Road Damage Detection Using Tensorflow Object Detection API and Tensorflow js
Deteksi Kerusakan Jalan Menggunakan Tensorflow Object Detection API dan Tensorflow js
DOI:
https://doi.org/10.21070/ups.7215Keywords:
deep learning, road damage, SSD Mobilenet, Tensorflow, TensorflowJSAbstract
Road damage significantly impacts mobility and the economy, with conventional detection methods being timeconsuming and costly. This research develops an automated road damage detection system using the Single Shot Detector (SSD) with the MobileNetV2 architecture and TensorFlow Object Detection API, implemented in a web application. The system detects three types of road damage on asphalt and concrete surfaces using a 2019 dataset. Testing involved varying parameters such as learning rate, batch size, data split and total training steps. The best performance was achieved for model without data augmentation achieved optimal performance with the following parameter combination: a learning rate of 0.1 (total loss: 0.34), a data split ratio of 60:30:10 (total loss: 0.39), 3,000
training steps (total loss: 0.33), and a batch size of 24 (total loss: 0.41). The implementation on a web application facilitates real-time road damage detection with improved accessibility.
Downloads
References
I. Himmati, I. P. Prawesti, and A. A. Brahmantya, “ROADECT: DETEKSI PERSOALAN INFRASTRUKTUR JALAN SEBAGAI SOLUSI STRATEGI DIGITAL DALAM PENYAMPAIAN ASPIRASI MASYARAKAT BERKELANJUTAN,” 2023.
A. N. Utomo and N. Lestari, “APLIKASI DETEKSI KERUSAKAN JALAN RAYA MENGGUNAKAN ALGORITMA K-NN (K-NEAREST NEIGHBOUR) ROAD DETECTION APPLICATION USING K-NN ALGORITHM (K-NEAREST NEIGHBOUR),” vol. 10, no. 1, 2021.
P. R. Aningtiyas, A. Sumin, and S. Wirawan, “Pembuatan Aplikasi Deteksi Objek Menggunakan TensorFlow Object Detection API dengan Memanfaatkan SSD MobileNet V2 Sebagai Model Pra -Terlatih,” Jurnal Ilmiah Komputasi, vol. 19, no. 3, Mar. 2020, doi: 10.32409/jikstik.19.3.68.
P. D. Arnesia, N. A. Pratama, and F. Sjafrina, “APLIKASI ARTIFICIAL INTELLIGENCE UNTUK MENDETEKSI OBJEK BERBASIS WEB MENGGUNAKAN LIBRARY TENSORFLOW JS, REACT JS DAN COCO DATASET,” 2022.
A. Basily, “Road Damage,” https://www.kaggle.com/datasets/alvarobasily/road-damage.
K. Kualitas Buah Salak dengan Transfer Learning Arsitektur VGG and A. Luthfiarta, “VGG16 Transfer Learning Architecture for Salak Fruit Quality Classification,” Jurnal Informatika dan Teknologi Informasi, vol. 18, no. 1, pp. 37–48, 2021, doi: 10.31515/telematika.v18i1.4025.
A. Julianto, A. Sunyoto, D. Ferry, and W. Wibowo, “OPTIMASI HYPERPARAMETER CONVOLUTIONAL NEURAL NETWORK UNTUK KLASIFIKASI PENYAKIT TANAMAN PADI (OPTIMIZATION OF CONVOLUTIONAL NEURAL NETWORK HYPERPARAMETERS FOR CLASSIFICATION OF RICE PLANT DISEASES).” [Online]. Available: https://www.kaggle.com/tedisetiady/leaf-rice-dis-
E. Muningsih, “KOMBINASI METODE K-MEANS DAN DECISION TREE DENGAN PERBANDINGAN KRITERIA DAN SPLIT DATA,” 2022.
Z. Syahputra, “Penerapan SSD-Mobilenet Dalam Identifikasi Jenis Buah Apel,” Indonesian Journal of Education And Computer Science, vol. 1, no. 1, 2023.
C. Anwar, “Deteksi Objek Berbasis Web Menggunakan Tensorflow Js dan Coco Dataset pada Framework React Js,” 2022.
“Klasifikasi Penderita Diabetes menggunakan Algoritma Machine Learning dan Z-Score”.
A. Pratama, “Deteksi Kerusakan Jalan Menggunakan Convolutional Neural Network Berbasis SSD EfficientNet Artikel Ilmiah,” 2020.
B. Sasmito, B. H. Setiadji, and R. Isnanto, “Deteksi Kerusakan Jalan Menggunakan Pengolahan Citra Deep Learning di Kota Semarang,” TEKNIK, vol. 44, no. 1, pp. 7–14, May 2023, doi: 10.14710/teknik.v44i1.51908.
Downloads
Additional Files
Posted
License
Copyright (c) 2025 UMSIDA Preprints Server
This work is licensed under a Creative Commons Attribution 4.0 International License.
