共查询到20条相似文献,搜索用时 15 毫秒
1.
A new pruning method for an FFT type of transform structure is proposed. Its novelty lies in the fact that, besides being able to prune the transform, it is able to complete a previously pruned transform or to progress from one level of pruning to another. The method can be directly applied to fast progressive image coding 相似文献
2.
The authors focus on the use of neural networks to approximate continuous decision functions. In this context, the parameters to be estimated are the synaptic weights of the network. The number of such parameters and the quantity of data (information) available for training greatly influence the quality of the solution obtained. A previous study analysed the influence and interaction of these two features. In order to reach the architecture of the net leading to the best fitting of the training data, two original pruning techniques are proposed. The evolution of the neural network performances, training and test rates, as the number of synaptic weights pruned increases, is shown experimentally. Two kinds of synaptic weights are obvious: irrelevant synaptic weights, which can be suppressed from the model; and relevant synaptic weights, which cannot be removed. In the test problem, it is possible to reduce the size of the network up to 42%. A 4% improvement of the performance in generalisation is observed 相似文献
3.
通道剪枝是深度模型压缩的主要方法之一。针对密集连接卷积神经网络中,每一层都接收其前部所有卷积层的输出特征图作为输入,但并非每个后部层都需要所有先前层的特征,网络中存在很大冗余的缺点。本文提出一种自学习剪枝密集连接网络中冗余通道的方法,得到稀疏密集连接卷积神经网络。首先,提出了一种衡量每个卷积层中每个输入特征图对输出特征图贡献度大小的方法,贡献度小的输入特征图即为冗余特征图;其次,介绍了通过自学习,网络分阶段剪枝冗余通道的训练过程,得到了稀疏密集连接卷积神经网络,该网络剪枝了密集连接网络中的冗余通道,减少了网络参数,降低了存储和计算量;最后,为了验证本文方法的有效性,在图像分类数据集CIFAR-10/100上进行了实验,在不牺牲准确率的前提下减小了模型冗余。 相似文献
4.
飞机目标识别是地面情报系统的一项重要关键技术。近年来火热的深度学习方法,如卷积神经网络,展现出对于图像识别任务的优越性能。但是,训练卷积神经网络需要大量的带标签样本以估计规模庞大的模型参数,因而限制了其在雷达目标识别领域中的应用。针对飞机目标识别中的小样本问题,文中引入适用于有限数据场景的迁移学习技术,预先在其他大样本高分辨距离像数据上训练一个初始卷积神经网络模型,再结合当前飞机目标识别任务调优模型参数。在实测数据上的实验结果显示,与仅使用卷积神经网络的方法相比,所提方法可显著提升识别准确率,验证了方法的有效性。 相似文献
5.
Deep convolutional neural networks (CNNs) have demonstrated its extraordinary power on various visual tasks like object detection and classification. However, it is still challenging to deploy state-of-the-art models into real-world applications, such as autonomous vehicles, due to their expensive computation costs. In this paper, to accelerate the network inference, we introduce a novel pruning method named Drop-path to reduce model parameters of 2D deep CNNs. Given a trained deep CNN, pruning paths with different lengths is achieved by ordering the influence of neurons in each layer on the probably approximately correct (PAC) Bayesian boundary of the model. We believe that the invariance of PAC-Bayesian boundary is an important factor to guarantee the generalization ability of deep CNN under the condition of optimizing as much as possible. To the best of our knowledge, this is the first time to reduce model size based on the generalization error boundary. After pruning, we observe that the convolutional kernels themselves become sparse, rather than some being removed directly. In fact, Drop-path is generic and can be well generalized on multi-layer and multi-branch models, since parameter ranking criterion can be applied to any kind of layer and the importance scores can still be propagated. Finally, Drop-path is evaluated on two image classification benchmark datasets (ImageNet and CIFAR-10) with multiple deep CNN models, including AlexNet, VGG-16, GoogLeNet, and ResNet-34/50/56/110. Experimental results demonstrate that Drop-path achieves significant model compression and acceleration with negligible accuracy loss. 相似文献
6.
为了解决高光谱图像领域中,传统卷积神经网络因部分特征信息损失而影响最终地物分类精度的问题,采用一种基于2维和3维的混合卷积神经网络的高光谱图像分类方法,从空间增强、光谱-空间两方面分别进行了特征提取.首先从空间增强角度提出一种3维-2维卷积神经网络混合结构,得到增强后的空间信息;其次从光谱-空间角度利用3维卷积网络结构... 相似文献
7.
针对特殊材料能伪造手指静脉从而欺骗识别系统,以及利用卷积神经网络进行手指静脉识别计算量大的问题,设计了具有活体检测功能和轻量化卷积神经网络结构的手指静脉识别系统。采用光容积法检测手指脉搏波的变化,从而判断被采集对象是否为活体;利用剪枝及通道恢复方法改进了ResNet-18卷积神经网络,并结合 L1正则化增加卷积神经网络的特征选择能力,在提升算法准确率的基础上,能有效地降低计算资源的消耗。实验表明,使用改进的剪枝及通道恢复优化结构,参数量降低了75.6%,计算量降低了25.6%,在山东大学和香港理工大学手指静脉数据库上得到的等误率分别为0.025%、0.085%,远低于ResNet-18得到的等误率(0.117%、0.213%)。 相似文献
8.
In this paper, puncturing and path pruning are combined for convolutional codes to construct a new coding scheme for unequal error protection (UEP), called the hybrid punctured and path-pruned convolutional codes. From an algebraic viewpoint, we show that the hybrid codes not only inherit all the advantages of the conventional rate-compatible punctured convolutional codes and path-compatible pruned convolutional codes but also can provide more flexible choices of protection capability for UEP. In addition, a data-multiplexing scheme originally proposed for path-pruned codes which can guarantee smooth transition between rates without additional zero-padding for frame termination is proven applicable to the hybrid codes to improve the system throughput. 相似文献
9.
Convolutional neural networks (CNNs) with large model size and computing operations are difficult to be deployed on embedded systems, such as smartphones or AI cameras. In this paper, we propose a novel structured pruning method, termed the structured feature sparsity training (SFST), to speed up the inference process and reduce the memory usage of CNNs. Unlike other existing pruning methods, which require multiple iterations of pruning and retraining to ensure stable performance, SFST only needs to fine-tune the pretrained model with additional regularization on the less important features and then prune them, no multiple pruning and retraining needed. SFST can be deployed to a variety of modern CNN architectures including VGGNet, ResNet and MobileNetv2. Experimental results on CIFAR, SVHN, ImageNet and MSTAR benchmark dataset demonstrate the effectiveness of our scheme, which achieves superior performance over the state-of-the-art methods. 相似文献
10.
We present methods for learning and pruning oblique decision trees. We propose a new function for evaluating different split rules at each node while growing the decision tree. Unlike the other evaluation functions currently used in the literature (which are all based on some notion of purity of a node), this new evaluation function is based on the concept of degree of linear separability. We adopt a correlation based optimization technique called the Alopex algorithm (K.P. Unnikrishnaan and K.P. Venugopal, 1994) for finding the split rule that optimizes our evaluation function at each node. The algorithm we present is applicable only for 2-class problems. Through empirical studies, we demonstrate that our algorithm learns good compact decision trees. We suggest a representation scheme for oblique decision trees that makes explicit the fact that an oblique decision tree represents each class as a union of convex sets bounded by hyperplanes in the feature space. Using this representation, we present a new pruning technique. Unlike other pruning techniques, which generally replace heuristically selected subtrees of the original tree by leaves, our method can radically restructure the decision tree. Through empirical investigation, we demonstrate the effectiveness of our method 相似文献
11.
We address the problem of network pruning for extending the service life of satellites in LEO constellations. Satellites in LEO constellations can spend over 30 % of their time under the earth’s umbra, time during which they are powered by batteries. While the batteries are recharged by solar energy, the depth of discharge they reach during eclipse significantly affects their lifetime—and by extension, the service life of the satellites themselves. For batteries of the type that power Iridium satellites, a 15 % increase to the depth of discharge can practically cut their service lives in half. In this paper, we present the design and evaluation of two forms of network pruning schemes that reduce the energy consumption of LEO satellite network. First, we propose a new lightweight traffic-agnostic metric for quantifiying the quality of a frugal topology, the Adequacy Index (ADI). After showing that the problem of minimizing the power consumption of a LEO network subject to a given ADI threshold is NP-hard, we propose heuristcs to solve it. Second, we propose traffic-aware metric for quantifiying the quality of a frugal topology, the maximum link utilization (MLU). Also, with the problem being NP-hard subject to a given MLU threshold, we propose heuristics to solve it. We evaluate both forms using realistic LEO topologies and traffic matrices. Results show that traffic-agnostic pruning and traffic-aware pruning can increase the satellite service life by as much as 85 and 80 %, respectively. This is accomplished by trading off very little in terms of average path length and congestion. 相似文献
12.
基于深度神经网络的低资源条件下关键词检索已经取得了很大的进展,但这些方法仍旧需要较多的参数才能保证模型的精度。为了进一步减少模型的参数量,本文将Squeeze-and-Excitation网络和深度可分离卷积应用在关键词检索任务中。首先利用Squeeze-and-Excitation网络对不同特征通道之间的相互依赖关系建模的能力进一步提升模型的精度,然后通过将标准卷积替换为深度可分离卷积来有效的减少模型所需要的参数。在谷歌语音命令数据集上的实验证明我们的模型可以在保证高精度的同时把参数量限制在一定的范围内。 相似文献
13.
Unconstrained face verification aims to verify whether two specify images contain the same person. In this paper, we propose a deep Bayesian convolutional neural network (DBCNN) framework to extract facial features and measure their similarity for face verification in unconstrained conditions. Specifically, we design a deep convolutional neural network and construct a Bayesian probabilistic model by transferring the Bayesian likelihood ratio function into linear decision function. By training a decision line rather than finding a suitable threshold, we further enlarge the distances between inter-class and intra-class in unconstrained environment. Finally, we comprehensively evaluate our method on LFW, CACD-VS and MegaFace datasets. The test results on LFW and CACD-VS datasets show that our method can shrink intra-class variations significantly. The performance of our DBCNN model on MegaFace dataset proves that our model can achieve comparable performance to state-of-the-art methods on face verification with relative small training data and only one single network. 相似文献
14.
In this paper we study the problem of pruning a binary tree by minimizing, over all pruned subtrees of the given tree, an objective function that combines an additive cost term with a penalty term that depends only on tree size. We present algorithms for general size-based penalties, although our focus is on subadditive penalties (roughly, penalties that grow more slowly than linear penalties with increasing tree size). Such penalties are motivated by recent results in statistical learning theory for decision trees, but may have wider application as well. We show that the family of pruned subtrees induced by a subadditive penalty is a subset of the family induced by an additive penalty. This implies (by known results about additive penalties) that the family induced by a subadditive penalty 1) is nested; 2) is unique; and 3) can be computed efficiently. It also implies that, when a single tree is to be selected by cross-validation from the family of prunings, subadditive penalties will never present a richer set of options than an additive penalty. 相似文献
15.
The number of short videos on the Internet is huge, but most of them are unlabeled. In this paper, a rough labelling method of short video based on the neural network of image classification is proposed. Convolutional auto-encoder is applied to train and learn unlabeled video frames, in order to obtain the feature in certain level of the network. Using these features, we extract key-frames of the video by our method of feature clustering. We put these key-frames which represent the video content into the image classification network, so that we can get the labels for every video clip. We also compare the different architectures of convolutional auto-encoder, while optimizing and selecting the better performance architecture through our experiment result. In addition, the video frame feature from the convolutional auto-encoder is compared with those features from other extraction methods. On the whole, this paper propose a method of image labels transferring for the realization of short video rough labelling, which can be applied to the video classes with few labeled samples. 相似文献
16.
为了针对高光谱图像中空间信息与光谱信息的不同特性进行特征提取,提出一种3维卷积递归神经网络(3-D-CRNN)的高光谱图像分类方法。首先采用3维卷积神经网络提取目标像元的局部空间特征信息,然后利用双向循环神经网络对融合了局部空间信息的光谱数据进行训练,提取空谱联合特征,最后使用Softmax损失函数训练分类器实现分类。3-D-CRNN模型无需对高光谱图像进行复杂的预处理和后处理,可以实现端到端的训练,并且能够充分提取空间与光谱数据中的语义信息。结果表明,与其它基于深度学习的分类方法相比,本文中的方法在Pavia University与Indian Pines数据集上分别取得了99.94%和98.81%的总体分类精度,有效地提高了高光谱图像的分类精度与分类效果。该方法对高光谱图像的特征提取具有一定的启发意义。 相似文献
17.
为了提高监控场景中行人检测的准确度,提出了一种基于上下文信息的行人检测方法.该方法将监控场景的上下文信息融入到卷积神经网络中,选择性地学习对行人检测有帮助的上下文信息.首先,利用一个截断的卷积神经网络提取输入图像的多张特征图.然后,将多张特征图通过两个包含上下文信息的卷积层,形成一张掩码图.最后,通过在掩码图上估计行人的边界框,获得行人检测的结果.实验表明,该方法能实现监控场景中准确且快速的行人检测. 相似文献
18.
Vehicle re-identification (V-ReID) aims at discovering an image of a specific vehicle from a set of images typically captured by different cameras. Vehicles are one of the most important objects in cross-camera target recognition systems, and recognizing them is one of the most difficult tasks due to the subtle differences in the visible characteristics of vehicle rigid objects. Compared to various methods that can improve re-identification accuracy, data augmentation is a more straightforward and effective technique. In this paper, we propose a novel data synthesis method for V-ReID based on local-region perspective transformation, transformation state adversarial learning and a candidate pool. Specifically, we first propose a parameter generator network, which is a lightweight convolutional neural network, to generate the transformation states. Secondly, an adversarial module is designed in our work, it ensures that noise information is added as much as possible while keeping the labeling and structure of the dataset intact. With this adversarial module, we are able to promote the performance of the network and generate more proper and harder training samples. Furthermore, we use a candidate pool to store harder samples for further selection to improve the performance of the model. Our system pays more balanced attention to the features of vehicles. Extensive experiments show that our method significantly boosts the performance of V-ReID on the VeRi-776, VehicleID and VERI-Wild datasets. 相似文献
19.
Traffic sign recognition (TSR) is an important component of automated driving systems. It is a rather challenging task to design a high-performance classifier for the TSR system. In this paper, we propose a new method for TSR system based on deep convolutional neural network. In order to enhance the expression of the network, a novel structure (dubbed block-layer below) which combines network-in-network and residual connection is designed. Our network has 10 layers with parameters (block-layer seen as a single layer): the first seven are alternate convolutional layers and block-layers, and the remaining three are fully-connected layers. We train our TSR network on the German traffic sign recognition benchmark (GTSRB) dataset. To reduce overfitting, we perform data augmentation on the training images and employ a regularization method named “dropout”. The activation function we employ in our network adopts scaled exponential linear units (SELUs), which can induce self-normalizing properties. To speed up the training, we use an efficient GPU to accelerate the convolutional operation. On the test dataset of GTSRB, we achieve the accuracy rate of 99.67%, exceed-ing the state-of-the-art results. 相似文献
20.
高质量的数据是深度卷积神经网络成功的关键因素之一。在计算机视觉领域,常用图像数据集通常以JPEG格式存储。这种有损压缩技术不可避免地会导致原始数据信息的丢失,进而造成利用压缩数据训练的卷积神经网络的性能降低。因此,为了增强卷积神经网络的性能,本文提出了一种面向压缩图像复原的增强训练方法,通过复原压缩图像实现卷积神经网络的性能增强。该方法具体为一个包含复原模块和任务模块的联合增强框架。复原模块致力于恢复有损压缩技术造成的信息丢失;任务模块专注于基于任务需求增强压缩图像。两个模块联合训练,使得压缩图像的复原增强更具有目的性。本文通过图像分类任务的实验表明,与压缩图像相比,该方法能有效地复原压缩图像,增强卷积神经网络的性能。此外,该方法中两个模块间的低耦合性和可替代性保证了该方法的适用性。 相似文献
|