An Image Classification Method Based on Mirror Invariant Convolutional Neural Network
A technology of convolutional neural network and classification method, which is applied in the direction of neural learning method, biological neural network model, neural architecture, etc., can solve the problem that the convolutional neural network does not have mirror invariance, and achieve the improvement of classification accuracy and the improvement of training process. Fast, less training time results
Active Publication Date: 2021-10-26
ZHEJIANG UNIV OF TECH
View PDF3 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0005] Aiming at the problem that the existing convolutional neural network does not have mirror invariance, the present invention proposes a mirror-invariant convolutional neural network for image classification, specifically a convolutional neural network that converts the bottom layer of the convolutional neural network to Part of the feature map of the convolutional layer and the corresponding convolution kernel are mirrored and flipped, so that the trained convolutional neural network model for image classification has mirror invariance
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment
[0029] A kind of image classification method based on mirror invariant convolutional neural network, comprises the steps:
[0030] Step 1: Read the weight file and parameter configuration file of the convolutional neural network to obtain the initial convolutional neural network, denoted as N.
[0031] Step 2: Prepare the training sample set I={(X i ,Y i )|i=1,2,3,…,m}, where X i Denotes the i-th sample image, Y i Indicates the label corresponding to the i-th sample image, Y i ∈{0,1,2,...,k-1}, k means that there are k categories in total in the image classification task, in this embodiment k=2, m means the number of samples in the training sample set, in this embodiment m=65000.
[0032] Step 3: Start the iterative training of the network. Each iteration randomly selects a batch of training samples from the training sample set I as a subset of the training sample set, denoted as I t , where t represents the t-th iteration of network training.
[0033] Step 4: Input a b...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The present invention proposes a convolutional neural network with mirror image invariance for image classification, specifically a partial feature map of the bottom convolutional layer of the convolutional neural network and the corresponding convolution kernel in the training process. The mirror flip method. By using the convolutional neural network of the present invention to train training samples, the trained convolutional neural network model can have mirror invariance, and better classification performance can be obtained for mirror-symmetrical images.
Description
technical field [0001] The invention relates to the technical field of image processing, in particular to an image classification method based on a mirror invariant convolutional neural network. Background technique [0002] With the continuous development of the field of computer vision and deep learning, new convolutional neural network models are emerging in an endless stream. The improvement of the network model mainly focuses on the deepening of the network depth, the increase of the network width and density. Due to the powerful image processing performance of the convolutional neural network itself, few people spend energy on improving the features extracted by the convolutional neural network. However, the features extracted by the convolutional neural network do have a great impact on the recognition performance of the convolutional neural network. [0003] At present, some scholars have proposed methods for improving the features extracted by convolutional neural...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06K9/62G06N3/04G06N3/08
CPCG06N3/084G06N3/045G06F18/24G06F18/214
Inventor 汪敏倩高飞葛一粟张元鸣卢书芳程振波陆佳炜
Owner ZHEJIANG UNIV OF TECH