Bangla LPDB - A

Bangla LPDB-A is the first complete publicly available dataset of Bangladeshi vehicle images with visible Bangla license plates. The vehicle image dataset consists of almost all types of vehicles present in Bangladesh, excluding army, police, and government vehicles. The images are in JPG, and PNG format. 

This dataset is of two parts-

1. Bangladeshi vehicle images with visible Bangla license plates which includes 1928 images.

2. Cropped Bangla license plate which includes 2662 images (720 synthetic images, 1942 images manually cropped from part-1).

Each image also contains its corresponding annotated text file in YOLO format. For Part-1 and Part-2, number of classes are 1 and 102, respectively.

This work has been accepted in The International Conference on Digital Image Computing: Techniques and Applications (DICTA) (2021). Corresponding paper link- https://ieeexplore.ieee.org/document/9647284 . Contact the authors if you do not have IEEE subscription.

If you use the dataset or paper in any form, cite the corresponding paper (not the dataset): 

IEEE Format: H. H. Shomee and A. Sams, "License Plate Detection and Recognition System for All Types of Bangladeshi Vehicles Using Multi-step Deep Learning Model," 2021 Digital Image Computing: Techniques and Applications (DICTA), Gold Coast, Australia, 2021, pp. 01-07, doi: 10.1109/DICTA52665.2021.9647284.

Bibtex: 
@INPROCEEDINGS{9647284,
  author={Shomee, Homaira Huda and Sams, Ataher},
  booktitle={2021 Digital Image Computing: Techniques and Applications (DICTA)}, 
  title={License Plate Detection and Recognition System for All Types of Bangladeshi Vehicles Using Multi-step Deep Learning Model}, 
  year={2021},
  volume={},
  number={},
  pages={01-07},
  keywords={Deep learning;Image quality;Image segmentation;Image recognition;Image resolution;Computational modeling;License plate recognition;Bangla license plate;Deep learning;Generative adversarial networks;Image processing;Object detection;YOLO},
  doi={10.1109/DICTA52665.2021.9647284}}