Analisis Modifikasi Algoritma YOLO dengan Implementasi Convolutional Block Attention Module (CBAM) terhadap Performa Deteksi Penyu di Lingkungan Bawah Laut

Juliantono, Fadly Rachman Drajad (2024) Analisis Modifikasi Algoritma YOLO dengan Implementasi Convolutional Block Attention Module (CBAM) terhadap Performa Deteksi Penyu di Lingkungan Bawah Laut. Other thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 5027201038-Undergraduate_Thesis.pdf] Text
5027201038-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only

Download (4MB) | Request a copy

Abstract

Eksplorasi biota laut di seluruh dunia, termasuk Indonesia, memainkan peran penting dalam memahami keanekaragaman hayati dan ekosistem bawah air. Indonesia, bagian dari segitiga karang dunia, memiliki keragaman spesies yang luar biasa, namun penelitian terhadap potensi ini masih terbatas. Penyu, reptil laut yang telah ada selama jutaan tahun, sering menjadi daya tarik wisata utama dan memainkan peran penting dalam ekosistem laut. Konservasi penyu menjadi perhatian penting karena ancaman dari aktivitas manusia dan perubahan iklim. Dengan berkembangnya modernisasi, sektor pariwisata maritim di Indonesia tumbuh pesat, menempatkan tekanan pada ekosistem laut dan menegaskan kebutuhan akan manajemen dan pemantauan yang berkelanjutan. Sistem deteksi objek di lingkungan laut menghadapi tantangan seperti air keruh dan variasi pencahayaan yang mengurangi efektivitas teknologi tradisional. Oleh karena itu, pengembangan model berbasis Deep Learning seperti You Only Look Once (YOLO) menjadi sangat relevan. Penelitian ini mengeksplorasi penerapan YOLO yang dilengkapi dengan Convolutional Block Attention Module (CBAM) dalam pengaruh performa deteksi penyu di lingkungan bawah laut dan bertujuan untuk menguji akurasi dan kinerja model deteksi pada kamera bawah laut yang menggunakan mesin Jetson Nano Dev Kit. Model dengan performa tingkat akurasi tinggi didapatkan oleh YOLOv8n dengan CBAM dengan hasil mAP@0,5: 0,95 senilai 57,1% dan beban Param 8.0M serta FLOPs 8.3. Nilai ini naik sebesar 4% dibandingkan dengan YOLOv8n tanpa CBAM. Dengan performa yang dimiliki, model yang dideploy dapat mendeteksi objek secara real-time dan mendapatkan hingga 10 fps. Dengan pembuatan model ini, dapat digunakan sebagai groundtruth dari sistem deteksi penyu secara akurat dan dapat dikembangkan dengan fitur klasifikasi untuk di masa yang akan datang.
===============================================================================================
Exploration of marine biota worldwide, including Indonesia, plays a crucial role in understanding biodiversity and underwater ecosystems. Indonesia, being part of the world's coral triangle, boasts an extraordinary diversity of species, yet research into this potential remains limited. Sea turtles, marine reptiles that have existed for millions of years, often become major tourist attractions and play a vital role in marine ecosystems. Turtle conservation is a pressing concern due to threats from human activities and climate change. As modernization progresses, the maritime tourism sector in Indonesia is rapidly growing, putting pressure on marine ecosystems and underscoring the need for ongoing management and monitoring. Object detection systems in marine environments face challenges such as turbid water and varying lighting conditions that reduce the effectiveness of traditional technologies. Therefore, the development of Deep Learning-based models like You Only Look Once (YOLO) becomes highly relevant. This research explores the application of YOLO equipped with the Convolutional Block Attention Module (CBAM) to study its impact on the performance of turtle detection in underwater environments and aims to test the accuracy and performance of the detection model on underwater cameras using the Jetson Nano Dev Kit. The model enhanced by YOLOv8n with CBAM achieved high-performance metrics with a mean Average Precision (mAP) @ 0,5 of 57.1% and a parameter load of 8.0M and FLOPs of 8.3, showing a 4% improvement compared to YOLOv8n without CBAM. With its current capabilities, the deployed model can detect objects in real-time at a rate of 10 fps. With the creation of this model, it can serve as the ground truth for accurately detecting turtles and can be further developed with classification features for future applications.

Item Type: Thesis (Other)
Uncontrolled Keywords: Convolutional Block Attention Module (CBAM), You Only Look Once (YOLO), Penyu, Deteksi Objek, Sea Turtle, Object Detection
Subjects: T Technology > T Technology (General) > T385 Visualization--Technique
T Technology > T Technology (General) > T57.5 Data Processing
T Technology > T Technology (General) > T57.62 Simulation
T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models.
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information Technology > 59201-(S1) Undergraduate Thesis
Depositing User: Fadly Rachman Drajad Juliantono
Date Deposited: 02 Aug 2024 03:11
Last Modified: 02 Aug 2024 03:11
URI: http://repository.its.ac.id/id/eprint/112272

Actions (login required)

View Item View Item