KLASIFIKASI MULTI-CATEGORY SPORTS IMAGE MENGGUNAKAN SWIN TRANSFORMERS

AQILLAH, FARCA RIZQI and Sutarno, Sutarno (2025) KLASIFIKASI MULTI-CATEGORY SPORTS IMAGE MENGGUNAKAN SWIN TRANSFORMERS. Undergraduate thesis, Sriwijaya University.

	Image RAMA_56201_09011282126082_cover.jpg Download (23kB)
	Text RAMA_56201_09011282126082.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (2MB) | Request a copy
	Text RAMA_56201_09011282126082_TURNITIN.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (7MB) | Request a copy
	Text RAMA_56201_09011282126082_0201117802_01_front_ref.pdf - Accepted Version Available under License Creative Commons Public Domain Dedication. Download (5MB)
	Text RAMA_56201_09011282126082_0201117802_02.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (4MB) | Request a copy
	Text RAMA_56201_09011282126082_0201117802_03.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (2MB) | Request a copy
	Text RAMA_56201_09011282126082_0201117802_04.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (1MB) | Request a copy
	Text RAMA_56201_09011282126082_0201117802_05.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (228kB) | Request a copy
	Text RAMA_56201_09011282126082_0201117802_06_ref.pdf - Bibliography Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (199kB) | Request a copy
	Text RAMA_56201_09011282126082_0201117802_07_lamp.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (409kB) | Request a copy

Abstract

This study explores the use of the Swin Transformer for multi-category sports image classification. Through its window-based self-attention architecture and hierarchical structure, it efficiently captures visual features despite variations in athlete poses, image angles, and complex backgrounds. The dataset contains over 13,000 images across 100 sports classes from Kaggle. Nine configurations were tested by varying batch size and learning rate (with 10 fixed epochs) and evaluated using accuracy, precision, recall, along with confusion matrix subset and heatmap. Model 5 (batch size 32, learning rate 0.001) achieved the highest performance with 99% validation accuracy and 0.99 for both precision and recall, followed by Model 7 (batch size 64, learning rate 0.01). The results confirm that the Swin Transformer is effective for complex sports image classification and has potential for further development through hyperparameter tuning and ensemble methods.

Item Type:	Thesis (Undergraduate)
Subjects:	Q Science > Q Science (General) > Q300-390 Cybernetics > Q325.5 Machine learning Q Science > Q Science (General) > Q334-342 Computer science. Artificial intelligence. Algorithms. Robotics. Automation.
Divisions:	09-Faculty of Computer Science > 56201-Computer Systems (S1)
Depositing User:	Farca Rizqi Aqillah
Date Deposited:	16 Sep 2025 07:49
Last Modified:	16 Sep 2025 07:49
URI:	http://repository.unsri.ac.id/id/eprint/184027

Actions (login required)

View Item