prikaz prve stranice dokumenta Usporedba performansi YOLOv5 i YOLOv8 algoritama u zadacima klasifikacije
  Download
PDF 2.91 MB
undergraduate thesis
Usporedba performansi YOLOv5 i YOLOv8 algoritama u zadacima klasifikacije
2024. urn:nbn:hr:166:073760
Drašković, Ema Andrea
University of Split
Faculty of Science
Department of Informatics

Cite this document

Drašković, E. A. (2024). Usporedba performansi YOLOv5 i YOLOv8 algoritama u zadacima klasifikacije (Undergraduate thesis). Split: University of Split, Faculty of Science. Retrieved from https://urn.nsk.hr/urn:nbn:hr:166:073760

Drašković, Ema Andrea. "Usporedba performansi YOLOv5 i YOLOv8 algoritama u zadacima klasifikacije." Undergraduate thesis, University of Split, Faculty of Science, 2024. https://urn.nsk.hr/urn:nbn:hr:166:073760

Drašković, Ema Andrea. "Usporedba performansi YOLOv5 i YOLOv8 algoritama u zadacima klasifikacije." Undergraduate thesis, University of Split, Faculty of Science, 2024. https://urn.nsk.hr/urn:nbn:hr:166:073760

Drašković, E. A. (2024). 'Usporedba performansi YOLOv5 i YOLOv8 algoritama u zadacima klasifikacije', Undergraduate thesis, University of Split, Faculty of Science, accessed 04 October 2024, https://urn.nsk.hr/urn:nbn:hr:166:073760

Drašković EA. Usporedba performansi YOLOv5 i YOLOv8 algoritama u zadacima klasifikacije [Undergraduate thesis]. Split: University of Split, Faculty of Science; 2024 [cited 2024 October 04] Available at: https://urn.nsk.hr/urn:nbn:hr:166:073760

E. A. Drašković, "Usporedba performansi YOLOv5 i YOLOv8 algoritama u zadacima klasifikacije", Undergraduate thesis, University of Split, Faculty of Science, Split, 2024. Available at: https://urn.nsk.hr/urn:nbn:hr:166:073760

Please login to the repository to save this object to your list.
Metadata
TitleUsporedba performansi YOLOv5 i YOLOv8 algoritama u zadacima klasifikacije
Title (english)Performance comparison of YOLOv5 and YOLOv8 algorithms in classification tasks
AuthorEma Andrea Drašković
MentorSaša Mladenović (mentor)
MentorDino Nejašmić (pomoć pri izradi, neposredni voditelj)
Committee memberDino Nejašmić (predsjednik povjerenstva)
Committee memberSaša Mladenović (član povjerenstva)
Committee memberNika Jerković (član povjerenstva)
GranterUniversity of Split
Faculty of Science
(Department of Informatics)
Split
Defense date and country2024-09-16, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Abstract
Umjetna inteligencija predstavlja široko interdisciplinarno područje informatike koje stvara sustave sposobne demonstrirati ljudske vještine poput percepcije, učenja i rješavanja problema. Rad uvodi u koncepte ključne za razvoj AI sustava, kao strojno učenje i duboko učenje. Posebno se ističe YOLO algoritam dubokog učenja za detekciju objekata, razvijen radi brze i precizne analize slika. Pregledavajući arhitekturu varijanti v5 i v8 YOLO modela, ističe se njegova sposobnost prilagodbe različitim zahtjevima te podrška za vlastiti skup podataka. Analiza metrika poput srednje prosječne preciznosti i F1 pruža uvid u evaluaciju performansi modela. Treniranje modela, ključno za postizanje visokih performansi, ističe važnost kvalitetnih podataka i finog podešavanja (engl. fine-tuning) za specifične potrebe aplikacije. Detekcija objekata kroz klasifikaciju i regresiju objašnjava kako YOLO model brzo i precizno detektira objekte u slikama. Kroz primjere primjene, rad ilustrira kako pripremiti skup podataka i trenirati na njemu YOLO model, te uspoređuje različite inačice algoritma preko metrika i performansi.
Abstract (english)
Artificial intelligence represents a broad interdisciplinary field of computer science that creates systems capable of demonstrating human-like skills such as perception, learning and problem solving. The paper introduces key concepts essential for the development of AI systems, such as machine learning and deep learning, where the YOLO algorithm for object detection is notable. By reviewing the architecture of the YOLO v5 and v8 model variants, the focus is on its adaptability to various requirements and support for custom datasets. The analysis of metrics such as mean average precision and F1 provides insight into the evaluation of model performance. Model training, crucial for achieving high performance, emphasizes the importance of quality data and fine-tuning for specific application needs. Object detection through classification and regression explains how YOLO model quickly and accurately detects objects in images. Through practical application, the paper illustrates how to prepare a dataset and train the YOLO model on it, comparing different versions of the algorithm based on metrics and performance.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:166:073760
Study programmeTitle: Computer Science
Study programme type: university
Study level: undergraduate
Academic / professional title: sveučilišni/a prvostupnik/prvostupnica (baccalaureus/baccalaurea) informatike (sveučilišni/a prvostupnik/prvostupnica (baccalaureus/baccalaurea) informatike)
Type of resourceText
File originBorn digital
Access conditionsOpen access
Terms of useLicence In copyright icon
Created on2024-09-18 12:29:14