CN109034249B - Convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernel - Google Patents

Convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernel Download PDF

Info

Publication number
CN109034249B
CN109034249B CN201810852407.5A CN201810852407A CN109034249B CN 109034249 B CN109034249 B CN 109034249B CN 201810852407 A CN201810852407 A CN 201810852407A CN 109034249 B CN109034249 B CN 109034249B
Authority
CN
China
Prior art keywords
convolution
feature map
radially symmetric
image
decomposing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810852407.5A
Other languages
Chinese (zh)
Other versions
CN109034249A (en
Inventor
黄文恺
胡凌恺
薛义豪
倪皓舟
彭广龙
何杰贤
吴羽
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou University
Original Assignee
Guangzhou University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangzhou University filed Critical Guangzhou University
Priority to CN201810852407.5A priority Critical patent/CN109034249B/en
Publication of CN109034249A publication Critical patent/CN109034249A/en
Application granted granted Critical
Publication of CN109034249B publication Critical patent/CN109034249B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/24Classification techniques
    • G06F18/241Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00Arrangements for image or video recognition or understanding
    • G06V10/20Image preprocessing

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

本发明公开了基于分解径向对称卷积核的卷积优化方法、装置、终端设备及计算机可读存储介质,所述方法包括:输入待识别图像,并对所述待识别图像进行预处理;分别利用预先分解m*m径向对称卷积核得到的1个1*1的卷积核和(m‑1)/2个1*m(m=2k+3,k∈N)的卷积核,对经过预处理的待识别图像进行卷积,得到1个1*1的第一特征图和(m‑1)/2个1*m(m=2k+3,k∈N)的第二特征图;进一步对第二特征图进行卷积,得到第三特征图;对第一特征图和(第三特征图进行求和,得到目标特征图,并输出所述目标特征图。本发明通过在降低径向对称卷积核计算量的基础上降低参数量,进而达到对卷积进行优化的目的。

Figure 201810852407

The invention discloses a convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernels. The method includes: inputting an image to be recognized, and preprocessing the image to be recognized; A 1*1 convolution kernel and (m‑1)/2 1*m (m=2k+3, k∈N) convolution kernels obtained by pre-decomposing m*m radially symmetric convolution kernels respectively kernel, convolve the preprocessed image to be recognized, and obtain a 1*1 first feature map and (m‑1)/2 1*m (m=2k+3, k∈N) first feature maps Two feature maps; further convolve the second feature map to obtain a third feature map; sum the first feature map and the (third feature map to obtain a target feature map, and output the target feature map. The present invention By reducing the amount of parameters on the basis of reducing the amount of calculation of the radially symmetric convolution kernel, the purpose of optimizing the convolution is achieved.

Figure 201810852407

Description

基于分解径向对称卷积核的卷积优化方法、装置、终端设备及 计算机可读存储介质Convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernel

技术领域technical field

本发明涉及神经网络技术领域,尤其涉及基于分解径向对称卷积核的卷积优化方法、装置、终端设备及计算机可读存储介质。The present invention relates to the technical field of neural networks, and in particular, to a convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernels.

背景技术Background technique

卷积神经网络(CNN)是近几年图像处理与识别领域最常用的一种神经网络,它具有较好的特征分类效果和易于高维数据处理的优点,但是卷积神经网络容易出现过拟合现象,且卷积神经网络的鲁棒性较低,因此,在传统的卷积神经网络构建中,常常会将原始训练集图片进行镜像以及大角度的旋转处理,以增加卷积神经网络的鲁棒性,使得卷积神经网络可识别任意角度的图片;但是这种传统方法却会产生增大数据量,使得训练时间增长的问题。Convolutional Neural Network (CNN) is the most commonly used neural network in the field of image processing and recognition in recent years. It has the advantages of good feature classification effect and easy processing of high-dimensional data, but convolutional neural network is prone to overfitting. Therefore, in the traditional construction of convolutional neural networks, the original training set images are often mirrored and rotated at a large angle to increase the performance of the convolutional neural network. Robustness, so that the convolutional neural network can recognize pictures from any angle; but this traditional method will increase the amount of data and increase the training time.

现有技术中,针对所述传统方法存在的问题,通常采用具有径向对称性质的卷积核,其在卷积神经网络的使用中能够提供良好的鲁棒性,且能够降低出现过拟合现象的可能性。然而所述径向对称卷积核在运算中计算量过大,现有的针对该计算量过大的问题的技术方案主要是,在卷积神经网络中采用卷积核裁剪法和多通道卷积优化算法。但是,发明人在实施现有技术方案时,发现这些优化算法在使用中会出现参数量过大的问题。In the prior art, in view of the problems existing in the traditional methods, convolution kernels with radial symmetry properties are usually used, which can provide good robustness in the use of convolutional neural networks, and can reduce the occurrence of overfitting. the possibility of the phenomenon. However, the radially symmetric convolution kernel requires too much computation in the operation. The existing technical solutions for the problem of the too large amount of computation are mainly to use the convolution kernel cropping method and the multi-channel volume in the convolutional neural network. Product optimization algorithm. However, when the inventor implements the prior art solutions, it is found that these optimization algorithms may have a problem of excessively large amount of parameters in use.

发明内容SUMMARY OF THE INVENTION

本发明所要解决的技术问题在于,提供基于分解径向对称卷积核的卷积优化方法、装置、终端设备及计算机可读存储介质,通过在降低径向对称卷积核计算量的基础上降低参数量,进而达到对卷积进行优化的目的。The technical problem to be solved by the present invention is to provide a convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernels. The amount of parameters, so as to achieve the purpose of optimizing the convolution.

为了解决上述技术问题,本发明的一个实施例提供了基于分解径向对称卷积核的卷积优化方法,适于在计算设备中执行,包括如下步骤:In order to solve the above technical problems, an embodiment of the present invention provides a convolution optimization method based on decomposing radially symmetric convolution kernels, suitable for execution in a computing device, including the following steps:

输入待识别图像,并对所述待识别图像进行预处理;inputting an image to be recognized, and preprocessing the image to be recognized;

分别利用预先分解m*m径向对称卷积核得到的1个1*1的卷积核和(m-1)/2 个1*m(m=2k+3,k∈N)的卷积核,对经过预处理的待识别图像进行卷积,得到1 个1*1的第一特征图和(m-1)/2个1*m(m=2k+3,k∈N)的第二特征图;再利用预先分解m*m径向对称卷积核得到的与(m-1)/2个1*m(m=2k+3,k∈N)的卷积核一一对应的(m-1)/2个m*1(m=2k+3,k∈N)的卷积核,对(m-1)/2个1*m(m=2k+3,k∈N) 的第二特征图进行卷积,得到(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图;Use a 1*1 convolution kernel and (m-1)/2 1*m (m=2k+3, k∈N) convolution kernels obtained by pre-decomposing m*m radially symmetric convolution kernels respectively. kernel, convolve the preprocessed image to be recognized, and obtain a 1*1 first feature map and (m-1)/2 1*m (m=2k+3, k∈N) first feature maps Two feature maps; then use the pre-decomposition m*m radially symmetric convolution kernel to obtain (m-1)/2 convolution kernels of 1*m (m=2k+3, k∈N) one-to-one correspondence (m-1)/2 convolution kernels of m*1 (m=2k+3, k∈N), for (m-1)/2 1*m (m=2k+3, k∈N) The second feature map is convolved to obtain (m-1)/2 m*1 (m=2k+3, k∈N) third feature maps;

对1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图进行求和,得到目标特征图,并输出所述目标特征图。Sum up one 1*1 first feature map and (m-1)/2 m*1 (m=2k+3, k∈N) third feature maps to obtain the target feature map, and output the target feature map.

进一步地,所述对所述待识别图像进行预处理,具体为:Further, the preprocessing of the to-be-recognized image is specifically:

根据预设参数,对所述待识图像进行随机拉伸和明暗调整,并加入特定的高斯噪声;According to preset parameters, randomly stretch and adjust the brightness and darkness of the image to be recognized, and add specific Gaussian noise;

进一步地,根据卷积处理的要求,对所述待识图像进行0~π/2角度的旋转和切割。Further, according to the requirements of the convolution processing, the to-be-recognized image is rotated and cut at an angle of 0-π/2.

进一步地,每一个卷积核的卷积核矩阵A满足如下公式:Further, the convolution kernel matrix A of each convolution kernel satisfies the following formula:

Figure BDA0001746392170000021
Figure BDA0001746392170000021

其中,

Figure BDA0001746392170000022
in,
Figure BDA0001746392170000022

进一步地,每一个m*1(m=2k+3,k∈N)的卷积核与对应的一个m*1(m=2k+3,k ∈N)的卷积核组成一个等比对称向量组,即ISV,具体的:Further, each m*1 (m=2k+3, k∈N) convolution kernel and a corresponding m*1 (m=2k+3, k∈N) convolution kernel form a proportional symmetry Vector group, or ISV, specifically:

设an为参数值,a1恒等于1,ISV的长度为m(m=2k+3,k∈N),Let a n be the parameter value, a 1 is always equal to 1, the length of the ISV is m (m=2k+3, k∈N),

ISV=(ISV_1,ISV_2);ISV=(ISV_1,ISV_2);

其中,ISV_1为一个1*m向量,ISV_2为一个m*1向量;Among them, ISV_1 is a 1*m vector, and ISV_2 is an m*1 vector;

Figure BDA0001746392170000023
Figure BDA0001746392170000023

Figure BDA0001746392170000024
Figure BDA0001746392170000024

进一步地,在所述对1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N) 的第三特征图进行求和,得到目标特征图,并输出所述目标特征图之后,还包括:Further, after summing the first feature map of 1*1 and the third feature map of (m-1)/2 m*1(m=2k+3,k∈N), we get target feature map, and after outputting the target feature map, it also includes:

判断所述图像的方向对识别结果是否有影响;Determine whether the direction of the image has an influence on the recognition result;

若是,则将1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图进行全局平均池化,并将输出的1+(m-1)/2个值进行softmax处理,得到目标特征图的概率。If so, perform global average pooling on a first feature map of 1*1 and a third feature map of (m-1)/2 m*1 (m=2k+3, k∈N). The output 1+(m-1)/2 values are processed by softmax to obtain the probability of the target feature map.

进一步地,所述的基于分解径向对称卷积核的卷积优化方法,还包括:Further, the described convolution optimization method based on decomposing radially symmetric convolution kernels also includes:

若否,则将1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图进行空间金字塔池化和全连接层处理,并将输出的1+(m-1)/2个值进行 softmax处理,得到目标特征图的判断结果。If not, perform spatial pyramid pooling and full-scale spatial pyramid pooling on a first feature map of 1*1 and a third feature map of (m-1)/2 m*1 (m=2k+3, k∈N). The connection layer is processed, and the output 1+(m-1)/2 values are subjected to softmax processing to obtain the judgment result of the target feature map.

本发明的一个实施例还提供了一种基于分解径向对称卷积核的卷积优化装置,包括:An embodiment of the present invention also provides a convolution optimization device based on decomposing radially symmetric convolution kernels, including:

输入模块,用于输入待识别图像,并对所述待识别图像进行预处理;an input module for inputting an image to be recognized and preprocessing the image to be recognized;

卷积优化模块,用于分别利用预先分解m*m径向对称卷积核得到的1个1*1 的卷积核和(m-1)/2个1*m(m=2k+3,k∈N)的卷积核,对经过预处理的待识别图像进行卷积,得到1个1*1的第一特征图和(m-1)/2个1*m(m=2k+3,k∈N)的第二特征图;再利用预先分解m*m径向对称卷积核得到的与(m-1)/2个1*m(m=2k+3,k ∈N)的卷积核一一对应的(m-1)/2个m*1(m=2k+3,k∈N)的卷积核,对(m-1)/2个 1*m(m=2k+3,k∈N)的第二特征图进行卷积,得到(m-1)/2个m*1(m=2k+3,k∈N) 的第三特征图;The convolution optimization module is used to use a 1*1 convolution kernel and (m-1)/2 1*m(m=2k+3, k∈N) convolution kernel, convolve the preprocessed image to be recognized, and obtain a 1*1 first feature map and (m-1)/2 1*m(m=2k+3 ,k∈N) of the second feature map; and then use the pre-decomposition m*m radially symmetric convolution kernel and (m-1)/2 1*m(m=2k+3,k ∈N) Convolution kernels correspond to (m-1)/2 m*1 (m=2k+3, k∈N) convolution kernels, for (m-1)/2 1*m (m=2k The second feature map of +3,k∈N) is convolved to obtain the third feature map of (m-1)/2 m*1(m=2k+3,k∈N);

输出模块,用于对1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图进行求和,得到目标特征图,并输出所述目标特征图。The output module is used to sum the first feature map of 1*1 and the third feature map of (m-1)/2 m*1 (m=2k+3, k∈N) to obtain the target feature map, and output the target feature map.

进一步地,每一个m*1(m=2k+3,k∈N)的卷积核与对应的一个m*1(m=2k+3,k ∈N)的卷积核组成一个等比对称向量组,即ISV,具体的:Further, each m*1 (m=2k+3, k∈N) convolution kernel and a corresponding m*1 (m=2k+3, k∈N) convolution kernel form a proportional symmetry Vector group, or ISV, specifically:

设an为参数值,a1恒等于1,ISV的长度为m(m=2k+3,k∈N),Let a n be the parameter value, a 1 is always equal to 1, the length of the ISV is m (m=2k+3, k∈N),

ISV=(ISV_1,ISV_2);ISV=(ISV_1,ISV_2);

其中,ISV_1为一个1*m向量,ISV_2为一个m*1向量;Among them, ISV_1 is a 1*m vector, and ISV_2 is an m*1 vector;

Figure BDA0001746392170000031
Figure BDA0001746392170000031

Figure BDA0001746392170000032
Figure BDA0001746392170000032

本发明的一个实施例还提供了一种基于分解径向对称卷积核的卷积优化终端设备,包括:处理器、存储器以及存储在所述存储器中且被配置为由所述处理器执行的计算机程序,所述处理器执行所述计算机程序时实现如上述的基于分解径向对称卷积核的卷积优化方法。An embodiment of the present invention also provides a convolution optimization terminal device based on decomposing radially symmetric convolution kernels, comprising: a processor, a memory, and a function stored in the memory and configured to be executed by the processor A computer program, when the processor executes the computer program, the above-mentioned convolution optimization method based on decomposing radially symmetric convolution kernels is implemented.

本发明的一个实施例还提供了一种计算机可读存储介质,所述计算机可读存储介质包括存储的计算机程序,其中,在所述计算机程序运行时控制所述计算机可读存储介质所在设备执行如上述的基于分解径向对称卷积核的卷积优化方法。An embodiment of the present invention further provides a computer-readable storage medium, where the computer-readable storage medium includes a stored computer program, wherein when the computer program runs, the device where the computer-readable storage medium is located is controlled to execute The above-mentioned convolution optimization method based on decomposing radially symmetric convolution kernels.

相比于现有技术,本发明的一个实施例的有益效果在于:Compared with the prior art, the beneficial effects of an embodiment of the present invention are:

本发明提供的基于分解径向对称卷积核的卷积优化方法、装置、终端设备及计算机可读存储介质,所述方法包括:输入待识别图像,并对所述待识别图像进行预处理;分别利用预先分解m*m径向对称卷积核得到的1个1*1的卷积核和 (m-1)/2个1*m(m=2k+3,k∈N)的卷积核,对经过预处理的待识别图像进行卷积,得到1个1*1的第一特征图和(m-1)/2个1*m(m=2k+3,k∈N)的第二特征图;再利用与(m-1)/2个1*m(m=2k+3,k∈N)的卷积核一一对应的(m-1)/2个m*1(m=2k+3,k ∈N)的卷积核,对(m-1)/2个1*m(m=2k+3,k∈N)的第二特征图进行卷积,得到 (m-1)/2个m*1(m=2k+3,k∈N)的第三特征图;对1个1*1的第一特征图和(m-1)/2 个m*1(m=2k+3,k∈N)的第三特征图进行求和,得到目标特征图,并输出所述目标特征图。本发明通过在降低径向对称卷积核计算量的基础上降低参数量,进而达到对卷积进行优化的目的。The convolution optimization method, device, terminal device, and computer-readable storage medium based on decomposing radially symmetric convolution kernels provided by the present invention include: inputting an image to be recognized, and preprocessing the image to be recognized; Use a 1*1 convolution kernel and (m-1)/2 1*m (m=2k+3, k∈N) convolution kernels obtained by pre-decomposing m*m radially symmetric convolution kernels respectively. kernel, convolve the preprocessed image to be recognized, and obtain a 1*1 first feature map and (m-1)/2 1*m (m=2k+3, k∈N) first feature maps Two feature maps; re-use (m-1)/2 m*1(m =2k+3,k∈N) convolution kernel, convolve the second feature map of (m-1)/2 1*m(m=2k+3,k∈N), get (m- 1)/2 m*1 (m=2k+3, k∈N) third feature maps; for 1 1*1 first feature map and (m-1)/2 m*1(m =2k+3,k∈N) and the third feature maps are summed to obtain the target feature map, and output the target feature map. The invention achieves the purpose of optimizing the convolution by reducing the amount of parameters on the basis of reducing the amount of calculation of the radially symmetric convolution kernel.

附图说明Description of drawings

图1是本发明第一实施例提供的基于分解径向对称卷积核的卷积优化方法的流程示意图;1 is a schematic flowchart of a convolution optimization method based on decomposing radially symmetric convolution kernels provided by the first embodiment of the present invention;

图2是本发明第一实施例中经过分解的3*3径向对称卷积核对图像的处理过程的流程示意图;2 is a schematic flowchart of a process of decomposing a 3*3 radially symmetric convolution check image in the first embodiment of the present invention;

图3本发明第一实施例中经过分解的5*5径向对称卷积核对图像的处理过程的流程示意图;3 is a schematic flow chart of the processing process of the decomposed 5*5 radially symmetric convolution check image in the first embodiment of the present invention;

图4本发明第一实施例中经过分解的m*m径向对称卷积核对图像的处理过程的流程示意图;4 is a schematic flow chart of a process of processing a decomposed m*m radially symmetric convolution check image in the first embodiment of the present invention;

图5是本发明第一实施例中包含分解对称卷积核的神经网络对猫狗图像分类的处理过程的流程示意图;5 is a schematic flowchart of a process for classifying cat and dog images by a neural network comprising a decomposed symmetric convolution kernel in the first embodiment of the present invention;

图6本发明第一实施例中包含分解对称卷积核的神经网络对MNIST图像分类的处理过程的流程示意图;6 is a schematic flowchart of a process for classifying MNIST images by a neural network including a decomposed symmetric convolution kernel in the first embodiment of the present invention;

图7是本发明第二实施例提供的基于分解径向对称卷积核的卷积优化装置的结构示意图。FIG. 7 is a schematic structural diagram of a convolution optimization apparatus based on decomposing radially symmetric convolution kernels provided by a second embodiment of the present invention.

具体实施方式Detailed ways

下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清楚、完整地描述,以使本发明的优点和特征能更易于被本领域技术人员理解,显然,所描述的实施例仅仅是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本发明保护的范围。The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, so that the advantages and features of the present invention can be more easily understood by those skilled in the art. Obviously, the described The embodiments are only some of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

本发明第一实施例:The first embodiment of the present invention:

请参阅图1-5。See Figures 1-5.

如图1所示,本实施例提供的一种基于分解径向对称卷积核的卷积优化方法,基于分解径向对称卷积核的卷积优化方法,适于在计算设备中执行,包括如下步骤:As shown in FIG. 1 , a convolution optimization method based on decomposing a radially symmetric convolution kernel provided in this embodiment, and a convolution optimization method based on decomposing a radially symmetric convolution kernel, are suitable for execution in a computing device, including: Follow the steps below:

S101、输入待识别图像,并对所述待识别图像进行预处理;S101. Input an image to be recognized, and preprocess the image to be recognized;

S102、分别利用预先分解m*m径向对称卷积核得到的1个1*1的卷积核和 (m-1)/2个1*m(m=2k+3,k∈N)的卷积核,对经过预处理的待识别图像进行卷积,得到1个1*1的第一特征图和(m-1)/2个1*m(m=2k+3,k∈N)的第二特征图;再利用预先分解m*m径向对称卷积核得到的与(m-1)/2个1*m(m=2k+3,k∈N)的卷积核一一对应的(m-1)/2个m*1(m=2k+3,k∈N)的卷积核,对(m-1)/2个1*m(m=2k+3,k ∈N)的第二特征图进行卷积,得到(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图;S102, respectively using a 1*1 convolution kernel and (m-1)/2 1*m (m=2k+3, k∈N) convolution kernels obtained by decomposing m*m radially symmetric convolution kernels in advance The convolution kernel, convolves the preprocessed image to be recognized, and obtains a 1*1 first feature map and (m-1)/2 1*m (m=2k+3, k∈N) The second feature map of Corresponding (m-1)/2 m*1 (m=2k+3, k∈N) convolution kernels, for (m-1)/2 1*m (m=2k+3, k∈N) The second feature map of N) is convolved to obtain the third feature map of (m-1)/2 m*1 (m=2k+3, k∈N);

S103、对1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图进行求和,得到目标特征图,并输出所述目标特征图。S103, summing up a first feature map of 1*1 and a third feature map of (m-1)/2 m*1 (m=2k+3, k∈N) to obtain a target feature map, And output the target feature map.

本实施例提出的基于分解径向对称卷积核的卷积优化方法,增强了鲁棒性,减少了参数的数量,加快了神经网络拟合速度并防止过拟合。The convolution optimization method based on decomposing radially symmetric convolution kernels proposed in this embodiment enhances robustness, reduces the number of parameters, speeds up neural network fitting and prevents overfitting.

对于步骤S101,所述对所述待识别图像进行预处理,具体为:For step S101, the preprocessing of the to-be-identified image is specifically:

根据预设参数,对所述待识图像进行随机拉伸和明暗调整,并加入特定的高斯噪声;According to preset parameters, randomly stretch and adjust the brightness and darkness of the image to be recognized, and add specific Gaussian noise;

进一步地,根据卷积处理的要求,对所述待识图像进行0~π/2角度的旋转和切割。Further, according to the requirements of the convolution processing, the to-be-recognized image is rotated and cut at an angle of 0-π/2.

在本实施例中,可以理解的是,为了增加鲁棒性,将要输入的待识别图像加上轻微的高斯噪声,进行小幅度的随机拉伸与明暗变化。由于径向对称卷积核的先天优势,这里待识别图片不需要进行镜像变换,同时预处理中的旋转步骤只需要旋转0~π/2的角度。由于该网络只应用了卷积神经网络,没有全连接层,对尺寸没有要求,除了对旋转的图片做必要的切割外,不再需要对图片尺寸做任何处理。In this embodiment, it can be understood that, in order to increase the robustness, a slight Gaussian noise is added to the to-be-recognized image to be input, and small-amplitude random stretching and light and dark changes are performed. Due to the inherent advantages of the radially symmetric convolution kernel, the image to be recognized does not need to be mirrored, and the rotation step in the preprocessing only needs to rotate by an angle of 0 to π/2. Since the network only applies the convolutional neural network, there is no fully connected layer, and there is no requirement for the size. Except for the necessary cutting of the rotated image, there is no need to do any processing on the image size.

对于步骤S102,其中,每一个卷积核的卷积核矩阵A满足如下公式:For step S102, the convolution kernel matrix A of each convolution kernel satisfies the following formula:

Figure BDA0001746392170000051
Figure BDA0001746392170000051

其中,

Figure BDA0001746392170000052
in,
Figure BDA0001746392170000052

在本实施例中,如以下两个矩阵:In this embodiment, such as the following two matrices:

Figure BDA0001746392170000053
Figure BDA0001746392170000053

Figure BDA0001746392170000061
Figure BDA0001746392170000061

考虑到图片在旋转非

Figure BDA0001746392170000062
倍数角度时像素数值会发生变化,这里仅考虑旋转角度
Figure BDA0001746392170000063
A卷积处理经过θ角度旋转的图片结果与直接处理原图得到的结果相等,即处理镜像图片也会与原图产生相同的结果。Taking into account that the image is rotating non-
Figure BDA0001746392170000062
The pixel value will change when the angle is multiple, only the rotation angle is considered here
Figure BDA0001746392170000063
The result of A convolution processing the image rotated by θ angle is equal to the result obtained by directly processing the original image, that is, processing the mirror image will produce the same result as the original image.

由此可以推知,在卷积核初始化时将其设置为满足A的矩阵即可达以到上述效果。It can be inferred from this that the above effect can be achieved by setting the convolution kernel to a matrix that satisfies A when it is initialized.

在本实施例中,该卷积核除了具有良好的鲁棒性外,还减少了庞大的参数量,该核的参数量为传统卷积核的

Figure BDA0001746392170000064
倍,
Figure BDA0001746392170000065
即此方法最高可以减少
Figure BDA0001746392170000066
的参数量。有效加快模型的收敛速度并防止过拟合。In this embodiment, the convolution kernel not only has good robustness, but also reduces the huge amount of parameters, and the parameter amount of the kernel is the same as that of the traditional convolution kernel.
Figure BDA0001746392170000064
times,
Figure BDA0001746392170000065
That is, this method can reduce the maximum
Figure BDA0001746392170000066
parameter amount. Effectively speed up the convergence of the model and prevent overfitting.

此方法在高维度的卷积网络中效果会更加明显,在x维的卷积神经网络中,该核最多能减少

Figure BDA0001746392170000067
的参数,优化效果极为显著。This method is more effective in high-dimensional convolutional networks. In x-dimensional convolutional neural networks, the kernel can reduce at most
Figure BDA0001746392170000067
parameters, the optimization effect is extremely significant.

在本实施例中,进一步地,每一个m*1(m=2k+3,k∈N)的卷积核与对应的一个m*1(m=2k+3,k∈N)的卷积核组成一个等比对称向量组,即ISV,具体的:In this embodiment, further, each m*1 (m=2k+3, k∈N) convolution kernel is convolutional with a corresponding m*1 (m=2k+3, k∈N) convolution kernel The kernel forms a proportional symmetric vector group, namely ISV, specifically:

设an为参数值,a1恒等于1,ISV的长度为m(m=2k+3,k∈N),Let a n be the parameter value, a 1 is always equal to 1, the length of the ISV is m (m=2k+3, k∈N),

ISV=(ISV_1,ISV_2);ISV=(ISV_1,ISV_2);

其中,ISV_1为一个1*m向量,ISV_2为一个m*1向量;Among them, ISV_1 is a 1*m vector, and ISV_2 is an m*1 vector;

Figure BDA0001746392170000068
Figure BDA0001746392170000068

Figure BDA0001746392170000069
Figure BDA0001746392170000069

在本实施例中,由于在常规的卷积操作中涉及了大量的重复运算,会消耗很多不必要的运算时间,所以,本实施例将长、宽为3的径向对称卷积核分解成了一对长度为3的ISV向量和1*1的卷积核组合,卷积核具体变量定义如下:In this embodiment, since a large number of repeated operations are involved in the conventional convolution operation, it will consume a lot of unnecessary operation time. Therefore, this embodiment decomposes the radially symmetric convolution kernel whose length and width are 3 into A pair of ISV vectors of length 3 and a 1*1 convolution kernel are combined. The specific variables of the convolution kernel are defined as follows:

ISV_1卷积核:

Figure BDA00017463921700000610
ISV_1 convolution kernel:
Figure BDA00017463921700000610

ISV_2卷积核:

Figure BDA00017463921700000611
ISV_2 convolution kernel:
Figure BDA00017463921700000611

1*1卷积核:

Figure BDA00017463921700000612
1*1 convolution kernel:
Figure BDA00017463921700000612

以上的卷积核中的an,b均为参数值。an and b in the above convolution kernels are both parameter values.

如图2所示,具体卷积过程如下:As shown in Figure 2, the specific convolution process is as follows:

假设原图为P,先将ISV_1卷积核和1*1卷积核分别对P进行卷积,得到特征图 P1、P2,接下来使用ISV_2卷积核对P1进行卷积,得到特征图P3,然后将P2 与P3相加得到输出的特征图P_output。Assuming that the original image is P, first convolve P with the ISV_1 convolution kernel and the 1*1 convolution kernel respectively to obtain the feature maps P1 and P2, and then use the ISV_2 convolution kernel to convolve P1 to obtain the feature map P3, Then add P2 and P3 to get the output feature map P_output.

经数学运算,上述卷积过程中的卷积过程等价于使用卷积核形为After mathematical operation, the convolution process in the above convolution process is equivalent to using the convolution kernel shape as

Figure BDA0001746392170000071
Figure BDA0001746392170000071

的径向对称卷积核进行卷积,而且,在上述卷积过程中对同一像素点只使用了7次乘法运算5次加法运算,传统的卷积操作要进行9次乘法和8次加法运算,减少了很多运算时间。The radially symmetric convolution kernel is used for convolution, and in the above convolution process, only 7 multiplications and 5 additions are used for the same pixel, while the traditional convolution operation requires 9 multiplications and 8 additions. , which reduces a lot of computation time.

在本实施例中,如图3所示,优选的,如要应用5*5的径向对称卷积核,可将其分解成两对长度分别为5,3的ISV向量和1*1的卷积核组合(每一对ISV的an都不相同)。In this embodiment, as shown in FIG. 3 , preferably, if a radially symmetric convolution kernel of 5*5 is to be applied, it can be decomposed into two pairs of ISV vectors with lengths of 5 and 3 and a 1*1 Combination of convolution kernels (an is different for each pair of ISVs).

进一步地,如图4所示,推广开来,如要应用m*m的径向对称卷积核(m=2k+3,k ∈N),可将其分解成(m-1)/2对长度分别为3,5,7,…,m的ISV向量和1*1的卷积核组合。Further, as shown in Figure 4, by extension, if a radially symmetric convolution kernel of m*m (m=2k+3, k ∈ N) is to be applied, it can be decomposed into (m-1)/2 Combine ISV vectors with lengths of 3, 5, 7, ..., m and 1*1 convolution kernels.

在本实施例中,根据上述描述,这种运算方式在卷积核尺寸较大时可以极大地减少运算量,当尺寸为m*m时,乘法的运算量为原来的

Figure BDA0001746392170000073
倍,加法的运算量为原来的
Figure BDA0001746392170000072
倍。In this embodiment, according to the above description, this operation method can greatly reduce the operation amount when the size of the convolution kernel is large. When the size is m*m, the operation amount of the multiplication is the same as the original.
Figure BDA0001746392170000073
times, the amount of operations for addition is the same as the original
Figure BDA0001746392170000072
times.

对于步骤S103、进一步地,在所述对1个1*1的第一特征图和(m-1)/2个 m*1(m=2k+3,k∈N)的第三特征图进行求和,得到目标特征图,并输出所述目标特征图之后,还包括:For step S103, further, the first feature map of 1*1 and the third feature map of (m-1)/2 m*1 (m=2k+3, k∈N) are performed. Summing to obtain the target feature map, and after outputting the target feature map, it also includes:

判断所述图像的方向对识别结果是否有影响;Determine whether the direction of the image has an influence on the recognition result;

若是,则将1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图进行全局平均池化,并将输出的1+(m-1)/2个值进行softmax处理,得到目标特征图的概率。If so, perform global average pooling on a first feature map of 1*1 and a third feature map of (m-1)/2 m*1 (m=2k+3, k∈N). The output 1+(m-1)/2 values are processed by softmax to obtain the probability of the target feature map.

在本实施例中,该卷积核拆分算法在高维卷积核中对运算量的减少会更明显。如图5所示,使用经分解的径向对称卷积网络对猫、狗进行分类识别。In this embodiment, the convolution kernel splitting algorithm reduces the computation amount more obviously in the high-dimensional convolution kernel. As shown in Figure 5, cats and dogs are classified and recognized using the decomposed radially symmetric convolutional network.

在神经网络的搭建中,在网络的前端使用了8层的3*3分解的径向对称卷积网络,并在其中连接残差层,防止梯度消失。In the construction of the neural network, an 8-layer 3*3 decomposed radially symmetric convolutional network is used at the front end of the network, and the residual layer is connected in it to prevent the gradient from disappearing.

由于图像的方向对识别结果没有影响,在神经网络的末端采用了全局平均池化的方法将最后的两张特征图平均池化,将输出的两个值进行softmax处理得到是猫和狗的概率Since the direction of the image has no effect on the recognition results, the global average pooling method is used at the end of the neural network to average pool the last two feature maps, and the two output values are processed by softmax to obtain the probability of cats and dogs.

进一步地,所述的基于分解径向对称卷积核的卷积优化方法,还包括:Further, the described convolution optimization method based on decomposing radially symmetric convolution kernels also includes:

若否,则将1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图进行空间金字塔池化和全连接层处理,并将输出的1+(m-1)/2个值进行 softmax处理,得到目标特征图的判断结果。If not, perform spatial pyramid pooling and full-scale spatial pyramid pooling on a first feature map of 1*1 and a third feature map of (m-1)/2 m*1 (m=2k+3, k∈N). The connection layer is processed, and the output 1+(m-1)/2 values are subjected to softmax processing to obtain the judgment result of the target feature map.

在本实施例中,如图6所示,使用经分解的径向对称卷积网络识别MNIST 数据集:In this example, as shown in Figure 6, a decomposed radially symmetric convolutional network is used to identify the MNIST dataset:

在图像处理阶段,仍采用上述的预处理方法,只是预处理中不包含旋转。神经网络的主体也采用在网络的前端使用了8层的3*3分解的径向对称卷积网络,并在其中连接残差层,防止梯度消失的方法。但由于数据的减少,相应的减少了神经网络的层数。In the image processing stage, the above-mentioned preprocessing method is still used, but rotation is not included in the preprocessing. The main body of the neural network also uses a radially symmetric convolutional network with 8 layers of 3*3 decomposition at the front end of the network, and connects the residual layer in it to prevent the gradient from disappearing. However, due to the reduction of data, the number of layers of the neural network is correspondingly reduced.

由于图像的方向对识别结果有影响,如识别6,9会因为方向不确定而识别错误,在神经网络的末端采用空间金字塔池化和全连接层搭配softmax输出判断为各数字的结果。Since the direction of the image has an impact on the recognition result, for example, the recognition of 6 and 9 will be wrong due to the uncertainty of the direction. At the end of the neural network, spatial pyramid pooling and fully connected layer with softmax output are used to judge the result of each number.

需要说明的是,这只是本实施例中的两个简单的例子,该网络可以用在更加庞大复杂的识别系统中并取得良好的效果。It should be noted that these are just two simple examples in this embodiment, and the network can be used in a larger and more complex identification system and achieve good results.

本实施例提供的一种基于分解径向对称卷积核的卷积优化方法,包括:输入待识别图像,并对所述待识别图像进行预处理;分别利用预先分解m*m径向对称卷积核得到的1个1*1的卷积核和(m-1)/2个1*m(m=2k+3,k∈N)的卷积核,对经过预处理的待识别图像进行卷积,得到1个1*1的第一特征图和(m-1)/2个1*m(m=2k+3,k∈N)的第二特征图;再利用与(m-1)/2个1*m(m=2k+3,k∈N)的卷积核一一对应的(m-1)/2个m*1(m=2k+3,k∈N)的卷积核,对(m-1)/2个1*m(m=2k+3,k ∈N)的第二特征图进行卷积,得到(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图;对1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图进行求和,得到目标特征图,并输出所述目标特征图。本实施例通过在降低径向对称卷积核计算量的基础上降低参数量,进而达到对卷积进行优化的目的。A convolution optimization method based on decomposing radially symmetric convolution kernels provided in this embodiment includes: inputting an image to be recognized, and preprocessing the to-be-recognized image; using pre-decomposing m*m radially symmetric convolutions respectively 1 convolution kernel of 1*1 and (m-1)/2 convolution kernels of 1*m (m=2k+3, k∈N) obtained by the accumulation kernel, perform preprocessing on the image to be recognized. Convolve to obtain a first feature map of 1*1 and (m-1)/2 second feature maps of 1*m (m=2k+3, k∈N); )/2 convolution kernels of 1*m (m=2k+3, k∈N) one-to-one correspondence (m-1)/2 volumes of m*1 (m=2k+3, k∈N) Product kernel, convolve the second feature map of (m-1)/2 1*m (m=2k+3, k ∈ N) to obtain (m-1)/2 m*1 (m= 2k+3, k∈N) of the third feature map; for a 1*1 first feature map and (m-1)/2 m*1 (m=2k+3, k∈N) the first feature map The three feature maps are summed to obtain the target feature map, and the target feature map is output. This embodiment achieves the purpose of optimizing the convolution by reducing the amount of parameters on the basis of reducing the amount of calculation of the radially symmetric convolution kernel.

本发明第二实施例:The second embodiment of the present invention:

请参阅图7。See Figure 7.

如图7所示,本实施例还提供了一种基于分解径向对称卷积核的卷积优化装置,包括:As shown in FIG. 7 , this embodiment also provides a convolution optimization device based on decomposing radially symmetric convolution kernels, including:

输入模块201,用于输入待识别图像,并对所述待识别图像进行预处理。The input module 201 is used for inputting an image to be recognized, and preprocessing the image to be recognized.

所述对所述待识别图像进行预处理,具体为:The preprocessing of the to-be-recognized image is specifically:

根据预设参数,对所述待识图像进行随机拉伸和明暗调整,并加入特定的高斯噪声;According to preset parameters, randomly stretch and adjust the brightness and darkness of the image to be recognized, and add specific Gaussian noise;

进一步地,根据卷积处理的要求,对所述待识图像进行0~π/2角度的旋转和切割。Further, according to the requirements of the convolution processing, the to-be-recognized image is rotated and cut at an angle of 0-π/2.

在本实施例中,可以理解的是,为了增加鲁棒性,将要输入的待识别图像加上轻微的高斯噪声,进行小幅度的随机拉伸与明暗变化。由于径向对称卷积核的先天优势,这里待识别图片不需要进行镜像变换,同时预处理中的旋转步骤只需要旋转0~π/2的角度。由于该网络只应用了卷积神经网络,没有全连接层,对尺寸没有要求,除了对旋转的图片做必要的切割外,不再需要对图片尺寸做任何处理。In this embodiment, it can be understood that, in order to increase the robustness, a slight Gaussian noise is added to the to-be-recognized image to be input, and small-amplitude random stretching and light and dark changes are performed. Due to the inherent advantages of the radially symmetric convolution kernel, the image to be recognized does not need to be mirrored, and the rotation step in the preprocessing only needs to rotate by an angle of 0 to π/2. Since the network only applies the convolutional neural network, there is no fully connected layer, and there is no requirement for the size. Except for the necessary cutting of the rotated image, there is no need to do any processing on the image size.

卷积优化模块202,用于分别利用预先分解m*m径向对称卷积核得到的1 个1*1的卷积核和(m-1)/2个1*m(m=2k+3,k∈N)的卷积核,对经过预处理的待识别图像进行卷积,得到1个1*1的第一特征图和(m-1)/2个1*m(m=2k+3,k∈N)的第二特征图;再利用预先分解m*m径向对称卷积核得到的与(m-1)/2个 1*m(m=2k+3,k∈N)的卷积核一一对应的(m-1)/2个m*1(m=2k+3,k∈N)的卷积核,对(m-1)/2个1*m(m=2k+3,k∈N)的第二特征图进行卷积,得到(m-1)/2个 m*1(m=2k+3,k∈N)的第三特征图。The convolution optimization module 202 is used for respectively using a 1*1 convolution kernel and (m-1)/2 1*m (m=2k+3) obtained by decomposing m*m radially symmetric convolution kernels in advance ,k∈N) convolution kernel, convolve the preprocessed image to be recognized, and obtain a 1*1 first feature map and (m-1)/2 1*m(m=2k+ 3, k∈N) of the second feature map; then use the pre-decomposition m*m radially symmetric convolution kernel and (m-1)/2 1*m (m=2k+3, k∈N) The convolution kernels correspond to (m-1)/2 m*1 (m=2k+3, k∈N) convolution kernels, for (m-1)/2 1*m(m= The second feature map of 2k+3, k∈N) is convolved to obtain the third feature map of (m-1)/2 m*1 (m=2k+3, k∈N).

其中,每一个卷积核的卷积核矩阵A满足如下公式:Among them, the convolution kernel matrix A of each convolution kernel satisfies the following formula:

Figure BDA0001746392170000091
Figure BDA0001746392170000091

其中,

Figure BDA0001746392170000092
in,
Figure BDA0001746392170000092

在本实施例中,如以下两个矩阵:In this embodiment, such as the following two matrices:

Figure BDA0001746392170000093
Figure BDA0001746392170000093

Figure BDA0001746392170000094
Figure BDA0001746392170000094

考虑到图片在旋转非

Figure BDA0001746392170000095
倍数角度时像素数值会发生变化,这里仅考虑旋转角度
Figure BDA0001746392170000096
A卷积处理经过θ角度旋转的图片结果与直接处理原图得到的结果相等,即处理镜像图片也会与原图产生相同的结果。Take into account that the image is rotating non-
Figure BDA0001746392170000095
The pixel value will change when the angle is multiple, only the rotation angle is considered here
Figure BDA0001746392170000096
The result of A convolution processing the image rotated by θ angle is equal to the result obtained by directly processing the original image, that is, processing the mirror image will produce the same result as the original image.

由此可以推知,在卷积核初始化时将其设置为满足A的矩阵即可达以到上述效果。It can be inferred from this that the above effect can be achieved by setting the convolution kernel to a matrix that satisfies A when it is initialized.

在本实施例中,该卷积核除了具有良好的鲁棒性外,还减少了庞大的参数量,该核的参数量为传统卷积核的

Figure BDA0001746392170000097
倍,
Figure BDA0001746392170000098
即此方法最高可以减少
Figure BDA0001746392170000099
的参数量。有效加快模型的收敛速度并防止过拟合。In this embodiment, the convolution kernel not only has good robustness, but also reduces the huge amount of parameters, and the parameter amount of the kernel is the same as that of the traditional convolution kernel.
Figure BDA0001746392170000097
times,
Figure BDA0001746392170000098
That is, this method can reduce the maximum
Figure BDA0001746392170000099
parameter amount. Effectively speed up the convergence of the model and prevent overfitting.

此方法在高维度的卷积网络中效果会更加明显,在x维的卷积神经网络中,该核最多能减少

Figure BDA0001746392170000101
的参数,优化效果极为显著。This method is more effective in high-dimensional convolutional networks. In x-dimensional convolutional neural networks, the kernel can reduce at most
Figure BDA0001746392170000101
parameters, the optimization effect is extremely significant.

在本实施例中,进一步地,每一个m*1(m=2k+3,k∈N)的卷积核与对应的一个m*1(m=2k+3,k∈N)的卷积核组成一个等比对称向量组,即ISV,具体的:In this embodiment, further, each m*1 (m=2k+3, k∈N) convolution kernel is convolutional with a corresponding m*1 (m=2k+3, k∈N) convolution kernel The kernel forms a proportional symmetric vector group, namely ISV, specifically:

设an为参数值,a1恒等于1,ISV的长度为m(m=2k+3,k∈N),Let a n be the parameter value, a 1 is always equal to 1, the length of the ISV is m (m=2k+3, k∈N),

ISV=(ISV_1,ISV_2);ISV=(ISV_1,ISV_2);

其中,ISV_1为一个1*m向量,ISV_2为一个m*1向量;Among them, ISV_1 is a 1*m vector, and ISV_2 is an m*1 vector;

Figure BDA0001746392170000102
Figure BDA0001746392170000102

Figure BDA0001746392170000103
Figure BDA0001746392170000103

在本实施例中,由于在常规的卷积操作中涉及了大量的重复运算,会消耗很多不必要的运算时间,所以,本实施例将长、宽为3的径向对称卷积核分解成了一对长度为3的ISV向量和1*1的卷积核组合,卷积核具体变量定义如下:In this embodiment, since a large number of repeated operations are involved in the conventional convolution operation, it will consume a lot of unnecessary operation time. Therefore, this embodiment decomposes the radially symmetric convolution kernel whose length and width are 3 into A pair of ISV vectors of length 3 and a 1*1 convolution kernel are combined. The specific variables of the convolution kernel are defined as follows:

ISV_1卷积核:

Figure BDA0001746392170000104
ISV_1 convolution kernel:
Figure BDA0001746392170000104

ISV_2卷积核:

Figure BDA0001746392170000105
ISV_2 convolution kernel:
Figure BDA0001746392170000105

1*1卷积核:

Figure BDA0001746392170000106
1*1 convolution kernel:
Figure BDA0001746392170000106

以上的卷积核中的an,b均为参数值。an and b in the above convolution kernels are both parameter values.

如图2所示,具体卷积过程如下:As shown in Figure 2, the specific convolution process is as follows:

假设原图为P,先将ISV_1卷积核和1*1卷积核分别对P进行卷积,得到特征图 P1、P2,接下来使用ISV_2卷积核对P1进行卷积,得到特征图P3,然后将P2 与P3相加得到输出的特征图P_output。Assuming that the original image is P, first convolve P with the ISV_1 convolution kernel and the 1*1 convolution kernel respectively to obtain the feature maps P1 and P2, and then use the ISV_2 convolution kernel to convolve P1 to obtain the feature map P3, Then add P2 and P3 to get the output feature map P_output.

经数学运算,上述卷积过程中的卷积过程等价于使用卷积核形为After mathematical operation, the convolution process in the above convolution process is equivalent to using the convolution kernel shape as

Figure BDA0001746392170000107
Figure BDA0001746392170000107

的径向对称卷积核进行卷积,而且,在上述卷积过程中对同一像素点只使用了7次乘法运算5次加法运算,传统的卷积操作要进行9次乘法和8次加法运算,减少了很多运算时间。The radially symmetric convolution kernel is used for convolution, and in the above convolution process, only 7 multiplications and 5 additions are used for the same pixel, while the traditional convolution operation requires 9 multiplications and 8 additions. , which reduces a lot of computation time.

在本实施例中,如图3所示,优选的,如要应用5*5的径向对称卷积核,可将其分解成两对长度分别为5,3的ISV向量和1*1的卷积核组合(每一对ISV的an都不相同)。In this embodiment, as shown in FIG. 3 , preferably, if a radially symmetric convolution kernel of 5*5 is to be applied, it can be decomposed into two pairs of ISV vectors with lengths of 5 and 3 and a 1*1 Combination of convolution kernels (an is different for each pair of ISVs).

进一步地,如图4所示,推广开来,如要应用m*m的径向对称卷积核(m=2k+3,k ∈N),可将其分解成(m-1)/2对长度分别为3,5,7,…,m的ISV向量和1*1的卷积核组合。Further, as shown in Figure 4, by extension, if a radially symmetric convolution kernel of m*m (m=2k+3, k ∈ N) is to be applied, it can be decomposed into (m-1)/2 Combine ISV vectors with lengths of 3, 5, 7, ..., m and 1*1 convolution kernels.

在本实施例中,根据上述描述,这种运算方式在卷积核尺寸较大时可以极大地减少运算量,当尺寸为m*m时,乘法的运算量为原来的

Figure BDA0001746392170000111
倍,加法的运算量为原来的
Figure BDA0001746392170000112
倍。In this embodiment, according to the above description, this operation method can greatly reduce the operation amount when the size of the convolution kernel is large. When the size is m*m, the operation amount of the multiplication is the same as the original.
Figure BDA0001746392170000111
times, the amount of operations for addition is the same as the original
Figure BDA0001746392170000112
times.

输出模块203,用于对1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈ N)的第三特征图进行求和,得到目标特征图,并输出所述目标特征图。The output module 203 is used for summing a first feature map of 1*1 and a third feature map of (m-1)/2 m*1 (m=2k+3, k∈ N) to obtain target feature map, and output the target feature map.

进一步地,在所述对1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N) 的第三特征图进行求和,得到目标特征图,并输出所述目标特征图之后,还包括:Further, after summing the first feature map of 1*1 and the third feature map of (m-1)/2 m*1(m=2k+3,k∈N), we get target feature map, and after outputting the target feature map, it also includes:

判断所述图像的方向对识别结果是否有影响;Determine whether the direction of the image has an influence on the recognition result;

若是,则将1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图进行全局平均池化,并将输出的1+(m-1)/2个值进行softmax处理,得到目标特征图的概率。If so, perform global average pooling on a first feature map of 1*1 and a third feature map of (m-1)/2 m*1 (m=2k+3, k∈N). The output 1+(m-1)/2 values are processed by softmax to obtain the probability of the target feature map.

在本实施例中,该卷积核拆分算法在高维卷积核中对运算量的减少会更明显。如图5所示,使用经分解的径向对称卷积网络对猫、狗进行分类识别。In this embodiment, the convolution kernel splitting algorithm reduces the computation amount more obviously in the high-dimensional convolution kernel. As shown in Figure 5, cats and dogs are classified and recognized using the decomposed radially symmetric convolutional network.

在神经网络的搭建中,在网络的前端使用了8层的3*3分解的径向对称卷积网络,并在其中连接残差层,防止梯度消失。In the construction of the neural network, an 8-layer 3*3 decomposed radially symmetric convolutional network is used at the front end of the network, and the residual layer is connected in it to prevent the gradient from disappearing.

由于图像的方向对识别结果没有影响,在神经网络的末端采用了全局平均池化的方法将最后的两张特征图平均池化,将输出的两个值进行softmax处理得到是猫和狗的概率Since the direction of the image has no effect on the recognition results, the global average pooling method is used at the end of the neural network to average pool the last two feature maps, and the two output values are processed by softmax to obtain the probability of cats and dogs.

进一步地,所述的基于分解径向对称卷积核的卷积优化方法,还包括:Further, the described convolution optimization method based on decomposing radially symmetric convolution kernels also includes:

若否,则将1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图进行空间金字塔池化和全连接层处理,并将输出的1+(m-1)/2个值进行 softmax处理,得到目标特征图的判断结果。If not, perform spatial pyramid pooling and full-scale spatial pyramid pooling on a first feature map of 1*1 and a third feature map of (m-1)/2 m*1 (m=2k+3, k∈N). The connection layer is processed, and the output 1+(m-1)/2 values are subjected to softmax processing to obtain the judgment result of the target feature map.

在本实施例中,如图6所示,使用经分解的径向对称卷积网络识别MNIST 数据集:In this example, as shown in Figure 6, a decomposed radially symmetric convolutional network is used to identify the MNIST dataset:

在图像处理阶段,仍采用上述的预处理方法,只是预处理中不包含旋转。神经网络的主体也采用在网络的前端使用了8层的3*3分解的径向对称卷积网络,并在其中连接残差层,防止梯度消失的方法。但由于数据的减少,相应的减少了神经网络的层数。In the image processing stage, the above-mentioned preprocessing method is still used, but rotation is not included in the preprocessing. The main body of the neural network also uses a radially symmetric convolutional network with 8 layers of 3*3 decomposition at the front end of the network, and connects the residual layer in it to prevent the gradient from disappearing. However, due to the reduction of data, the number of layers of the neural network is correspondingly reduced.

由于图像的方向对识别结果有影响,如识别6,9会因为方向不确定而识别错误,在神经网络的末端采用空间金字塔池化和全连接层搭配softmax输出判断为各数字的结果。Since the direction of the image has an impact on the recognition result, for example, the recognition of 6 and 9 will be wrong due to the uncertainty of the direction. At the end of the neural network, spatial pyramid pooling and fully connected layer with softmax output are used to judge the result of each number.

需要说明的是,这只是本实施例中的两个简单的例子,该网络可以用在更加庞大复杂的识别系统中并取得良好的效果。It should be noted that these are just two simple examples in this embodiment, and the network can be used in a larger and more complex identification system and achieve good results.

本实施例提供的一种基于分解径向对称卷积核的卷积优化装置,通过输入待识别图像,并对所述待识别图像进行预处理;分别利用预先分解m*m径向对称卷积核得到的1个1*1的卷积核和(m-1)/2个1*m(m=2k+3,k∈N)的卷积核,对经过预处理的待识别图像进行卷积,得到1个1*1的第一特征图和(m-1)/2个 1*m(m=2k+3,k∈N)的第二特征图;再利用与(m-1)/2个1*m(m=2k+3,k∈N)的卷积核一一对应的(m-1)/2个m*1(m=2k+3,k∈N)的卷积核,对(m-1)/2个1*m(m=2k+3,k ∈N)的第二特征图进行卷积,得到(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图;对1个1*1的第一特征图和(m-1)/2个m*1(m=2k+3,k∈N)的第三特征图进行求和,得到目标特征图,并输出所述目标特征图。本实施例通过在降低径向对称卷积核计算量的基础上降低参数量,进而达到对卷积进行优化的目的。This embodiment provides a convolution optimization device based on decomposing radially symmetric convolution kernels, by inputting an image to be recognized, and preprocessing the image to be recognized; using pre-decomposed m*m radially symmetric convolution respectively 1 convolution kernel of 1*1 and (m-1)/2 convolution kernels of 1*m (m=2k+3, k∈N) obtained by the kernel, the preprocessed image to be recognized is rolled product, to obtain a first feature map of 1*1 and (m-1)/2 second feature maps of 1*m (m=2k+3, k∈N); then use and (m-1) /2 convolution kernels of 1*m (m=2k+3, k∈N) one-to-one correspondence (m-1)/2 convolutions of m*1 (m=2k+3, k∈N) kernel, convolve the second feature map of (m-1)/2 1*m (m=2k+3, k ∈ N) to get (m-1)/2 m*1 (m=2k +3,k∈N) the third feature map; for 1 1*1 first feature map and (m-1)/2 m*1(m=2k+3,k∈N) the third feature map The feature maps are summed to obtain a target feature map, and the target feature map is output. This embodiment achieves the purpose of optimizing the convolution by reducing the amount of parameters on the basis of reducing the amount of calculation of the radially symmetric convolution kernel.

本发明的一个实施例还提供了一种基于分解径向对称卷积核的卷积优化终端设备,包括:处理器、存储器以及存储在所述存储器中且被配置为由所述处理器执行的计算机程序,所述处理器执行所述计算机程序时实现如上述的基于分解径向对称卷积核的卷积优化方法。An embodiment of the present invention also provides a convolution optimization terminal device based on decomposing radially symmetric convolution kernels, comprising: a processor, a memory, and a function stored in the memory and configured to be executed by the processor A computer program, when the processor executes the computer program, the above-mentioned convolution optimization method based on decomposing radially symmetric convolution kernels is implemented.

本发明的一个实施例还提供了一种计算机可读存储介质,所述计算机可读存储介质包括存储的计算机程序,其中,在所述计算机程序运行时控制所述计算机可读存储介质所在设备执行如上述的基于分解径向对称卷积核的卷积优化方法。An embodiment of the present invention further provides a computer-readable storage medium, where the computer-readable storage medium includes a stored computer program, wherein when the computer program runs, the device where the computer-readable storage medium is located is controlled to execute The above-mentioned convolution optimization method based on decomposing radially symmetric convolution kernels.

相比于现有技术,本发明的一个实施例的有益效果在于:Compared with the prior art, the beneficial effects of an embodiment of the present invention are:

本发明提供的基于分解径向对称卷积核的卷积优化方法、装置、终端设备及计算机可读存储介质,所述方法包括:输入待识别图像,并对所述待识别图像进行预处理;分别利用预先分解m*m径向对称卷积核得到的1个1*1的卷积核和(m-1)/2个1*m(m=2k+3,k∈N)的卷积核,对经过预处理的待识别图像进行卷积,得到1个1*1的第一特征图和(m-1)/2个1*m(m=2k+3,k∈N)的第二特征图;再利用与(m-1)/2个1*m(m=2k+3,k∈N)的卷积核一一对应的(m-1)/2个m*1(m=2k+3,k ∈N)的卷积核,对(m-1)/2个1*m(m=2k+3,k∈N)的第二特征图进行卷积,得到 (m-1)/2个m*1(m=2k+3,k∈N)的第三特征图;对1个1*1的第一特征图和(m-1)/2 个m*1(m=2k+3,k∈N)的第三特征图进行求和,得到目标特征图,并输出所述目标特征图。本发明通过在降低径向对称卷积核计算量的基础上降低参数量,进而达到对卷积进行优化的目的The convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernels provided by the present invention include: inputting an image to be recognized, and preprocessing the image to be recognized; Use a 1*1 convolution kernel and (m-1)/2 1*m (m=2k+3, k∈N) convolution kernels obtained by pre-decomposing m*m radially symmetric convolution kernels respectively. kernel, convolve the preprocessed image to be recognized, and obtain a 1*1 first feature map and (m-1)/2 1*m (m=2k+3, k∈N) first feature maps Two feature maps; re-use (m-1)/2 m*1(m =2k+3,k∈N) convolution kernel, convolve the second feature map of (m-1)/2 1*m(m=2k+3,k∈N), get (m- 1)/2 m*1 (m=2k+3, k∈N) third feature maps; for 1 1*1 first feature map and (m-1)/2 m*1(m =2k+3,k∈N) and the third feature maps are summed to obtain the target feature map, and output the target feature map. The invention achieves the purpose of optimizing the convolution by reducing the amount of parameters on the basis of reducing the amount of calculation of the radially symmetric convolution kernel

以上所述是本发明的优选实施方式,应当指出,对于本技术领域的普通技术人员来说,在不脱离本发明原理的前提下,还可以做出若干改进和润饰,这些改进和润饰也视为本发明的保护范围。The above are the preferred embodiments of the present invention. It should be pointed out that for those skilled in the art, without departing from the principles of the present invention, several improvements and modifications can be made, and these improvements and modifications may also be regarded as It is the protection scope of the present invention.

本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机程序来指令相关的硬件来完成,所述的程序可存储于一计算机可读取存储介质中,该程序在执行时,可包括如上述各方法的实施例的流程。其中,所述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory,ROM) 或随机存储记忆体(Random AccessMemory,RAM)等。Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. During execution, the processes of the embodiments of the above-mentioned methods may be included. The storage medium may be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM) or the like.

Claims (7)

1.一种基于分解径向对称卷积核的卷积优化方法,适于在计算设备中执行,其特征在于,包括如下步骤:1. a convolution optimization method based on decomposing radially symmetric convolution kernel, is suitable for carrying out in computing equipment, it is characterized in that, comprise the steps: 输入待识别图像,并对所述待识别图像进行预处理,包括:根据预设参数,对所述待识别图像进行随机拉伸和明暗调整,并加入特定的高斯噪声;根据卷积处理的要求,对所述待识别图像进行0~π/2角度的旋转和切割;Inputting the image to be recognized, and preprocessing the image to be recognized, including: randomly stretching and adjusting the brightness and darkness of the image to be recognized according to preset parameters, and adding specific Gaussian noise; according to the requirements of convolution processing , rotate and cut the to-be-recognized image at an angle of 0 to π/2; 分别利用预先分解m*m径向对称卷积核得到的1个1*1的卷积核和(m-1)/2个1*m的卷积核,对经过预处理的待识别图像进行卷积,得到1个1*1的第一特征图和(m-1)/2个1*m的第二特征图;再利用预先分解m*m径向对称卷积核得到的与(m-1)/2个1*m的卷积核一一对应的(m-1)/2个m*1的卷积核,对(m-1)/2个1*m的第二特征图进行卷积,得到(m-1)/2个m*1的第三特征图;Using a 1*1 convolution kernel and (m-1)/2 1*m convolution kernels obtained by decomposing m*m radially symmetric convolution kernels in advance, the preprocessed images to be recognized are processed. Convolve to obtain a 1*1 first feature map and (m-1)/2 1*m second feature maps; then use the pre-decomposed m*m radially symmetric convolution kernel to obtain and (m -1)/2 1*m convolution kernels one-to-one (m-1)/2 m*1 convolution kernels, for (m-1)/2 1*m second feature maps Perform convolution to obtain (m-1)/2 m*1 third feature maps; 对1个1*1的第一特征图和(m-1)/2个m*1的第三特征图进行求和,得到目标特征图,并输出所述目标特征图;其中m=2k+3,k∈N。Summing a first feature map of 1*1 and a third feature map of (m-1)/2 m*1 to obtain a target feature map, and output the target feature map; where m=2k+ 3, k∈N. 2.根据权利要求1所述的基于分解径向对称卷积核的卷积优化方法,其特征在于,每一个卷积核的卷积核矩阵满足如下公式:2. the convolution optimization method based on decomposing radially symmetric convolution kernel according to claim 1, is characterized in that, the convolution kernel matrix of each convolution kernel satisfies following formula:
Figure FDA0003118710850000011
Figure FDA0003118710850000011
其中,
Figure FDA0003118710850000012
in,
Figure FDA0003118710850000012
3.根据权利要求1所述的基于分解径向对称卷积核的卷积优化方法,其特征在于,在所述对1个1*1的第一特征图和(m-1)/2个m*1的第三特征图进行求和,得到目标特征图,并输出所述目标特征图之后,还包括:3. The convolution optimization method based on decomposing radially symmetric convolution kernels according to claim 1, characterized in that, in the pair of 1 first feature map of 1*1 and (m-1)/2 The third feature map of m*1 is summed to obtain the target feature map, and after outputting the target feature map, it also includes: 判断所述图像的方向对识别结果是否有影响;Determine whether the direction of the image has an influence on the recognition result; 若是,则将1个1*1的第一特征图和(m-1)/2个m*1的第三特征图进行全局平均池化,并将输出的1+(m-1)/2个值进行softmax处理,得到目标特征图的概率。If so, perform global average pooling on one 1*1 first feature map and (m-1)/2 m*1 third feature maps, and combine the output 1+(m-1)/2 Each value is subjected to softmax processing to obtain the probability of the target feature map. 4.根据权利要求3所述的基于分解径向对称卷积核的卷积优化方法,其特征在于,还包括:4. the convolution optimization method based on decomposing radially symmetric convolution kernel according to claim 3, is characterized in that, also comprises: 若否,则将1个1*1的第一特征图和(m-1)/2个m*1的第三特征图进行空间金字塔池化和全连接层处理,并将输出的1+(m-1)/2个值进行softmax处理,得到目标特征图的判断结果。If not, perform spatial pyramid pooling and fully connected layer processing on one 1*1 first feature map and (m-1)/2 m*1 third feature maps, and convert the output 1+( m-1)/2 values are processed by softmax to obtain the judgment result of the target feature map. 5.一种基于分解径向对称卷积核的卷积优化装置,其特征在于,包括:5. a convolution optimization device based on decomposing radially symmetric convolution kernel, is characterized in that, comprises: 输入模块,用于输入待识别图像,并对所述待识别图像进行预处理,包括:根据预设参数,对所述待识别图像进行随机拉伸和明暗调整,并加入特定的高斯噪声;根据卷积处理的要求,对所述待识别图像进行0~π/2角度的旋转和切割;The input module is used for inputting the to-be-recognized image and preprocessing the to-be-recognized image, including: randomly stretching and shading the to-be-recognized image according to preset parameters, and adding specific Gaussian noise; According to the requirements of convolution processing, the image to be recognized is rotated and cut at an angle of 0 to π/2; 卷积优化模块,用于分别利用预先分解m*m径向对称卷积核得到的1个1*1的卷积核和(m-1)/2个1*m的卷积核,对经过预处理的待识别图像进行卷积,得到1个1*1的第一特征图和(m-1)/2个1*m的第二特征图;再利用预先分解m*m径向对称卷积核得到的与(m-1)/2个1*m的卷积核一一对应的(m-1)/2个m*1的卷积核,对(m-1)/2个1*m的第二特征图进行卷积,得到(m-1)/2个m*1的第三特征图;The convolution optimization module is used to use a 1*1 convolution kernel and (m-1)/2 1*m convolution kernels obtained by decomposing m*m radially symmetric convolution kernels in advance. The preprocessed image to be recognized is convolved to obtain a 1*1 first feature map and (m-1)/2 1*m second feature maps; then use the pre-decomposed m*m radial symmetry volume The product kernel obtained corresponds to (m-1)/2 1*m convolution kernels one-to-one (m-1)/2 m*1 convolution kernels, for (m-1)/2 1 The second feature map of *m is convolved to obtain (m-1)/2 third feature maps of m*1; 输出模块,用于对1个1*1的第一特征图和(m-1)/2个m*1的第三特征图进行求和,得到目标特征图,并输出所述目标特征图;其中m=2k+3,k∈N。The output module is used to sum the first feature map of 1*1 and the third feature map of (m-1)/2 m*1 to obtain the target feature map, and output the target feature map; Where m=2k+3, k∈N. 6.一种基于分解径向对称卷积核的卷积优化终端设备,其特征在于,包括:处理器、存储器以及存储在所述存储器中且被配置为由所述处理器执行的计算机程序,所述处理器执行所述计算机程序时实现如权利要求1至4任一项所述的基于分解径向对称卷积核的卷积优化方法。6. a convolution optimization terminal device based on decomposing radially symmetric convolution kernel, is characterized in that, comprises: processor, memory and the computer program that is stored in described memory and is configured to be executed by described processor, When the processor executes the computer program, the method for convolution optimization based on decomposing radially symmetric convolution kernels according to any one of claims 1 to 4 is implemented. 7.一种计算机可读存储介质,其特征在于,所述计算机可读存储介质包括存储的计算机程序,其中,在所述计算机程序运行时控制所述计算机可读存储介质所在设备执行如权利要求1至4任一项所述的基于分解径向对称卷积核的卷积优化方法。7. A computer-readable storage medium, characterized in that the computer-readable storage medium comprises a stored computer program, wherein, when the computer program is run, the device where the computer-readable storage medium is located is controlled to perform as claimed in the claims The convolution optimization method based on decomposing radially symmetric convolution kernels according to any one of 1 to 4.
CN201810852407.5A 2018-07-27 2018-07-27 Convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernel Active CN109034249B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810852407.5A CN109034249B (en) 2018-07-27 2018-07-27 Convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernel

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810852407.5A CN109034249B (en) 2018-07-27 2018-07-27 Convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernel

Publications (2)

Publication Number Publication Date
CN109034249A CN109034249A (en) 2018-12-18
CN109034249B true CN109034249B (en) 2021-08-06

Family

ID=64646640

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810852407.5A Active CN109034249B (en) 2018-07-27 2018-07-27 Convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernel

Country Status (1)

Country Link
CN (1) CN109034249B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109886273B (en) * 2019-02-26 2022-12-16 四川大学华西医院 A CMR Image Segmentation and Classification System
CN110334798B (en) * 2019-03-13 2021-06-08 北京地平线机器人技术研发有限公司 Feature data extraction method and device and instruction generation method and device
CN109978137B (en) * 2019-03-20 2021-03-16 厦门美图之家科技有限公司 Processing method of convolutional neural network
CN112712461B (en) * 2019-10-24 2024-04-19 Tcl科技集团股份有限公司 Image deconvolution processing method and device and terminal equipment
CN111680678B (en) * 2020-05-25 2022-09-16 腾讯科技(深圳)有限公司 Target area identification method, device, equipment and readable storage medium
CN113591025B (en) * 2021-08-03 2024-06-14 深圳思谋信息科技有限公司 Feature map processing method and device, convolutional neural network accelerator and medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104346621A (en) * 2013-07-30 2015-02-11 展讯通信(天津)有限公司 Method and device for creating eye template as well as method and device for detecting eye state
CN105139017A (en) * 2015-08-27 2015-12-09 重庆理工大学 License plate positioning algorithm fusing affine invariant corner feature and visual color feature
CN106557743A (en) * 2016-10-26 2017-04-05 桂林电子科技大学 A kind of face characteristic extraction system and method based on FECNN
US9953236B1 (en) * 2017-03-10 2018-04-24 TuSimple System and method for semantic segmentation using dense upsampling convolution (DUC)
CN107967459A (en) * 2017-12-07 2018-04-27 北京小米移动软件有限公司 convolution processing method, device and storage medium

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104346621A (en) * 2013-07-30 2015-02-11 展讯通信(天津)有限公司 Method and device for creating eye template as well as method and device for detecting eye state
CN105139017A (en) * 2015-08-27 2015-12-09 重庆理工大学 License plate positioning algorithm fusing affine invariant corner feature and visual color feature
CN106557743A (en) * 2016-10-26 2017-04-05 桂林电子科技大学 A kind of face characteristic extraction system and method based on FECNN
US9953236B1 (en) * 2017-03-10 2018-04-24 TuSimple System and method for semantic segmentation using dense upsampling convolution (DUC)
CN107967459A (en) * 2017-12-07 2018-04-27 北京小米移动软件有限公司 convolution processing method, device and storage medium

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于数字图像检测和虚拟现实的桥梁施工控制技术研究;胡帮义;<<中国优秀硕士学位论文全文数据库 工程科技II辑>>;20170315(第3期);正文第20-60页 *

Also Published As

Publication number Publication date
CN109034249A (en) 2018-12-18

Similar Documents

Publication Publication Date Title
CN109034249B (en) Convolution optimization method, device, terminal device and computer-readable storage medium based on decomposing radially symmetric convolution kernel
CN109658344B (en) Image denoising method, device, device and storage medium based on deep learning
Zhang et al. Learning deep CNN denoiser prior for image restoration
EP3407266B1 (en) Artificial neural network calculating device and method for sparse connection
WO2021027193A1 (en) Face clustering method and apparatus, device and storage medium
WO2020056791A1 (en) Method and apparatus for super-resolution reconstruction of multi-scale dilated convolution neural network
CN110232661A (en) Low illumination colour-image reinforcing method based on Retinex and convolutional neural networks
CN109949255A (en) Image reconstruction method and device
CN110781895A (en) Image semantic segmentation method based on convolutional neural network
CN108510485A (en) It is a kind of based on convolutional neural networks without reference image method for evaluating quality
CN108470320A (en) A kind of image stylizing method and system based on CNN
CN111489364A (en) Medical image segmentation method based on lightweight full convolution neural network
WO2023065759A1 (en) Video action recognition method based on spatial-temporal enhanced network
CN114511576B (en) Scale-adaptive feature-enhanced deep neural network for image segmentation method and system
US11663705B2 (en) Image haze removal method and apparatus, and device
CN113870283B (en) Portrait matting method, device, computer equipment and readable storage medium
CN109409210B (en) Face detection method and system based on SSD (solid State disk) framework
CN116452810B (en) Multi-level semantic segmentation method and device, electronic equipment and storage medium
CN112446379A (en) Self-adaptive intelligent processing method for dynamic large scene
CN114241230B (en) Target detection model pruning method and target detection method
CN112036461B (en) Handwriting digital image recognition method, device, equipment and computer storage medium
CN111382839B (en) Method and device for pruning neural network
CN113095333A (en) Unsupervised feature point detection method and unsupervised feature point detection device
CN110321964A (en) Identification model update method and relevant apparatus
CN116051408A (en) Image depth denoising method based on residual error self-coding

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant