Identifying Mangrove Species using Deep Learning Model and Recording for Diversity Analysis: A Mobile Approach

Abstract

Mangrove forests are essential to the coastal ecosystem and provide numerous ecological and economic benefits. Accurate identification and monitoring of mangrove species are crucial for their conservation and management. However, traditional species identification and documentation methods are often time-consuming, labor-intensive, and require taxonomic expertise. The preservation of mangroves relies heavily on the ability to identify different species and monitor their diversity. This research paper proposes a mobile application for mangrove species identification using MobileNetV3, a deep-learning model and recording for diversity analysis. We collected a dataset of 5,000 images of five mangrove species commonly found in Clarin, Bohol, and trained a deep-learning model using transfer learning. The model MobileNetV3Large achieved an accuracy of 98.4% on a test set of images, indicating that it effectively identifies mangrove species. The trained model was integrated into a mobile application that can capture and identify mangrove leaves using a smartphone camera. The application's user-friendly interface, real-time data recording, and cloud-based architecture make it suitable for large-scale biodiversity monitoring and management, enabling faster and more efficient data collection and analysis.