Enhancing mosquito classification through self-supervised learning DOI Creative Commons

Ratana Charoenpanyakul,

Veerayuth Kittichai, Songpol Eiamsamang

и другие.

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

Опубликована: Ноя. 7, 2024

Traditional mosquito identification methods, relied on microscopic observation and morphological characteristics, often require significant expertise experience, which can limit their effectiveness. This study introduces a self-supervised learning-based image classification model using the Bootstrap Your Own Latent (BYOL) algorithm, designed to enhance species efficiently. The BYOL algorithm offers key advantage by eliminating need for labeled data during pretraining, as it autonomously learns important features. During fine-tuning, requires only small fraction of achieve accurate results. Our approach demonstrates impressive performance, achieving over 96.77% accuracy in analysis, with minimized both false positives negatives. Additionally, model's overall accuracy, measured area under ROC curve, surpasses 99.55%, highlighting its robustness reliability. A notable finding is that fine-tuning just 10% produces results comparable full dataset. particularly valuable resource-limited settings limited access advanced equipment expertise. provides practical solution identification, overcoming challenges traditional such time-consuming process reliance specialized knowledge healthcare services. Overall, this supports personnel resource-constrained environments facilitating vector density analysis paving way future methodologies.

Язык: Английский

Enhancing mosquito classification through self-supervised learning DOI Creative Commons

Ratana Charoenpanyakul,

Veerayuth Kittichai, Songpol Eiamsamang

и другие.

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

Опубликована: Ноя. 7, 2024

Traditional mosquito identification methods, relied on microscopic observation and morphological characteristics, often require significant expertise experience, which can limit their effectiveness. This study introduces a self-supervised learning-based image classification model using the Bootstrap Your Own Latent (BYOL) algorithm, designed to enhance species efficiently. The BYOL algorithm offers key advantage by eliminating need for labeled data during pretraining, as it autonomously learns important features. During fine-tuning, requires only small fraction of achieve accurate results. Our approach demonstrates impressive performance, achieving over 96.77% accuracy in analysis, with minimized both false positives negatives. Additionally, model's overall accuracy, measured area under ROC curve, surpasses 99.55%, highlighting its robustness reliability. A notable finding is that fine-tuning just 10% produces results comparable full dataset. particularly valuable resource-limited settings limited access advanced equipment expertise. provides practical solution identification, overcoming challenges traditional such time-consuming process reliance specialized knowledge healthcare services. Overall, this supports personnel resource-constrained environments facilitating vector density analysis paving way future methodologies.

Язык: Английский

Процитировано

0