Abstract

Understanding and monitoring aquatic biodiversity is critical for ecologicalhealth and conservation efforts. This paper proposes SuoiAI, an end-to-endpipeline for building a dataset of aquatic invertebrates in Vietnam andemploying machine learning (ML) techniques for species classification. Weoutline the methods for data collection, annotation, and model training,focusing on reducing annotation effort through semi-supervised learning andleveraging state-of-the-art object detection and classification models. Ourapproach aims to overcome challenges such as data scarcity, fine-grainedclassification, and deployment in diverse environmental conditions.

 

Quick Read (beta)