ArTS: Arabic Traffic Sign Dataset

Published: 7 March 2020| Version 1 | DOI: 10.17632/4tznkn45mx.1
Contributors:
Ghazanfar Latif,
Jaafar Alghazo,
Danyah A. Alghmgham,
Loay Alzubaidi

Description

A new dataset for Arabic Traffic Signs is developed for the selected most common 24 Arabic traffic signs. The dataset consists of 2,718 real captured images and 57,078 augmented images for 24 Arabic traffic signs. The images are captured from three connected cities (Khobar, Dammam and Dhahran) in the Eastern Province of Saudi Arabia. The newly developed dataset consisting of 2,718 real images is randomly partitioned into 80% percent training set (2,200 images) and 20% percent testing set (518) images. Augmented dataset of 57,078 images with 10,878 images for testing and 46,200 images for testing. Due to large file size, the Augmented training dataset is uploaded as two compressed files.

Files