Historically, the pursuit of efficient inference has been one of the driving forces behind research into new deep learning architectures and building blocks. Some recent examples include: the squeeze-and-excitation module, depthwise separable convolutions in Xception, and the inverted bottleneck in MobileNet v2. Notably, in all of these cases, the resulting building blocks enabled not only higher efficiency, but also higher accuracy, and found wide adoption in the field. In this work, we further expand the arsenal of efficient building blocks for neural network architectures; but instead of combining standard primitives (such as convolution), we advocate for the replacement of these dense primitives with their sparse counterparts. While the idea of using sparsity to decrease the parameter count is not new, the conventional wisdom is that this reduction in theoretical FLOPs does not translate into real-world efficiency gains. We aim to correct this misconception by introducing a family of efficient sparse kernels for ARM and WebAssembly, which we open-source for the benefit of the community as part of the XNNPACK library. Equipped with our efficient implementation of sparse primitives, we show that sparse versions of MobileNet v1, MobileNet v2 and EfficientNet architectures substantially outperform strong dense baselines on the efficiency-accuracy curve. On Snapdragon 835 our sparse networks outperform their dense equivalents by $1.3-2.4\times$ -- equivalent to approximately one entire generation of MobileNet-family improvement. We hope that our findings will facilitate wider adoption of sparsity as a tool for creating efficient and accurate deep learning architectures.
SenseTime Group（商汤科技集团）；The Chinese University of Hong Kong（香港中文大学）；Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences（中国科学院深圳先进技术研究院）
This paper proposes to learn high-performance deep ConvNets with sparse neural connections, referred to as sparse ConvNets, for face recognition. The sparse ConvNets are learned in an iterative way, each time one additional layer is sparsified and the entire model is re-trained given the initial weights learned in previous iterations. One important finding is that directly training the sparse ConvNet from scratch failed to find good solutions for face recognition, while using a previously learned denser model to properly initialize a sparser model is critical to continue learning effective fea…
3D neural networks have become prevalent for many 3D vision tasks including object detection, segmentation, registration, and various perception tasks for 3D inputs. However, due to the sparsity and irregularity of 3D data, custom 3D operators or network designs have been the primary focus of research, while the size of networks or efficacy of parameters has been overlooked. In this work, we perform the first comprehensive study on the weight sparsity of spatially sparse 3D convolutional networks and propose a compact weight-sparse and spatially sparse 3D convnet (WS^3-Convnet) for semantic an…
We revisit the idea of brain damage, i.e. the pruning of the coefficients of a neural network, and suggest how brain damage can be modified and used to speedup convolutional layers. The approach uses the fact that many efficient implementations reduce generalized convolutions to matrix multiplications. The suggested brain damage process prunes the convolutional kernel tensor in a group-wise fashion by adding group-sparsity regularization to the standard training process. After such group-wise pruning, convolutions can be reduced to multiplications of thinned dense matrices, which leads to spee…
Fast-ParC: Capturing Position Aware Global Feature for ConvNets and ViTs
Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））；Hong Kong Polytechnic University（香港理工大学）；Northwestern Polytechnical University（西北工业大学）；Intellifusion（云天励飞）
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…
The implementation of Deep Convolutional Neural Networks (ConvNets) on tiny end-nodes with limited non-volatile memory space calls for smart compression strategies capable of shrinking the footprint yet preserving predictive accuracy. There exist a number of strategies for this purpose, from those that play with the topology of the model or the arithmetic precision, e.g. pruning and quantization, to those that operate a model agnostic compression, e.g. weight encoding. The tighter the memory constraint, the higher the probability that these techniques alone cannot meet the requirement, hence m…
In many robotics and VR/AR applications, 3D-videos are readily-available sources of input (a continuous sequence of depth images, or LIDAR scans). However, those 3D-videos are processed frame-by-frame either through 2D convnets or 3D perception algorithms. In this work, we propose 4-dimensional convolutional neural networks for spatio-temporal perception that can directly process such 3D-videos using high-dimensional convolutions. For this, we adopt sparse tensors and propose the generalized sparse convolution that encompasses all discrete convolutions. To implement the generalized sparse conv…
Learning image representations with ConvNets by pre-training on ImageNet has proven useful across many visual understanding tasks including object detection, semantic segmentation, and image captioning. Although any image representation can be applied to video frames, a dedicated spatiotemporal representation is still vital in order to incorporate motion patterns that cannot be captured by appearance based models alone. This paper presents an empirical ConvNet architecture search for spatiotemporal feature learning, culminating in a deep 3-dimensional (3D) Residual ConvNet. Our proposed archit…
The Chinese University of Hong Kong（香港中文大学）；SenseTime Research（商汤科技研究院）
Normalization methods improve both optimization and generalization of ConvNets. To further boost performance, the recently-proposed switchable normalization (SN) provides a new perspective for deep learning: it learns to select different normalizers for different convolution layers of a ConvNet. However, SN uses softmax function to learn importance ratios to combine normalizers, leading to redundant computations compared to a single normalizer. This work addresses this issue by presenting Sparse Switchable Normalization (SSN) where the importance ratios are constrained to be sparse. Unlike $\e…
Efficient Deformable ConvNets: Rethinking Dynamic and Sparse Operator for Vision Applications
OpenGVLab, Shanghai AI Laboratory（上海人工智能实验室 OpenGVLab）；Tsinghua University（清华大学）；SenseTime Research（商汤科技研究院）；The Chinese University of Hong Kong（香港中文大学）；Nanjing University（南京大学）；University of Toronto（多伦多大学）；CAIR, HKISI, CAS（中国科学院香港创新研究院人工智能与机器人创新中心）
Facebook AI Research（脸书人工智能研究院）；Dartmouth College（达特茅斯学院）
We propose a simple, yet effective approach for spatiotemporal feature learning using deep 3-dimensional convolutional networks (3D ConvNets) trained on a large scale supervised video dataset. Our findings are three-fold: 1) 3D ConvNets are more suitable for spatiotemporal feature learning compared to 2D ConvNets; 2) A homogeneous architecture with small 3x3x3 convolution kernels in all layers is among the best performing architectures for 3D ConvNets; and 3) Our learned features, namely C3D (Convolutional 3D), with a simple linear classifier outperform state-of-the-art methods on 4 different …
Peking University（北京大学）；Bytedance Inc.（字节跳动公司）；University of Oxford（牛津大学）
We identify and overcome two key obstacles in extending the success of BERT-style pre-training, or the masked image modeling, to convolutional networks (convnets): (i) convolution operation cannot handle irregular, random-masked input images; (ii) the single-scale nature of BERT pre-training is inconsistent with convnet's hierarchical structure. For (i), we treat unmasked pixels as sparse voxels of 3D point clouds and use sparse convolution to encode. This is the first use of sparse convolution for 2D masked modeling. For (ii), we develop a hierarchical decoder to reconstruct images from multi…
Semantic image segmentation plays a pivotal role in many vision applications including autonomous driving and medical image analysis. Most of the former approaches move towards enhancing the performance in terms of accuracy with a little awareness of computational efficiency. In this paper, we introduce LiteSeg, a lightweight architecture for semantic image segmentation. In this work, we explore a new deeper version of Atrous Spatial Pyramid Pooling module (ASPP) and apply short and long residual connections, and depthwise separable convolution, resulting in a faster and efficient model. LiteS…
Deep 3-dimensional (3D) Convolutional Network (ConvNet) has shown promising performance on video recognition tasks because of its powerful spatio-temporal information fusion ability. However, the extremely intensive requirements on memory access and computing power prohibit it from being used in resource-constrained scenarios, such as portable and edge devices. So in this paper, we first propose a two-stage Fully Separable Block (FSB) to significantly compress the model sizes of 3D ConvNets. Then a feature enhancement approach named Temporal Residual Gradient (TRG) is developed to improve the …
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 expensive 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 u…
Hong Kong University of Science and Technology（香港科技大学）
3D point cloud interpretation is a challenging task due to the randomness and sparsity of the component points. Many of the recently proposed methods like PointNet and PointCNN have been focusing on learning shape descriptions from point coordinates as point-wise input features, which usually involves complicated network architectures. In this work, we draw attention back to the standard 3D convolutions towards an efficient 3D point cloud interpretation. Instead of converting the entire point cloud into voxel representations like the other volumetric methods, we voxelize the sub-portions of th…
Beijing National Research Center for Information Science and Technology (BNRist)（北京信息科学与技术国家研究中心）；School of Software（软件学院）；Tsinghua University（清华大学）；MEGVII Technology（旷视科技）；Hong Kong University of Science and Technology（香港科技大学）；Aberystwyth University（阿伯里斯特威斯大学）
We present a simple but powerful architecture of convolutional neural network, which has a VGG-like inference-time body composed of nothing but a stack of 3x3 convolution and ReLU, while the training-time model has a multi-branch topology. Such decoupling of the training-time and inference-time architecture is realized by a structural re-parameterization technique so that the model is named RepVGG. On ImageNet, RepVGG reaches over 80% top-1 accuracy, which is the first time for a plain model, to the best of our knowledge. On NVIDIA 1080Ti GPU, RepVGG models run 83% faster than ResNet-50 or 101…
U-DECN: End-to-End Underwater Object Detection ConvNet with Improved DeNoising Training
National Key Laboratory of Autonomous Marine Vehicle Technology, Harbin Engineering University（自主水下机动车辆技术国家重点实验室，哈尔滨工程大学）；Hong Kong Polytechnic University（香港理工大学）
Beijing University of Posts and Telecommunications（北京邮电大学）；ByteDance AI Lab（字节跳动人工智能实验室）
Although convolutional networks (ConvNets) have enjoyed great success in computer vision (CV), it suffers from capturing global information crucial to dense prediction tasks such as object detection and segmentation. In this work, we innovatively propose ConTNet (ConvolutionTransformer Network), combining transformer with ConvNet architectures to provide large receptive fields. Unlike the recently-proposed transformer-based models (e.g., ViT, DeiT) that are sensitive to hyper-parameters and extremely dependent on a pile of data augmentations when trained from scratch on a midsize dataset (e.g.…
WaveDH: Wavelet Sub-bands Guided ConvNet for Efficient Image Dehazing
Artificial Intelligence Graduate School, Gwangju Institute of Science and Technology (GIST)（光州科学技术院人工智能研究生院）；School of Electronic Engineering, Xidian University（西安电子科技大学电子工程学院）；School of Electrical Engineering and Computer Science, Gwangju Institute of Science and Technology (GIST)（光州科学技术院电气与计算机工程学院）
Almost all previous deep learning-based multi-view stereo (MVS) approaches focus on improving reconstruction quality. Besides quality, efficiency is also a desirable feature for MVS in real scenarios. Towards this end, this paper presents a Fast-MVSNet, a novel sparse-to-dense coarse-to-fine framework, for fast and accurate depth estimation in MVS. Specifically, in our Fast-MVSNet, we first construct a sparse cost volume for learning a sparse and high-resolution depth map. Then we leverage a small-scale convolutional neural network to encode the depth dependencies for pixels within a local reg…
We describe a method for fast approximation of sparse coding. The input space is subdivided by a binary decision tree, and we simultaneously learn a dictionary and assignment of allowed dictionary elements for each leaf of the tree. We store a lookup table with the assignments and the pseudoinverses for each node, allowing for very fast inference. We give an algorithm for learning the tree, the dictionary and the dictionary element assignment, and In the process of describing this algorithm, we discuss the more general problem of learning the groups in group structured sparse modelling. We sho…
The University of Hong Kong（香港大学）；Tencent Games（腾讯游戏）；Texas A&M University（得克萨斯农工大学）
We introduce SparseNeuS, a novel neural rendering based method for the task of surface reconstruction from multi-view images. This task becomes more difficult when only sparse images are provided as input, a scenario where existing neural reconstruction approaches usually produce incomplete or distorted results. Moreover, their inability of generalizing to unseen new scenes impedes their application in practice. Contrarily, SparseNeuS can generalize to new scenes and work well with sparse images (as few as 2 or 3). SparseNeuS adopts signed distance function (SDF) as the surface representation,…
Sparse approximations using highly over-complete dictionaries is a state-of-the-art tool for many imaging applications including denoising, super-resolution, compressive sensing, light-field analysis, and object recognition. Unfortunately, the applicability of such methods is severely hampered by the computational burden of sparse approximation: these algorithms are linear or super-linear in both the data dimensionality and size of the dictionary. We propose a framework for learning the hierarchical structure of over-complete dictionaries that enables fast computation of sparse representations…
King Abdullah University of Science and Technology (KAUST)（沙特国王 Abdullah 科技大学（KAUST））
Convolutional sparse coding (CSC) is an important building block of many computer vision applications ranging from image and video compression to deep learning. We present two contributions to the state of the art in CSC. First, we significantly speed up the computation by proposing a new optimization framework that tackles the problem in the dual domain. Second, we extend the original formulation to higher dimensions in order to process a wider range of inputs, such as RGB images and videos. Our results show up to 20 times speedup compared to current state-of-the-art CSC solvers.
Limited by the performance factor, it is arduous to recognize target object and locate it in Augmented Reality (AR) scenes on low-end mobile devices, especially which using monocular cameras. In this paper, we proposed an algorithm which can quickly locate the target object through image object detection in the circumstances of having very sparse feature points. We introduce YOLOv3-Tiny to our algorithm as the object detection module to filter the possible points and using Principal Component Analysis (PCA) to determine the location. We conduct the experiment in a manually designed scene by ho…
Johns Hopkins University（约翰斯·霍普金斯大学）；West Virginia University（西弗吉尼亚大学）
Sparse representation-based classifiers have shown outstanding accuracy and robustness in image classification tasks even with the presence of intense noise and occlusion. However, it has been discovered that the performance degrades significantly either when test image is not aligned with the dictionary atoms or the dictionary atoms themselves are not aligned with each other, in which cases the sparse linear representation assumption fails. In this paper, having both training and test images misaligned, we introduce a novel sparse coding framework that is able to efficiently adapt the diction…
Massachusetts Institute of Technology（麻省理工学院）；Princeton University（普林斯顿大学）
Convolutional networks (ConvNets) have become a popular approach to computer vision. It is important to accelerate ConvNet training, which is computationally costly. We propose a novel parallel algorithm based on decomposition into a set of tasks, most of which are convolutions or FFTs. Applying Brent's theorem to the task dependency graph implies that linear speedup with the number of processors is attainable within the PRAM model of parallel computation, for wide network architectures. To attain such performance on real shared-memory machines, our algorithm computes convolutions converging o…
Instituto de Telecomunicações（电信研究院）；Instituto Superior Técnico（高等技术学院）；Universidade de Lisboa（里斯本大学）
This paper introduces two very fast and competitive hyperspectral image (HSI) restoration algorithms: fast hyperspectral denoising (FastHyDe), a denoising algorithm able to cope with Gaussian and Poissonian noise, and fast hyperspectral inpainting (FastHyIn), an inpainting algorithm to restore HSIs where some observations from known pixels in some known bands are missing. FastHyDe and FastHyIn fully exploit extremely compact and sparse HSI representations linked with their low-rank and self-similarity characteristics. In a series of experiments with simulated and real data, the newly introduce…
Courant Institute of Mathematical Sciences（柯朗数学科学研究所）；New York University（纽约大学）
Adaptive sparse coding methods learn a possibly overcomplete set of basis functions, such that natural image patches can be reconstructed by linearly combining a small subset of these bases. The applicability of these methods to visual object recognition tasks has been limited because of the prohibitive cost of the optimization algorithms required to compute the sparse representation. In this work we propose a simple and efficient algorithm to learn basis functions. After training, this model also provides a fast and smooth approximator to the optimal representation, achieving even better accu…
Michigan State University（密歇根州立大学）；EPFL（洛桑联邦理工学院）；University of Applied Sciences Western Switzerland (HES-SO)（瑞士西部应用科学大学）
We propose an algorithm for rotational sparse coding along with an efficient implementation using steerability. Sparse coding (also called dictionary learning) is an important technique in image processing, useful in inverse problems, compression, and analysis; however, the usual formulation fails to capture an important aspect of the structure of images: images are formed from building blocks, e.g., edges, lines, or points, that appear at different locations, orientations, and scales. The sparse coding problem can be reformulated to explicitly account for these transforms, at the cost of incr…
Massachusetts General Hospital（麻省总医院）；Harvard Medical School（哈佛医学院）
We introduce SparseVM, a method that registers clinical-quality 3D MR scans both faster and more accurately than previously possible. Deformable alignment, or registration, of clinical scans is a fundamental task for many clinical neuroscience studies. However, most registration algorithms are designed for high-resolution research-quality scans. In contrast to research-quality scans, clinical scans are often sparse, missing up to 86% of the slices available in research-quality scans. Existing methods for registering these sparse images are either inaccurate or extremely slow. We present a lear…
