Kirana, Naufal Laksana and Dr. Diva Kurnianingtyas, S.Kom and Ir. Indriati, S.T., M.Kom (2024) Perbandingan Kinerja Model YOLOv6 dan YOLOv7 dalam Mendeteksi Sampah Perairan. Sarjana thesis, Universitas Brawijaya.
Abstract
Deep learning merupakan cabang machine learning dengan banyak layer, termasuk You Only Look Once(YOLO)yangmerupakan metode yang digunakan global.YOLOv6 dan YOLOv7 menonjol dengan nilai mean average precision(mAP) tinggi.Penerapan deep learning dapat dilakukan untuk deteksi sampah perairan mengingat perairan Indonesia lebih besar dibanding daratan. Sampah plastik merupakan salah satu jenis sampah terbesar yang mencemari Indonesia. Penelitian diawali dengan studi literatur untuk penggalian informasi,dilanjutkan dengan mengumpulkan data,dilanjutkan dengan merancang model YOLOv6 dan YOLOv7. Setelah model dibuat, dilanjutkan dengan pelatihan model dan dilanjutkandenganevaluasidananalisiskinerjamodel.Penelitiandiakhiridengan penarikan kesimpulan terkait perbandinganmodel. Data yang digunakan pada penelitian ini yakni FloW-IMG yang didapat dengan cararequest pada website orcaboat.Kedua model yang telah dirancang berhasil diimplementasikan dengan dilatih menggunakan data yang telah didapat dan telah dilakukan pemrosesan awal. Hasil penelitian menunjukkan bahwa YOLOv6 dan YOLOv7 mampu melakukan deteksi sampah dengan baik.Didapatkan hasil bahwa YOLOv6 lebih unggul dibanding YOLOv7 pada nilai mAP yakni0,873 untuk YOLOv6 dan 0,512 untuk YOLOv7. YOLOv6 juga memiliki kecepatan deteksi yang lebih tinggi dibanding YOLOv7 yakni4,21ms untuk YOLOv6 dan 13,7ms untuk YOLOv7.
English Abstract
Deep learning is a branch of machine learning with many layers, including You Only Look Once (YOLO) which is a globally used method. YOLOv6 and YOLOv7 stand out with high Mean Average Precision (mAP) values. The application of deep learning can be done for marine debris detection considering Indonesia's water areas are larger than land areas. Plastic waste is one of the largest types of waste polluting Indonesia. The research begins with a literature study to extract information, followed by collecting data,followed by designing the YOLOv6 and YOLOv7 models. After the model is made, continued with model training and continued with evaluation and analysis of model performance. The research ends with drawing conclusions related to model comparison. The data used in this research is FloW-IMG which is obtained by requesting on the orcaboat website. Both models that have been designed are successfully implemented by being trained using data that has been obtained and has been pre-processed. The results showed that YOLOv6 and YOLOv7 were able to perform garbage detection well. It was found that YOLOv6 is superior to YOLOv7 in them AP value which is 0.873 for YOLOv6 and 0.512 for YOLOv7.In addition to them AP value, YOLOv6 also has a higher detection speed than YOLOv7, which is 4.21ms for YOLOv6 and 13.7 for YOLOv7.
| Item Type: | Thesis (Sarjana) |
|---|---|
| Identification Number: | 0524150308 |
| Uncontrolled Keywords: | You Only Look Once (YOLO), FloW-IMG, deep learning, sampah perairan-You Only Look Once (YOLO), FloW-IMG, deep learning, marine debris |
| Divisions: | Fakultas Ilmu Komputer > Teknik Informatika |
| Depositing User: | soegeng sugeng |
| Date Deposited: | 10 Jul 2024 08:23 |
| Last Modified: | 10 Jul 2024 08:23 |
| URI: | http://repository.ub.ac.id/id/eprint/222404 |
![[thumbnail of DALAM MASA EMBARGO]](http://repository.ub.ac.id/style/images/fileicons/text.png) |
Text (DALAM MASA EMBARGO)
Naufal Laksana Kirana.pdf Restricted to Registered users only Download (7MB) |
Actions (login required)
 |
View Item |