Time-coherent embeddings for Wireless Capsule Endoscopy DOI
Guillem Pascual, Jordi Vitrià, Santi Seguí

и другие.

2022 26th International Conference on Pattern Recognition (ICPR), Год журнала: 2022, Номер 6791, С. 4248 - 4255

Опубликована: Авг. 21, 2022

Deep learning models thrive with high amounts of data where the classes are, usually, appropriately balanced. In medical imaging, however, we often encounter opposite case. Wireless Capsule Endoscopy is not an exception; even if huge could be obtained, labeling each frame a video take up to twelve hours for expert physician. Those videos would show no pathologies most patients, while minority have few frames associated pathology. Overall, there low and great unbalance. Self-supervised provides means use unlabelled initialize that can perform better under described circumstance. We propose novel contrastive loss derived from Triplet Loss, crafted leverage temporal information in endoscopy videos. our model outperforms existing other methods several tasks.

Язык: Английский

The analysis of generative adversarial network in sports education based on deep learning DOI Creative Commons

Eerdenisuyila Eerdenisuyila,

Hongming Li, Wei Chen

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Дек. 5, 2024

The importance of mental health is increasingly emphasized in modern society. assessment qualities among college and university students as the future workforce holds significant significance. Therefore, this study, aiming to streamline process writing quality evaluations enhance fairness comments, explores use Generative Adversarial Network (GAN) technology deep learning evaluate through unique avenue sports. Firstly, GAN Sequence (SeqGAN) models are introduced. Secondly, employed construct a model for generating evaluation texts, encompassing construction generator discriminator, along with introduction reward function. Finally, constructed utilized train on texts related engaged sports, validating effectiveness model. results indicate: (1) pre-training text generation stabilizes after 10th epoch. In contrast, discriminator gradually 35th epoch, demonstrating overall good training effectiveness. (2) When generator's update speed surpasses that model's loss does not converge. However, reduction ratio rounds between two, there noticeable improvement convergence (3) mean score adaptability highest four indicators, suggesting strong correlation comment quality. validate proposed semantic control. This study aims advance level education sports domain, providing theoretical references enhancing assessments other subjects well.

Язык: Английский

Процитировано

0
Abnormal Event Detection using Additive Summarization Model for Intelligent Transportation Systems DOI Open Access

G. Balamurugan,

J. Jayabharathy

International Journal of Advanced Computer Science and Applications, Год журнала: 2022, Номер 13(5)

Опубликована: Янв. 1, 2022

Video surveillance is used for capturing the abnormal events on roadsides that are caused due to improper driving, accidents, and hindrances resulting in transportation lags life-critical issues. It essential highlight accident keyframes videos achieve intelligent video surveillance. summarization plays a vital role summarizing keyframe an event from stacked input. The observed converted into frames analyzed providing accurate analysis forecast guiding users avoiding such events. main issues arise inconsistency between spatiotemporal redundancies classification of sequence verification This article introduces Additive Event Summarization Method (AESM) projecting classified through gated recurrent unit learning paradigm. In this process, gates assigned unclassified active verification. Based sequence, abnormality summarized with higher accuracy than state art techniques. proposed method relies heterogeneous features classifying better structural indices. method’s performance using metrics accuracy, false rate, time, SSIM, F1-Score.

Язык: Английский

Процитировано

2
COHETS: A highlight extraction method using textual streams of streaming videos DOI
Chien Chin Chen,

Liang-Wei Lo,

Sheng-Jie Lin

и другие.

Knowledge-Based Systems, Год журнала: 2022, Номер 258, С. 110000 - 110000

Опубликована: Окт. 17, 2022

Язык: Английский

Процитировано

2
Abstractive text summarization employing ontology-based knowledge-aware multi-focus conditional generative adversarial network (OKAM-CGAN) with hybrid pre-processing methodology DOI

Nafees Muneera M,

P Sriramya

Multimedia Tools and Applications, Год журнала: 2022, Номер 82(15), С. 23305 - 23331

Опубликована: Ноя. 17, 2022

Язык: Английский

Процитировано

1
Time-coherent embeddings for Wireless Capsule Endoscopy DOI
Guillem Pascual, Jordi Vitrià, Santi Seguí

и другие.

2022 26th International Conference on Pattern Recognition (ICPR), Год журнала: 2022, Номер 6791, С. 4248 - 4255

Опубликована: Авг. 21, 2022

Deep learning models thrive with high amounts of data where the classes are, usually, appropriately balanced. In medical imaging, however, we often encounter opposite case. Wireless Capsule Endoscopy is not an exception; even if huge could be obtained, labeling each frame a video take up to twelve hours for expert physician. Those videos would show no pathologies most patients, while minority have few frames associated pathology. Overall, there low and great unbalance. Self-supervised provides means use unlabelled initialize that can perform better under described circumstance. We propose novel contrastive loss derived from Triplet Loss, crafted leverage temporal information in endoscopy videos. our model outperforms existing other methods several tasks.

Язык: Английский

Процитировано

1