Semi-supervised learning advances species recognition for aquatic biodiversity monitoring DOI Creative Commons
Dongliang Ma,

Jine Wei,

Likai Zhu

et al.

Frontiers in Marine Science, Journal Year: 2024, Volume and Issue: 11

Published: May 28, 2024

Aquatic biodiversity monitoring relies on species recognition from images. While deep learning (DL) streamlines the process, performance of these method is closely linked to large-scale labeled datasets, necessitating manual processing with expert knowledge and consume substantial time, labor, financial resources. Semi-supervised (SSL) offers a promising avenue improve DL models by utilizing extensive unlabeled samples. However, complex collection environments long-tailed class imbalance aquatic make SSL difficult implement effectively. To address challenges in within scheme, we propose Wavelet Fusion Network Consistency Equilibrium Loss function. The former mitigates influence data environment fusing image information at different frequencies decomposed through wavelet transform. latter improves scheme refining consistency loss function adaptively adjusting margin for each class. Extensive experiments are conducted FishNet dataset. As expected, our existing up 9.34% overall classification accuracy. With accumulation data, improved limited shows potential advance conservation.

Language: Английский

Water quality classification framework for IoT-enabled aquaculture ponds using deep learning based flexible temporal network model DOI
Arepalli Peda Gopi, K. Jairam Naik

Earth Science Informatics, Journal Year: 2025, Volume and Issue: 18(2)

Published: March 31, 2025

Language: Английский

Citations

0
An Evaluation of Copyright Infringements Committed Through Generative Artificial Intelligence and Especially Deep Fakes DOI
Barış GÖZÜBÜYÜK

Advances in public policy and administration (APPA) book series, Journal Year: 2025, Volume and Issue: unknown, P. 195 - 212

Published: Jan. 17, 2025

Deepfake technology, a form of Generative Artificial Intelligence (Gen-AI), allows for the manipulation individuals' voices and images to generate fake videos where people appear be saying or doing things they never actually said did. This has led concerns about copyright infringement other rights violations, though this study specifically focuses on issues. Dealing with these violations global scale presents significant urgent challenges. However, necessity providing fair compensation creators using their work is not just step, but fundamental requirement towards allowing legal use deepfake technology. As solution, suggests remuneration model address infringements related deepfakes.

Language: Английский

Citations

0
Dynamic monitoring of surface area and water volume of reservoirs using satellite imagery, computer vision and deep learning DOI
Ariane Marina de Albuquerque Teixeira, Leonardo Vidal Batista, Richarde Marques da Silva

et al.

Remote Sensing Applications Society and Environment, Journal Year: 2024, Volume and Issue: 35, P. 101205 - 101205

Published: April 21, 2024

Language: Английский

Citations

3
Charting the aquaculture internet of things impact: Key applications, challenges, and future trend DOI Creative Commons
Ahmad Fikri Abdullah, Hasfalina Che Man, Mohammed Abdulsalam

et al.

Aquaculture Reports, Journal Year: 2024, Volume and Issue: 39, P. 102358 - 102358

Published: Sept. 20, 2024

Language: Английский

Citations

3
Employing an innovative underwater camera to improve electronic monitoring in the commercial Gulf of Mexico reef fish fishery DOI Creative Commons

Carole L. Neidig,

Max Lee,

Genevieve Patrick

et al.

PLoS ONE, Journal Year: 2024, Volume and Issue: 19(3), P. e0298588 - e0298588

Published: March 8, 2024

Vessel electronic monitoring (EM) systems used in fisheries around the world apply a variety of cameras to record catch as it is brought on deck and during fish processing activities. In EM work conducted by Center for Fisheries Electronic Monitoring at Mote (CFEMM) Gulf Mexico commercial reef fishery, there was need improve upon current technologies enhance camera views accurate species identification large sharks, particularly those that were released while underwater vessel side or underneath hull. This paper describes how this problem addressed with development first known system integrated (UCAM) specialized vessel-specific deployment device bottom longline vessel. Data are presented based blind video reviews from CFEMM trained reviewers resulting UCAM footage compared only overhead 68 gear retrievals collected eight fishing trips. Results revealed successful tool capturing clear (>2m) sharks enable individual identification, determination, fate 34.4%. important obtaining data incidental catches protected shark species. It also provided imagery presence potential predators such marine mammals close vessel, more specifically bottlenose dolphin ( Tursiops truncatus ) retrieval, which often damaged removed catch. information intended assist researchers gathering critical bycatch proximity conventional limited.

Language: Английский

Citations

2
Adaptive density guided network with CNN and Transformer for underwater fish counting DOI Creative Commons

Shijian Zheng,

Rujing Wang,

Shitao Zheng

et al.

Journal of King Saud University - Computer and Information Sciences, Journal Year: 2024, Volume and Issue: 36(6), P. 102088 - 102088

Published: June 8, 2024

Accurate assessment of high-density underwater fish resources is vital to the aquaculture industry. It directly related formulation fishery insurance strategies and implementation breeding plans. However, accurately counting in environments becomes challenging due uneven distribution density individual fish's different sizes postures. To break through this technical bottleneck, we developed an advanced adaptive density-guided network. In detail, first all, network adopts a multi-layer feature fusion structure similar UNet, which significantly enhances matching between targets scales pyramid levels, effectively alleviating problems caused by scale changes morphological deformations. Secondly, also introduces selection module, can intelligently judge applicability Convolutional Neural Network Transformer blocks areas, thereby achieving robust information transfer interaction blocks. Finally, verify effectiveness method, specially constructed two data sets: simulated image set (SHUFD) real (RHUFD). The proposed method has significant improvements over state-of-the-art (CUT) on SHUFD RHUFD datasets, with mean absolute error, square background region bias, foreground bias map indicators improving 3.44% 6.47%, 11.43% 4.41%, 23.91% 29.48%, 4.43% 10.33%, 8.3% 13.14%, respectively.

Language: Английский

Citations

2
A hands‐on guide to use network video recorders, internet protocol cameras, and deep learning models for dynamic monitoring of trout and salmon in small streams DOI Creative Commons
Konrad Karlsson

Ecology and Evolution, Journal Year: 2024, Volume and Issue: 14(5)

Published: May 1, 2024

Abstract This study outlines a method for using surveillance cameras and an algorithm that calls deep learning model to generate video segments featuring salmon trout in small streams. automated process greatly reduces the need human intervention surveillance. Furthermore, comprehensive guide is provided on setting up configuring equipment, along with instructions training tailored specific requirements. Access data knowledge about models makes monitoring of dynamic hands‐on, as collected can be used train further improve models. Hopefully, this setup will encourage fisheries managers conduct more equipment relatively cheap compared customized solutions fish monitoring. To make effective use data, natural markings camera‐captured individual identification. While speeds initial sorting detection fish, manual identification based still requires effort involvement. Individual encounter hold many potential applications, such capture–recapture relative abundance models, evaluating passages streams hydropower by spatial recaptures, is, same identified at different locations. There much gain technique camera captures are better option fish's welfare less time‐consuming physical tagging.

Language: Английский

Citations

1
Deep Learning Techniques for Enhanced Underwater Remote Sensing: Applications in Marine Biodiversity and Infrastructure Inspection DOI

Ayush Kumar Ojha

Journal of Image Processing and Intelligent Remote Sensing, Journal Year: 2024, Volume and Issue: 11, P. 11 - 22

Published: June 27, 2024

Underwater remote sensing has become an essential tool for marine biodiversity studies and underwater infrastructure inspection. However, the unique challenges posed by environments, such as light absorption, scattering, low visibility, necessitate advanced image processing techniques. This research explores application of deep learning methods tailored specifically interpreting images videos. By leveraging convolutional neural networks (CNNs), generative adversarial (GANs), other state-of-the-art architectures, this study aims to enhance clarity, accuracy, interpretability imagery. The proposed focus on several key areas: improving quality through noise reduction color correction, object detection classification species identification, anomaly We conducted extensive experiments using diverse datasets evaluate performance these deep-learning models. results demonstrate significant improvements in enhancement, accurate identification species, reliable structural anomalies. provides valuable insights into integration with sensing, offering potential advancements monitoring maintenance infrastructure. findings highlight transformative artificial intelligence overcoming limitations traditional techniques, paving way more effective efficient exploration conservation efforts.

Language: Английский

Citations

1
A method for custom measurement of fish dimensions using the improved YOLOv5-keypoint framework with multi-attention mechanisms DOI Creative Commons

Danying Cao,

Cheng Guo, Mijuan Shi

et al.

Water Biology and Security, Journal Year: 2024, Volume and Issue: unknown, P. 100293 - 100293

Published: Sept. 1, 2024

Language: Английский

Citations

1
Applications of Computer Vision, 2nd Edition DOI Open Access
Eva Cernadas

Electronics, Journal Year: 2024, Volume and Issue: 13(18), P. 3779 - 3779

Published: Sept. 23, 2024

Computer vision (CV) is a broad term mainly used to refer processing image and video data [...]

Language: Английский

Citations

1