Nepal's Embossed License Plate and Character Set Dataset (Bagmati Province, Four-Wheeler)
Photographs of 167 embossed Nepali license plates from Bagmati Province, covering 83 distinct registrations, together with 1,057 individually cropped and labelled alphanumeric characters segmented from them.
- Labelled characters
- 1,057
- Plate photographs
- 167
- Distinct registrations
- 83
- License
- CC BY 4.0
Nepali license plate data for recognition and OCR research
Nepal has moved to embossed, machine-readable license plates, but public training data for them is close to nonexistent, which makes recognition work on Nepali plates hard to reproduce or compare. This release is the underlying data collected for the AA-LPCR study, published so that others can train, benchmark, and stress-test plate recognition on real plates instead of synthetic stand-ins.
The archive holds 167 plate photographs covering 83 distinct registrations from Bagmati Province, Nepal, together with 1,057 alphanumeric characters cropped from those plates and labelled individually. The character set is the part most directly usable for classification: one glyph per image, as it appears in the field, with the dirt, glare, and wear that real plates carry.
Because the plates were photographed on vehicles in traffic rather than staged, the images keep the conditions that break production systems. That is what makes the set useful for robustness work, including the adversarial-tampering experiments the accompanying paper reports.
Files
How to cite
Cite the dataset with its concept DOI, which always resolves to the newest version. If you use the method as well as the data, cite the paper too.
How to cite
@dataset{shrestha2026nepalplates,
author = {Shrestha, Bishal},
title = {Nepal's Embossed License Plate and Character Set Dataset
(Bagmati Province, Four-Wheeler)},
year = {2026},
publisher = {Zenodo},
version = {1.0.0},
doi = {10.5281/zenodo.22861165},
url = {https://doi.org/10.5281/zenodo.22861165}
}