Inconsistency-based active learning with adaptive pseudo-labeling for fish species identification DOI
M M Nabi, Chiranjibi Shah, Simegnew Yihunie Alaba

и другие.

Опубликована: Июнь 6, 2024

The deep neural network has found widespread application in object detection due to its high accuracy. However, performance typically depends on the availability of a substantial volume accurately labeled data. Several active learning approaches have been proposed reduce labeling dependency based confidence detector. Nevertheless, these tend exhibit biases toward high-performing classes, resulting datasets that do not adequately represent testing In this study, we introduce comprehensive framework for considers both uncertainty and robustness detector, ensuring superior across all classes. robustness-based score is calculated using consistency between an image augmented version. Additionally, leverage pseudo-labeling mitigate potential distribution drift enhance model performance. To address challenge setting threshold, adaptive threshold mechanism. This adaptability crucial, as fixed can negatively impact performance, particularly low-performing classes or during initial stages training. For our experiment, employ Southeast Area Monitoring Assessment Program Dataset 2021 (SEAMAPD21), comprising 130 fish species with 28,328 samples. results show outperforms state-of-the-art method significantly reduces annotation cost. Furthermore, benchmark model's against public dataset (PASCAL VOC07), showcasing effectiveness comparison existing methods.

Язык: Английский

HafsAuga MobileEM: mobile electronic monitoring for fisheries management and research DOI Creative Commons
Lachlan C. Fetterplace,

Emilia Benavente Norrman,

Kristin Öhman

и другие.

Research Ideas and Outcomes, Год журнала: 2025, Номер 11

Опубликована: Апрель 28, 2025

Electronic monitoring (EM) using video cameras is valuable for documenting fisheries catch and bycatch, but it remains challenging to implement in small-scale fisheries. Current barriers include high costs, technical installation needs limited power supply on small vessels. In addition, as most EM systems the market are difficult quickly move between vessels, they do not allow random data collection, which may be required obtain reliable estimates of bycatch across a fleet. Basic available, designed use fisheries, image-based, have low frame rates always capable recording enough quality identify species with precision. The Swedish fishery consists over 700 boats (under 12 m length), key target including cod, herring, sprat flatfish. To meet requirements gather sufficient machine-learning applications, we created HafsAuga MobileEM: low-cost mobile multi-camera, GPS remote offload system catch, effort. It records (up 60 fps), compact (~ 2 kg) deployable under 30 minutes. Designed simple operate install, modifiable allows users connect vessel's 12v or an internal battery record high-quality footage continuously week. This ideal also well-suited situations where fleets need randomly sampled by moving Here, describe MobileEM outline its Sweden, has been since 2020. date, twenty vessels had mounted them 1000 fishing days successfully recorded. provides innovative new tool potential applications other regions.

Язык: Английский

Процитировано

0
Systems Thinking of Marine Policies in Promoting Environmental Law, Sustainability, and Digital Technologies: Social Challenges in Belt and Road Initiative Countries DOI Creative Commons

WU Xiao-ping,

Muhammad Bilawal Khaskheli

Systems, Год журнала: 2024, Номер 12(10), С. 400 - 400

Опубликована: Сен. 27, 2024

The success of marine environmental regulations in terms social challenges Belt and Road Initiative (BRI) countries is the main subject this study, which compares contrasts them with an eye toward sustainability, integration digital technologies, law, reducing ecological degradation. Environmental solid governance essential as BRI increase their activity, important part world economy by systems thinking; industry includes a broad range operations about ocean its resources through to promote legislation emissions participating BRI. This study evaluated effects institutional quality technical advancements policies between 2013 2024. project aims examine how various policy contexts relate conservation, well they comply international regulations, technology can improve monitoring implementation thinking. determine common obstacles best methods for enforcing examining research from different countries. results deepen our understanding these be utilized meet sustainable development objectives while preventing degradation ecosystems due economic growth business.

Язык: Английский

Процитировано

1
Inconsistency-based active learning with adaptive pseudo-labeling for fish species identification DOI
M M Nabi, Chiranjibi Shah, Simegnew Yihunie Alaba

и другие.

Опубликована: Июнь 6, 2024

The deep neural network has found widespread application in object detection due to its high accuracy. However, performance typically depends on the availability of a substantial volume accurately labeled data. Several active learning approaches have been proposed reduce labeling dependency based confidence detector. Nevertheless, these tend exhibit biases toward high-performing classes, resulting datasets that do not adequately represent testing In this study, we introduce comprehensive framework for considers both uncertainty and robustness detector, ensuring superior across all classes. robustness-based score is calculated using consistency between an image augmented version. Additionally, leverage pseudo-labeling mitigate potential distribution drift enhance model performance. To address challenge setting threshold, adaptive threshold mechanism. This adaptability crucial, as fixed can negatively impact performance, particularly low-performing classes or during initial stages training. For our experiment, employ Southeast Area Monitoring Assessment Program Dataset 2021 (SEAMAPD21), comprising 130 fish species with 28,328 samples. results show outperforms state-of-the-art method significantly reduces annotation cost. Furthermore, benchmark model's against public dataset (PASCAL VOC07), showcasing effectiveness comparison existing methods.

Язык: Английский

Процитировано

0