KLASIFIKASI KERUSAKAN JALAN DI SIDOARJO MENGGUNAKAN CNN BERBASIS ARSITEKTUR INCEPTION RESNET-V2

Zahrah, Fathima (2025) KLASIFIKASI KERUSAKAN JALAN DI SIDOARJO MENGGUNAKAN CNN BERBASIS ARSITEKTUR INCEPTION RESNET-V2. Undergraduate thesis, UNIVERSITAS PEMBANGUNAN NASIONAL VETERAN JAWA TIMUR.

	Text (Cover) 21083010043_Cover.pdf Download (1MB)
	Text (Bab I) 21083010043_Bab 1.pdf Download (207kB)
	Text (Bab II) 21083010043_Bab 2.pdf Restricted to Registered users only until 19 June 2028. Download (463kB)
	Text (Bab III) 21083010043_Bab 3.pdf Restricted to Registered users only until 19 June 2028. Download (440kB)
	Text (Bab IV) 21083010043_Bab 4.pdf Restricted to Registered users only until 19 June 2028. Download (1MB)
	Text (Bab V) 21083010043_Bab 5.pdf Restricted to Registered users only until 19 June 2028. Download (169kB)
	Text (Daftar Pustaka) 21083010043_Daftar Pustaka.pdf Download (169kB)
	Text (Lampiran) 21083010043_Lampiran.pdf Restricted to Registered users only Download (159kB)

Abstract

Road damage is a serious issue in Sidoarjo Regency, posing risks to the safety of road users. This study aims to classify road surface conditions using a Convolutional Neural Network (CNN) model based on the Inception ResNet-V2 architecture. The image-based classification model was developed by combining secondary data obtained through scraping Google Maps Platform APIs and primary data collected manually. The training process involved strategies such as data augmentation, class balancing, early stopping, and model checkpointing. A total of 1,080 images were used in this study and categorized into three classes: potholes, cracks, and undamaged roads. The model was trained for 50 epochs, with early stopping triggered at epoch 29 when the validation accuracy reached 77.71%. Evaluation on the test data showed an accuracy of 76%. The undamaged road class achieved the highest performance with an F1-score of 0.83, followed by the pothole class with an F1-score of 0.78. The lowest performance was observed in the cracked road class, with an F1-score of 0.59, indicating the model’s limitations in detecting fine crack features. These limitations are likely due to class imbalance and visual similarities between classes. Although the model demonstrated good generalization for the two majority classes, the gap between validation and test accuracy highlights the need for improvement in detecting minority classes. Future research is recommended to explore more advanced augmentation techniques, enhance data representation for minority classes, and consider alternative architectures or ensemble methods to improve the model’s sensitivity to subtle road damage features.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Diyasa, I Gede Susrama Mas	NIDN0019067008	igsusrama.if@upnjatim.ac.id
Thesis advisor	Saputra, Wahyu Syaifullah Jauharis	NIDN0725088601	wahyu.s.j.saputra.if@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA76.6 Computer Programming
Q Science > QA Mathematics > QA76.87 Neural computers

Divisions:

Faculty of Computer Science > Departemen of Data Science

Depositing User:

Fathima Zahrah

Date Deposited:

20 Jun 2025 02:31

Last Modified:

20 Jun 2025 02:31

URI:

https://repository.upnjatim.ac.id/id/eprint/38636