A Step Toward Instrument-Agnostic Plankton Classification in the Shared Label Space of IFCB and FlowCam Data

This project focuses on automating the classification of plankton images into 10 distinct classes, which is essential for marine research and biodiversity monitoring. The system leverages EfficientFuzeNet, a hybrid deep learning model that combines EfficientNetB0 for transfer learning with custom layers tailored to plankton identification. The model uses transfer learning from ImageNet and fine-tunes its performance using plankton images collected from publicly available datasets. The image dataset includes various plankton species, such as Acantharia Protist, Copepod Calanoid, and Trichodesmium Puff. The system employs a two-phase training strategy—initially freezing the EfficientNetB0 base layers and later fine-tuning them for enhanced accuracy. The output layer uses softmax activation for multi-class classification. This model, once trained, offers a robust, automated solution for real-time plankton classification, providing valuable insights into marine ecosystems. The project aims to reduce the manual effort traditionally required for identifying plankton species while improving classification speed and accuracy. The tool is expected to serve in a variety of ecological research applications, including marine biodiversity monitoring, conservation efforts, and oceanography studies.

Keywords:

Plankton Classification , Deep Learning , EfficientFuzeNet , Transfer Learning, EfficientNetB0 , Multi-Class Classification , CNN (Convolutional Neural Network) , Marine Research , Biodiversity Monitoring , Image Recognition