Optimized bilateral segmentation deep learning network for wild animal detection

Srikanth Busa, Johnwesily Chappidi, Gunakala Archana, Anitha Rani Palakayala, Guggilam Venkata Vidya Lakshmi, Khasim Syed and Shaik Salma Asiya Begum

PLOS ONE, 2026, vol. 21, issue 10, 1-25

Abstract: The use of Deep learning (DL) for recognition of wild animals has become an important development in wildlife monitoring. The advances will include the use of multimodal data for improved accuracy and development of lightweight models for use in edge devices. As a result, DL will help contribute to global conservation efforts as it tackles the issues. The key novelty of the proposed framework is the integration of bilateral segmentation with DL to enhance relevant animal feature extraction while reducing the influence of complex background information. The paper proposes the Optimized Bilateral Segmentation DL Network that uses Harris Hawks Optimization (HHO) to improve Convolutional Neural Network (CNN) parameters for wild animal identification. The first step in the process was collecting datasets from various online sources. The image quality was improved through image processing methods such as Adaptive Histogram Equalization (AHE) as well as median filtering. The classifier was trained and validated using pre-processed images. This classifier combines the Bilateral Segmentation Deep Learning Network (BSDLN) with the HHO algorithm. CNN is utilized for detection and the weight parameter is utilized by HHO. In order to enhance model convergence also with performance, HHO dynamically adjusts the learning rate during training within the BSDLN. This balances speed and accuracy by adapting to gradients and losses behavior. By adaptively adjusting the learning rate, HHO and BSDLN optimize the CNN learning process and enhance convergence and classification performance. The suggested approach is put into practice in Python, and statistical metrics are used to assess its effectiveness. The results are compared with other approaches to verify the strategy’s efficacy. With just 0.2% FPR and 0.3% FNR, the suggested BSDLN performed the best overall, achieving 99.4% precision, 99.8% accuracy, 99.6% recall, 99.5% F1-score, and 99.3% specificity. The experimental findings show that the combination of HHO and bilateral segmentation enhances the efficacy of wild animal detection, offering a strong foundation for automated wildlife monitoring.

Date: 2026
References: Add references at CitEc
Citations:

Downloads: (external link)
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0360422 (text/html)
https://journals.plos.org/plosone/article/file?id= ... 60422&type=printable (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:plo:pone00:0360422

DOI: 10.1371/journal.pone.0360422

Access Statistics for this article

More articles in PLOS ONE from Public Library of Science
Bibliographic data for series maintained by plosone ().