电子科学与工程学院国家专用集成电路系统工程技术研究中心）；Singapore University of Technology and Design（新加坡科技设计大学）
In many real-time applications, the deployment of deep neural networks is constrained by high computational cost and efficient lightweight neural networks are widely concerned. In this paper, we propose that depthwise convolution (DWC) is used to expand the number of channels in a bottleneck block, which is more efficient than 1 x 1 convolution. The proposed Pointwise-Standard-Depthwise network (PSDNet) based on channel expansion with DWC has fewer number of parameters, less computational cost and higher accuracy than corresponding ResNet on CIFAR datasets. To design more efficient lightweight…
Generative Adversarial Networks have been employed successfully to generate high-resolution augmented images of size 1024^2. Although the augmented images generated are unprecedented, the training time of the model is exceptionally high. Conventional GAN requires training of both Discriminator as well as the Generator. In Progressive GAN, which is the current state-of-the-art GAN for image augmentation, instead of training the GAN all at once, a new concept of progressing growing of Discriminator and Generator simultaneously, was proposed. Although the lower stages such as 4x4 and 8x8 train ra…
Convolutional neural networks (CNNs) have shown remarkable performance in various computer vision tasks in recent years. However, the increasing model size has raised challenges in adopting them in real-time applications as well as mobile and embedded vision applications. Many works try to build networks as small as possible while still have acceptable performance. The state-of-the-art architecture is MobileNets. They use Depthwise Separable Convolution (DWConvolution) in place of standard Convolution to reduce the size of networks. This paper describes an improved version of MobileNet, called…
Selective Depthwise Separable Convolution for Lightweight Joint Source-Channel Coding in Wireless Image Transmission
Singapore University of Technology and Design（新加坡科技设计大学）；National Mobile Communications Research Laboratory（国家移动通信研究中心）；Southeast University（东南大学）；School of Electronic and Optical Engineering（电子与光学工程学院）
University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
Despite their tremendous successes, convolutional neural networks (CNNs) incur high computational/storage costs and are vulnerable to adversarial perturbations. Recent works on robust model compression address these challenges by combining model compression techniques with adversarial training. But these methods are unable to improve throughput (frames-per-second) on real-life hardware while simultaneously preserving robustness to adversarial perturbations. To overcome this problem, we propose the method of Generalized Depthwise-Separable (GDWS) convolution -- an efficient, universal, post-tra…
Indian Institute of Technology, Madras（印度理工学院马德拉斯分校）；Microsoft Research（微软研究院）
Massively parallel systolic arrays and resource-efficient depthwise separable convolutions are two promising techniques to accelerate DNN inference on the edge. Interestingly, their combination is inefficient: Computational patterns of depthwise separable convolutions do not exhibit a rhythmic systolic flow and lack sufficient data reuse to saturate systolic arrays. We formally analyse this inefficiency and propose an efficient operator, an optimal hardware dataflow, and a superior training methodology towards alleviating this. The efficient operator, called FuSeConv, is a drop-in replacement …
Accelerating Vision Foundation Models with Drop-in Depthwise Convolution
University of Modena and Reggio Emilia（摩德纳和雷吉奥艾米利亚大学）；INSAIT, Sofia University ``St. Kliment Ohridski'', Sofia, Bulgaria（INSAIT，索菲亚大学『圣克莱门特·欧赫里迪斯』，索菲亚，保加利亚）
We introduce blueprint separable convolutions (BSConv) as highly efficient building blocks for CNNs. They are motivated by quantitative analyses of kernel properties from trained models, which show the dominance of correlations along the depth axis. Based on our findings, we formulate a theoretical foundation from which we derive efficient implementations using only standard layers. Moreover, our approach provides a thorough theoretical derivation, interpretation, and justification for the application of depthwise separable convolutions (DSCs) in general, which have become the basis of many mo…
University of California, San Diego（加利福尼亚大学圣地亚哥分校）；IBM Research AI（IBM研究院人工智能部门）；University of Central Florida（中佛罗里达大学）
There is a growing interest in designing models that can deal with images from different visual domains. If there exists a universal structure in different visual domains that can be captured via a common parameterization, then we can use a single model for all domains rather than one model per domain. A model aware of the relationships between different domains can also be trained to work on new domains with less resources. However, to identify the reusable structure in a model is not easy. In this paper, we propose a multi-domain learning architecture based on depthwise separable convolution…
University of Waterloo（滑铁卢大学）；Waterloo Artificial Intelligence Institute（滑铁卢人工智能研究所）
As the "Mobile AI" revolution continues to grow, so does the need to understand the behaviour of edge-deployed deep neural networks. In particular, MobileNets are the go-to family of deep convolutional neural networks (CNN) for mobile. However, they often have significant accuracy degradation under post-training quantization. While studies have introduced quantization-aware training and other methods to tackle this challenge, there is limited understanding into why MobileNets (and potentially depthwise-separable CNNs (DWSCNN) in general) quantize so poorly compared to other CNN architectures. …
Deep learning methods have shown considerable potential for hyperspectral image (HSI) classification, which can achieve high accuracy compared with traditional methods. However, they often need a large number of training samples and have a lot of parameters and high computational overhead. To solve these problems, this paper proposes a new network architecture, LiteDepthwiseNet, for HSI classification. Based on 3D depthwise convolution, LiteDepthwiseNet can decompose standard convolution into depthwise convolution and pointwise convolution, which can achieve high classification performance wit…
Indian Institute of Technology Gandhinagar（印度理工学院甘地讷格尔分校）；Osaka University（大阪大学）
Conventional 3D convolutional neural networks (CNNs) are computationally expensive, memory intensive, prone to overfitting, and most importantly, there is a need to improve their feature learning capabilities. To address these issues, we propose spatio-temporal short term Fourier transform (STFT) blocks, a new class of convolutional blocks that can serve as an alternative to the 3D convolutional layer and its variants in 3D CNNs. An STFT block consists of non-trainable convolution layers that capture spatially and/or temporally local Fourier information using a STFT kernel at multiple low freq…
Electronics and Telecommunications Research Institute (ETRI)（韩国电子通信研究院）
Video classification researches that have recently attracted attention are the fields of temporal modeling and 3D efficient architecture. However, the temporal modeling methods are not efficient or the 3D efficient architecture is less interested in temporal modeling. For bridging the gap between them, we propose an efficient temporal modeling 3D architecture, called VoV3D, that consists of a temporal one-shot aggregation (T-OSA) module and depthwise factorized component, D(2+1)D. The T-OSA is devised to build a feature hierarchy by aggregating temporal features with different temporal recepti…
We present a logarithmic-scale efficient convolutional neural network architecture for edge devices, named WaveletNet. Our model is based on the well-known depthwise convolution, and on two new layers, which we introduce in this work: a wavelet convolution and a depthwise fast wavelet transform. By breaking the symmetry in channel dimensions and applying a fast algorithm, WaveletNet shrinks the complexity of convolutional blocks by an O(logD/D) factor, where D is the number of channels. Experiments on CIFAR-10 and ImageNet classification show superior and comparable performances of WaveletNet …
National Taiwan University of Science and Technology（台湾科技大学）
Automatic license plate recognition plays a crucial role in modern transportation systems such as for traffic monitoring and vehicle violation detection. In real-world scenarios, license plate recognition still faces many challenges and is impaired by unpredictable interference such as weather or lighting conditions. Many machine learning based ALPR solutions have been proposed to solve such challenges in recent years. However, most are not convincing, either because their results are evaluated on small or simple datasets that lack diverse surroundings, or because they require powerful hardwar…
We propose a random convolutional neural network to generate a feature space in which we study image classification and retrieval performance. Put briefly we apply random convolutional blocks followed by global average pooling to generate a new feature, and we repeat this k times to produce a k-dimensional feature space. This can be interpreted as partitioning the space of image patches with random hyperplanes which we formalize as a random depthwise convolutional neural network. In the network's final layer we perform image classification and retrieval with the linear support vector machine a…
Convolutional neural network (CNN) has achieved great success on image super-resolution (SR). However, most deep CNN-based SR models take massive computations to obtain high performance. Downsampling features for multi-resolution fusion is an efficient and effective way to improve the performance of visual recognition. Still, it is counter-intuitive in the SR task, which needs to project a low-resolution input to high-resolution. In this paper, we propose a novel Hybrid Pixel-Unshuffled Network (HPUN) by introducing an efficient and effective downsampling module into the SR task. The network c…
Parameter-Efficient 3D Segmentation of Liver and Liver tumors: Depthwise factorization Scales Better Than Dense Convolution with Spatial Dimensionality
Convolutional neural network (CNN) offers significant accuracy in image detection. To implement image detection using CNN in the internet of things (IoT) devices, a streaming hardware accelerator is proposed. The proposed accelerator optimizes the energy efficiency by avoiding unnecessary data movement. With unique filter decomposition technique, the accelerator can support arbitrary convolution window size. In addition, max pooling function can be computed in parallel with convolution by using separate pooling unit, thus achieving throughput improvement. A prototype accelerator was implemente…
University of Louisiana at Lafayette（路易斯安那大学拉法叶分校）
While convolutional neural networks (CNNs) demonstrate outstanding performance on computer vision tasks, their computational costs remain high. Several techniques are used to reduce these costs, like reducing channel count, and using separable and depthwise separable convolutions. This paper reduces computational costs by introducing a novel architecture, axial CNNs, which replaces spatial 2D convolution operations with two consecutive depthwise separable 1D operations. The axial CNNs are predicated on the assumption that the dataset supports approximately separable convolution operations with…
IMSE: Efficient U-Net-based Speech Enhancement using Inception Depthwise Convolution and Amplitude-Aware Linear Attention
College of Electronics and Information Engineering, Shenzhen University（深圳大学电子与信息工程学院）；Shenzhen, China Information Center, Shenzhen University（深圳大学深圳信息中心）
The University of Hong Kong（香港大学）；Sun Yat-sen University（中山大学）
Recently deep convolutional neural networks have achieved significant success in salient object detection. However, existing state-of-the-art methods require high-end GPUs to achieve real-time performance, which makes them hard to adapt to low-cost or portable devices. Although generic network architectures have been proposed to speed up inference on mobile devices, they are tailored to the task of image classification or semantic segmentation, and struggle to capture intra-channel and inter-channel correlations that are essential for contrast modeling in salient object detection. Motivated by…
University of Wisconsin-Madison（威斯康星大学麦迪逊分校）；Amazon Lab126（亚马逊Lab126）；Korea University（高丽大学）；Amazon Inc（亚马逊公司）
Deep convolutional neural networks have achieved remarkable success in computer vision. However, deep neural networks require large computing resources to achieve high performance. Although depthwise separable convolution can be an efficient module to approximate a standard convolution, it often leads to reduced representational power of networks. In this paper, under budget constraints such as computational cost (MAdds) and the parameter count, we propose a novel basic architectural block, ANTBlock. It boosts the representational power by modeling, in a high dimensional space, interdependency…
WBMM: Windowed Batch Matrix Multiplication for Efficient Large Receptive Field Convolution
Hefei University of Technology, Hefei, China（合肥工业大学）；Lingyang Industrial Internet Co., Ltd., Hefei, China（灵阳工业互联网有限公司）；Kochi University of Technology, Kochi, Japan（Kochi大学技术）
Shift operation is an efficient alternative over depthwise separable convolution. However, it is still bottlenecked by its implementation manner, namely memory movement. To put this direction forward, a new and novel basic component named Sparse Shift Layer (SSL) is introduced in this paper to construct efficient convolutional neural networks. In this family of architectures, the basic block is only composed by 1x1 convolutional layers with only a few shift operations applied to the intermediate feature maps. To make this idea feasible, we introduce shift operation penalty during optimization …
Carnegie Mellon University（卡内基梅隆大学）；Bosch Center for AI（博世人工智能中心）
Neural network weights are typically initialized at random from univariate distributions, controlling just the variance of individual weights even in highly-structured operations like convolutions. Recent ViT-inspired convolutional networks such as ConvMixer and ConvNeXt use large-kernel depthwise convolutions whose learned filters have notable structure; this presents an opportunity to study their empirical covariances. In this work, we first observe that such learned filters have highly-structured covariance matrices, and moreover, we find that covariances calculated from small networks may …
Neural Architecture Search (NAS), together with model scaling, has shown remarkable progress in designing high accuracy and fast convolutional architecture families. However, as neither NAS nor model scaling considers sufficient hardware architecture details, they do not take full advantage of the emerging datacenter (DC) accelerators. In this paper, we search for fast and accurate CNN model families for efficient inference on DC accelerators. We first analyze DC accelerators and find that existing CNNs suffer from insufficient operational intensity, parallelism, and execution efficiency. Thes…
Huazhong University of Science and Technology（华中科技大学）
Modern efficient Convolutional Neural Networks(CNNs) always use Depthwise Separable Convolutions(DSCs) and Neural Architecture Search(NAS) to reduce the number of parameters and the computational complexity. But some inherent characteristics of networks are overlooked. Inspired by visualizing feature maps and N$\times$N(N$>$1) convolution kernels, several guidelines are introduced in this paper to further improve parameter efficiency and inference speed. Based on these guidelines, our parameter-efficient CNN architecture, called \textit{VGNetG}, achieves better accuracy and lower latency than …
Photonic Microring Resonator (MRR) based hardware accelerators have been shown to provide disruptive speedup and energy-efficiency improvements for processing deep Convolutional Neural Networks (CNNs). However, previous MRR-based CNN accelerators fail to provide efficient adaptability for CNNs with mixed-sized tensors. One example of such CNNs is depthwise separable CNNs. Performing inferences of CNNs with mixed-sized tensors on such inflexible accelerators often leads to low hardware utilization, which diminishes the achievable performance and energy efficiency from the accelerators. In this …
Graph Convolutional Networks (GCNs) have been widely demonstrated their powerful ability in graph data representation and learning. Existing graph convolution layers are mainly designed based on graph signal processing and transform aspect which usually suffer from some limitations, such as over-smoothing, over-squashing and non-robustness, etc. As we all know that Convolution Neural Networks (CNNs) have received great success in many computer vision and machine learning. One main aspect is that CNNs leverage many learnable convolution filters (kernels) to obtain rich feature descriptors and t…
