Sorokh-Poth: A Balanced Bangladeshi Road Vehicle Image Dataset

Published: 11 April 2023| Version 1 | DOI: 10.17632/7xvcvxgphb.1
Nazmul Hasan,
, Sabiul Islam, Istiak Ahmad


The gigantic vehicle image data on the Internet can potentially promote more advanced object detection and classification models and algorithms. But organized, balanced, and valuable dataset remains a critical problem. We developed a new comprehensive Bangladesh road transport-based balanced image dataset called "Sorokh-Poth" is proposed, which harmonious with several CNN-based architectures such as YOLO, VGG-16, R-CNN, and DPM. Most of the dataset images were collected from a smartphone. The dataset comprises 9809 labeled and annotated images of 10 categories of vehicle images like Auto-rickshaw, bike, bus, bicycle, car, CNG, leguna, rickshaw, truck, and van.



Daffodil International University


Computer Science, Artificial Intelligence, Computer Vision, Image Processing, Annotation, Image Classification, Neural Network