HRP4K in YOLO format
HRP4K, a high-resolution perspective-view road image dataset for pothole detection, in YOLO format, taken from version 1.01 of the Zenodo record. 2,000 of its images show road without potholes on purpose and have empty label files.
Original dataset: https://doi.org/10.5281/zenodo.21941931. Dataset paper: https://doi.org/10.1038/s41597-026-07317-w. Created by Hanshen Chen, Zhoulin Tu, Yu Zhao and Jianfeng Ye.
| View | Street level, forward-facing |
| Place | Urban and rural roads in Zhejiang Province, China |
| Capture | Vehicle-mounted full-frame mirrorless cameras; faces, licence plates and traffic signs masked by the dataset authors |
| Image size | 3840 x 2160 pixels |
| Images | 6,003 (train 4,203, val 900, test 900) |
| Boxes | 7,217 |
| Configuration | data.yaml |
| Number | Class | Meaning | Train | Val | Test | Total |
|---|---|---|---|---|---|---|
| 0 | pothole | Pothole | 5,259 | 1,037 | 921 | 7,217 |
Use with Ultralytics
from huggingface_hub import snapshot_download
from ultralytics import YOLO
folder = snapshot_download("ashmeera/hrp4k-yolo", repo_type="dataset")
YOLO("yolo26n.pt").train(data=f"{folder}/data.yaml")
Each data.yaml points to its images/train, images/val and images/test folders relative to its own location. Each image has a label file of the same name in the matching labels folder, one line per box: class number, x centre, y centre, width and height, all as fractions of the image size.
Split
The split shipped with the dataset is kept. The dataset authors split by source video, so all frames of one video sit in the same split.
Changes from the original
- The
validfolder is renamed toval. - Boxes that extend past the image edge are cut at the edge.
- Label lines are written with six decimals.
Limitations
- All images come from three cities in Zhejiang Province, China, as the dataset paper notes.
Licence
Shared under CC BY 4.0. The Zenodo record and the dataset paper list CC BY 4.0 for the data. The paper itself is published under CC BY-NC-ND 4.0, which covers the article text only. The original creators provide the data as it is, with no warranty. Please credit the creators and cite the works below.
Citation
@article{Chen_2026,
title = {A high-resolution perspective-view road image dataset for pothole detection},
volume = {13},
number = {1},
journal = {Scientific Data},
publisher = {Springer Science and Business Media LLC},
author = {Chen, Hanshen and Tu, Zhoulin and Zhao, Yu and Ye, Jianfeng},
year = {2026},
doi = {10.1038/s41597-026-07317-w}
}
@misc{Chen_2026_data,
title = {HRP4K},
publisher = {Zenodo},
author = {Chen, Hanshen},
year = {2026},
doi = {10.5281/zenodo.21941931}
}
- Downloads last month
- 54

