Constructing a mobile visual search framework for Dunhuang murals based on fine-tuned CNN and ontology semantic distance

Zeng, Ziming; Sun, Shouqiang; Sun, Jingjing; Yin, Jie; Shen, Yueyan

doi:10.1108/el-09-2021-0173

Cited by 5 publications

(3 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Multiple information sources in the form of two- dimensional images are transmitted to the CNN input layer of different image channels, and it is studied whether using multimodal images as input can improve the segmentation results. Their results show better performance than those using single modal input [9].…”

Section: Convolutional Neural Networkmentioning

confidence: 93%

Semantic segmentation of remote sensing image based on U-NET

Yao,

Jia,

Dai

2023

Third International Conference on Image Processing and Intelligent Control (IPIC 2023)

View full text Add to dashboard Cite

At present, the image processing of remote sensing technology mainly depends on the transcendental ability of human beings, and it needs to spend a lot of artificial resources to mark. Therefore, this paper proposes a research and application of semantic segmentation method for remote sensing images based on convolutional neural network. Normalize the data, subtract the mean value and divide it by the standard deviation to standardize, divide the data, introduce data enhancement to further enhance the training data, and create a convolutional neural network and a training network. Each layer of U-NET is composed of three layers of convolution, and features are extracted and integrated by pooling or up-sampling. At the last layer, all the previously extracted features are classified into two categories to realize the semantic segmentation of the image. The experimental results show that the F1 score, Recall score and Precision score of this method are 84.31%, 89.59% and 79.62%, respectively. By introducing U-NET, the semantic segmentation accuracy of remote sensing images is improved. Compared with the traditional full convolution neural network, U-NET has been improved. Through the stronger connection between layers, plus up-sampling and down-convolution, features can be fully extracted and accurate segmentation can be achieved with fewer training samples.

show abstract

Section: Convolutional Neural Networkmentioning

confidence: 93%

Semantic segmentation of remote sensing image based on U-NET

Yao,

Jia,

Dai

2023

Third International Conference on Image Processing and Intelligent Control (IPIC 2023)

View full text Add to dashboard Cite

show abstract

“…To enable users to effectively search for same, related and diverse Dunhuang mural images, Zeng et al proposed a CNN model tuned in the data set of Dunhuang murals. The finely tuned ResNet152 is the best choice for searching for similar images at the visual feature level, while improving the diversity of search results [13]. Wang et al studied the robustness of graph convolutional network (GCN), and proposed a new "false node attack" to attack GCN by adding malicious fake nodes.…”

Section: Related Workmentioning

confidence: 99%

Application of Health Education Program based on Intelligent Recommendation Algorithm in the Development of School-age Children's Hand Hygiene Behavior

Zhao,

Li,

Wang

et al. 2024

SCPE

View full text Add to dashboard Cite

The cultivation of hand hygiene behavior among school-age children is an important way to prevent the spread of diseases and ensure children’s health. However, traditional health education methods lack personalized programs tailored to each child, which cannot effectively improve their hand hygiene awareness and behavior. In response to this issue, the study combines multi-objective particle swarm optimization algorithm to provide personalized hand hygiene behavior development recommendations for school-age children, improving their hand hygiene awareness and behavioral level. The study adopted multi-objective particle swarm optimization algorithm and convolutional neural network, combined with the personalized needs of school-age children and the goal of cultivating hand hygiene behavior, and proposed a health education plan based on intelligent recommendation algorithm. Through deep learning, match the hand hygiene needs of school-age children with corresponding health education plans to achieve personalized recommendations. The average accuracy of the health education recommendation plan for hand hygiene behavior cultivation of school-age children based on multi-objective particle swarm optimization algorithm reaches 99.07%. Meanwhile, after introducing convolutional neural networks, the feature matching error between the recommended scheme and school-age children ranges from 10-1 to 10-2. After testing, the designed algorithm performs more stably and has less fluctuations under different sparsity conditions. Health education solutions based on intelligent recommendation algorithms can provide personalized solutions for school-age children, effectively cultivate their hand hygiene behavior, and meet various health needs.

show abstract

“…The convolutional operation of the convolutional layer calculates a linear response; the pooling layer pools the feature data after convolution, and the number of training parameters decreases significantly after down sampling; the activation function layer adds non-linear variation. The activation function layer increases the nonlinear variation to update the generalization ability (Zeng et al 2022). As the audio data of musical tones will be interspersed with different levels of noise, the noise will have an impact on the model recognition and detection.…”

Section: Design Of Music Sound Recognition Model Fusing Rnn and Cnnmentioning

confidence: 99%

Research on Musical Tone Recognition Method Based on Improved RNN for Vocal Music Teaching Network Courses

Long

2023

International Journal of Web-Based Learning and Teaching Technologies

View full text Add to dashboard Cite

The test results show that the fast Fourier process with multiple time superposition and a dimension length of 40 is most beneficial to the accuracy of the model. The loss curve value of the convolutional recurrent network model (CRN) is much lower than the other three models. The music tone recognition model learns better. The accuracy rate value and recall rate value of the CRN are the highest, and the accuracy rates of the four music tone indicators are 94.6%, 92.4%, 93.5%, 92.5%, and the recall rates were 93.2%, 94.9%, 95.2%, and 88.6% respectively; the improved algorithm was the most accurate in terms of F1 values and is suitable for use in vocal music teaching courses. The results show that the algorithm can be broadly performed in the zone of music tone recognition and has a certain contribution to the development of the field of music tone recognition.

show abstract

Constructing a mobile visual search framework for Dunhuang murals based on fine-tuned CNN and ontology semantic distance

Cited by 5 publications

References 51 publications

Semantic segmentation of remote sensing image based on U-NET

Semantic segmentation of remote sensing image based on U-NET

Application of Health Education Program based on Intelligent Recommendation Algorithm in the Development of School-age Children's Hand Hygiene Behavior

Research on Musical Tone Recognition Method Based on Improved RNN for Vocal Music Teaching Network Courses

Contact Info

Product

Resources

About