Unlock instant, AI-driven research and patent intelligence for your innovation.
An optimization method for mutual blog intelligence evolution between multiple groups of convolutional neural networks
What is Al technical title?
Al technical title is built by PatSnap Al team. It summarizes the technical point description of the patent document.
A convolutional neural network and neural network technology, applied in the field of artificial intelligence deep learning, can solve the problems of consuming manpower and material resources, heavy workload of collecting and labeling labels, and closed data sets
Active Publication Date: 2021-05-11
SHANDONG LINGNENG ELECTRONIC TECH CO LTD
View PDF13 Cites 0 Cited by
Summary
Abstract
Description
Claims
Application Information
AI Technical Summary
This helps you quickly interpret patents by identifying the three key elements:
Problems solved by technology
Method used
Benefits of technology
Problems solved by technology
[0003] Although convolutional neural networks are widely used, one of the key factors limiting their development is the collection of training data sets.
In order to train a convolutional neural network to achieve target detection, it is necessary to feed the network a large data set. Taking the ImageNet data set as an example, it contains 1000 object categories, and each category has 1200 pictures. The training The data set has reached 1,200,000 pictures. The workload of collecting and labeling these pictures is very large, which consumes a lot of manpower and material resources.
And the data sets collected by many companies are not open to the public, which severely limits the development and application of deep learning
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 more
Image
Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
Click on the blue label to locate the original text in one second.
Reading with bidirectional positioning of images and text.
Smart Image
Examples
Experimental program
Comparison scheme
Effect test
Embodiment 1
[0032] An optimization method for mutual blog intelligence evolution between multiple groups of convolutional neural networks, such as figure 1 shown, including:
[0033] (1) Select m convolutional neural networks, including net1,...,neti,...,netm, m≥2 and m is an integer, 1≤i≤m, neti refers to any volume in m convolutional neural networks Convolutional neural network; the initialization states of m convolutional neural networks are all trained based on the same data set, which has n categories, and one category has multiple training pictures, including 1,...,j,...,n; j refers to any one of the n categories;
[0034] (2) Use m convolutional networks to infer and classify the single images collected by the camera, and the convolutional neural network neti obtains the classification results, that is, n probability values X i1 ,...,X ij …X in ,X ij It refers to the probability that the currently collected image is category j, and the maximum value is selected from n probab...
Embodiment 2
[0039] An optimization method for mutual blog intelligence evolution between multiple groups of convolutional neural networks, such as figure 2 shown, including:
[0040] (1) Select AlexNet and GoogleNet, which won the category recognition champions in the ImageNet Large Scale Visual Recognition Challenge in 2012 and 2014, respectively, and apply them to embedded device security monitoring. These two convolutional neural networks The networks are all trained on the IamgeNet training dataset with 1000 categories.
[0041] (2) Security monitoring captures images at a limited time point, and at a certain moment, an image A is captured by monitoring and sent to AlexNet and GoogleNet for real-time recognition. AlexNet calculates that the target in image A is the probability of 1000 classifications in the IamgeNet data set. x 1 , X 2 , X 3 …X 1000 , where the largest probability value is X i , that is, AlexNet believes that the target in the A image is most similar to the i-t...
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 relates to an optimization method for mutual blogging intelligent evolution between multiple convolutional neural networks. By comparing the accuracy of recognition images between multiple convolutional neural networks, the images are collected in real time by a camera as a data set and fed back to the recognition effect. Fine-tuning a bad network to optimize the network can better classify images of this category; security monitoring can be used to automatically obtain a training set and feed it to a network with poor classification effect, and control its automatic fine-tuning (Finturn), saving a lot of manpower Material manually collects pictures and writes labels, and realizes automatic training evolution between convolutional neural networks.
Description
technical field [0001] The invention relates to an optimization method for mutual blog intelligent evolution between multiple groups of convolutional neural networks, belonging to the field of deep learning of artificial intelligence. Background technique [0002] In recent years, artificial intelligence technology has developed rapidly, and the convolutional neural network (CNN) in the field of deep learning has become one of the research hotspots in many scientific fields, especially in the field of pattern classification, because the convolutional neural network avoids the complex early preprocessing of images. , can directly input the original image, and has a wide range of applications in target recognition, behavior recognition, face recognition and other fields. [0003] Although convolutional neural networks are widely used, a very critical factor limiting their development is the collection of training data sets. In order to train a convolutional neural network to ...
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
Application Date:The date an application was filed.
Publication Date:The date a patent or application was officially published.
First Publication Date:The earliest publication date of a patent with the same application number.
Issue Date:Publication date of the patent grant document.
PCT Entry Date:The Entry date of PCT National Phase.
Estimated Expiry Date:The statutory expiry date of a patent right according to the Patent Law, and it is the longest term of protection that the patent right can achieve without the termination of the patent right due to other reasons(Term extension factor has been taken into account ).
Invalid Date:Actual expiry date is based on effective date or publication date of legal transaction data of invalid patent.
Login to View More 
Login to View More