Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/143989
Title: Identifikasi Fitoplankton menggunakan Deep Learning dengan Algoritma YOLOv8 Diimplementasikan pada Website
Other Titles: Phytoplankton identification using Deep Learning with the YOLOv8 Algorithm Implemented on the Website
Authors: Jaya, Indra
Iqbal, Muhammad
Lestari, Dea Fauzia
Lotaldy, Alnodio
Issue Date: 2024
Publisher: IPB University
Abstract: Fitoplankton memiliki peran penting dalam ekosistem dan dapat berfungsi sebagai bioindikator kualitas perairan. Genus Bacteriastrum, Chaetoceros, dan Thalassiothrix merupakan diatom yang dominan karena morfologinya yang dapat hidup dalam lingkungan tercemar. Identifikasi fitoplankton menjadi krusial untuk analisis dan mencegah potensi kerusakan ekosistem. Metode konvensional memerlukan waktu dan keahlian observasi karena kemiripan morfologi, sehingga diperlukan metode alternatif yang mudah dan efisien. Oleh karena itu, tujuan penelitian ini menggunakan metode deep learning algoritma YOLOv8 yang diimplementasikan pada website dengan framework Flask untuk menjadi solusi. Pengambilan sampel dilakukan pada dua tempat yaitu Perairan Palabuhanratu dan Perairan Pulau Kelapa Dua dengan aktif dan pasif vertikal. Pelabelan melalui platform Roboflow dan training dilakukan pada Google Colaboratory dengan dua model epoch berbeda yaitu 1000 dan 3000, 16 batch, serta learning rate 0,01. Implementasi website dilakukan dengan framework Flask. Sampel yang diperoleh sebesar 700 dataset yang seimbang dengan genus Bacteriastrum, Chaetoceros, dan Thalassiothrix. Performa model training epoch 3000 menghasilkan akurasi sebesar 97,14%, recall, precision, dan F1-score berturut-turut 0,9855. Model tersebut dapat bekerja dengan baik diimplementasi menggunakan framework Flask setelah diuji.
Phytoplankton has an important role in the ecosystem and serve as a bioindicator of water quality. The genera Bacteriastrum, Chaetoceros, and Thalassiothrix are the dominant diatoms because of their morphology that can live in polluted environments. Pythoplankton identification is crucial for analysis and prevent potential ecosystem damage. Conventional methods require time and observation skills due to morphological similarities, so an easy and efficient alternative method is needed. Therefore, the purpose of this research is to use deep learning YOLOv8 algorithm implemented on websites with Flask framework method can be a solution. Sampling was conducted in two places, namely Palabuhanratu and Kelapa Dua Island Waters with active and passive vertical. Labeling uses the Roboflow platform and training is carried out on Google Colaboratory with two different epoch models, namely 1000 and 3000, 16 batches, and a learning rate of 0.01. Website implementation uses Flask framework. Balanced 700 datasets obtained with the genera Bacteriastrum, Chaetoceros, and Thalassiothrix. The performance of the model from epoch 3000 training results in high accuracy of 97.14%, recall, precision, and F1-score of 0.9855 respectively. This model can work well implemented using the Flask framework after testing.
URI: http://repository.ipb.ac.id/handle/123456789/143989
Appears in Collections:UT - Marine Science And Technology

Files in This Item:
File Description SizeFormat 
Cover.pdf
  Restricted Access
Cover526.35 kBAdobe PDFView/Open
C54190002_Alnodio Lotaldy.pdf
  Restricted Access
Fullteks9.52 MBAdobe PDFView/Open
Lampiran.pdf
  Restricted Access
Lampiran793.4 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.