FGBNet: A Bio-Subspecies Classification Network with Multi-Level Feature Interaction DOI Creative Commons
Yang Yuan, Danping Huang,

Bingbin Cai

et al.

Diversity, Journal Year: 2025, Volume and Issue: 17(4), P. 237 - 237

Published: March 27, 2025

Biodiversity is a foundation for maintaining ecosystem health and stability, while precise species identification crucial to monitoring protecting ecosystems. Subspecies of organisms, as carriers genetic diversity, play key roles in stability adaptive evolution. Accurate subspecies helps deepen our understanding distribution, ecological relationships, change trends, providing scientific basis effective protection strategies. Therefore, this study proposes FineGrained-BioNet (FGBNet), deep learning network model specifically constructed fine-grained bio-subspecies image classification. The combines detail information supplement module, multi-level feature interaction, coordinate attention (CA) mechanism improve the accuracy efficiency Through experimentation optimization, ConvNeXt selected backbone FGBNet extraction, effectiveness interaction method verified. Additionally, optimal placement CA within also explored. experimental results show that, compared with ConvNeXt-Tiny, achieved an increase 6.204% by increasing parameter quantity only 5.702%, reaching 90.748%. This indicates that significantly improves classification computational efficiency. proposed facilitates more accurate classification, promoting development biodiversity strong technical support conservation.

Language: Английский

FGBNet: A Bio-Subspecies Classification Network with Multi-Level Feature Interaction DOI Creative Commons
Yang Yuan, Danping Huang,

Bingbin Cai

et al.

Diversity, Journal Year: 2025, Volume and Issue: 17(4), P. 237 - 237

Published: March 27, 2025

Biodiversity is a foundation for maintaining ecosystem health and stability, while precise species identification crucial to monitoring protecting ecosystems. Subspecies of organisms, as carriers genetic diversity, play key roles in stability adaptive evolution. Accurate subspecies helps deepen our understanding distribution, ecological relationships, change trends, providing scientific basis effective protection strategies. Therefore, this study proposes FineGrained-BioNet (FGBNet), deep learning network model specifically constructed fine-grained bio-subspecies image classification. The combines detail information supplement module, multi-level feature interaction, coordinate attention (CA) mechanism improve the accuracy efficiency Through experimentation optimization, ConvNeXt selected backbone FGBNet extraction, effectiveness interaction method verified. Additionally, optimal placement CA within also explored. experimental results show that, compared with ConvNeXt-Tiny, achieved an increase 6.204% by increasing parameter quantity only 5.702%, reaching 90.748%. This indicates that significantly improves classification computational efficiency. proposed facilitates more accurate classification, promoting development biodiversity strong technical support conservation.

Language: Английский

Citations

0