Image recognition method based on deep course learning

An image recognition and course technology, applied in the field of image recognition, can solve problems such as inability to fit real annotation results, deep convolutional neural network penalties, etc., and achieve the effects of reduced training workload, fast update speed, and high prediction accuracy.

Active Publication Date: 2020-05-15
HEFEI UNIV OF TECH +1
View PDF4 Cites 15 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In the existing technology, the deep convolutional neural network is often punished due to the large difference between the prediction result and the correct labeling result, which cannot fit the real labeling result.

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
  • Image recognition method based on deep course learning
  • Image recognition method based on deep course learning
  • Image recognition method based on deep course learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0030] An image recognition method based on deep course learning. In this embodiment, given image data sets such as CIFAR10 and CIFAR100, we use different deep neural network models to verify the effectiveness of BLCL learning.

[0031] Each data set is first divided into two parts, the training set and the test set, and the training data is divided into batches for training.

[0032] Specific steps are as follows:

[0033] S10. Construct a teacher network and a student network based on a deep convolutional neural network; the convolutional neural network in this embodiment uses ResNet-8, ResNet-20, ResNet-32 and DensNet-BC.

[0034] S20. Using the training samples to perform image recognition and classification training on the teacher network; the teacher network learns the training samples and predicts the probability that the training samples belong to each category;

[0035] The training samples are CIFAR10 and CIFAR100 data sets. The training set contains 50,000 pictures...

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

No PUM Login to view more

Abstract

The invention discloses an image recognition method based on deep course learning, and belongs to the field of image recognition. The method comprises the following steps: constructing teacher and student networks based on a deep convolutional neural network; performing image classification training on the teacher network by using a training sample, and predicting the probability that the trainingsample belongs to each category; calculating the difference between the prediction of the teacher network and the labels to update the parameters; transmitting the prediction information to a studentnetwork; training the student network; guiding student network training according to the prediction information result of the teacher network; calculating a difference updating parameter between thestudent network prediction result and the label; completing student network classification training; and the trained student network realizes recognition and classification of the images. According tothe method, the process of human learning from easiness to difficulty is simulated, the training process is reasonable, the workload is greatly reduced, the network parameters are updated quickly, the influence of the samples is balanced by gradient differences generated by different samples, the prediction precision is higher, and the performance is more reliable and stable.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to an image recognition method based on deep course learning. Background technique [0002] Due to the large increase in computing power and the increase in training data, the training conditions of deep neural networks have been met. Due to its strong fitting ability, deep neural networks have achieved the best results in many tasks. However, since the parameters of deep neural networks are updated based on a large number of samples, samples play a crucial role in the training process. Therefore, the training work is intensive, the training process is complicated, and the parameters of the deep neural network are updated slowly. [0003] At the same time, the deep convolutional neural network will calculate the gradient of each parameter according to the prediction results of the training samples and the correctly classified labels of the deep convolutional neural networ...

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
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/2415G06F18/214Y02T10/40
Inventor 胡珍珍秦伟刘祥龙洪日昌汪萌
Owner HEFEI UNIV OF TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products