Comparative Analysis of Deep Learning based models for Fresh Water Algae Identification

Abstract

Freshwater algae are fundamental to aquatic ecosystem functioning and are widely recognised as sensitive indicators of water quality and ecological change. Conventional algae identification methods based on manual microscopic examination are labour-intensive, time-consuming, and constrained by taxonomic subjectivity which restricts their scalability for widespread environmental monitoring. This study presents a comparative evaluation of deep learning based object detection models YOLO, Single Shot MultiBox Detector (SSD), and Faster Region-Based Convolutional Neural Networks (Faster R-CNN) for freshwater algae identification.

The analysis is conducted using secondary image data referenced from publicly accessible and licensed repositories. Model performance is evaluated using robust statistical measures, including precision, recall, F1-score, and mean Average Precision (mAP). The results demonstrate statistically significant differences among the models, with Faster R-CNN consistently achieving superior detection accuracy and robustness, particularly in handling morphological variability and overlapping algal structures. YOLO exhibits competitive performance with improved computational efficiency, while SSD shows comparatively lower accuracy and higher variability.