Implementasi Model Deep Learning Untuk Deteksi Objek Candi Prambanan, Candi Borobudur, Dan Candi Ratu Boko Menggunakan YOLO V5

Levina, Anora (2021) Implementasi Model Deep Learning Untuk Deteksi Objek Candi Prambanan, Candi Borobudur, Dan Candi Ratu Boko Menggunakan YOLO V5. Undergraduate Thesis thesis, Institut Teknologi Telkom Purwokerto.

	Text 1. COVER.pdf Download (416kB)
	Text 2. ABSTRACT.pdf Download (27kB)
	Text 3. ABSTRAK.pdf Download (28kB)
	Text 4. BAB I.pdf Download (103kB)
	Text 5. BAB II.pdf Download (714kB)
	Text 6. BAB III.pdf Download (1MB)
	Text 7. BAB IV.pdf Restricted to Registered users only Download (2MB) | Request a copy
	Text 8. BAB V.pdf Download (30kB)
	Text 9. DAFTAR PUSTAKA.pdf Download (106kB)
	Text 10. LAMPIRAN.pdf Restricted to Registered users only Download (174kB) | Request a copy

Abstract

The Covid-19 pandemic that has spread in Indonesia has resulted in a rapid decline in the Indonesia's foreign exchange in the tourism sector and temple tourism visits in the Yogyakarta area and its surroundings, which is one of the well-known tourist destinations, have also greatly decreased. To overcome this, digital tourism is needed in the form of a temple object detection system to increase interest in post-pandemic tourist visits. Object detection is one of the applications of computer vision where the trained AI system can perform certain tasks. In this study, the main task carried out by the AI system created is to detect temple objects which consist of 3 classes, namely Prambanan Temple, Borobudur Temple, and Ratu Boko Temple which are located in Yogyakarta and its surroundings. The research was conducted with the aim of obtaining the best performance of the YOLOv5 model for detecting temple objects. The AI project cycle is used as the stages and methods used in this research. The evaluation parameters used are confusion matrix, mean Average Precision (mAP), Precision, Recall and Accuracy. Comparisons were made on the 25, 50, 75, and 100 epochs with a combination of batch sizes 16, 32, and 64. The data preprocessing and training model were carried out using Roboflow and Google Colab. The experimental results show that the best model is achieved in the scenario of epoch 25 in batch size 16 with a Mean Average Precision (mAP) of 0.955, precision is 0.90, and recall is 0.947 with accuracy of testing data is 78.5%. Keywords: Object Detection, YOLOv5, Deep Learning, Artificial Intelligence, Computer Vision

Actions (login required)

View Item