NEU-DET: Hot-Rolled Steel Strip Surface Defect Detection (No Data Mirror -- Adapter Only)

This card describes a DetectionBench dataset adapter for NEU-DET. It does NOT host or redistribute the dataset itself -- no license is stated anywhere for NEU-DET, so there is no redistribution grant to act on. See Getting the Data for the official download links.

Disclaimer

DetectionBench is not an official release of NEU-DET and does not host any NEU-DET images, annotations, or derived files anywhere -- not on Hugging Face, not in this repository.

NEU-DET was created by Kechen Song's group at Northeastern University (China), who retain all rights. Neither the current official homepage nor the dataset's associated papers state a license -- only a citation request. This repository does not claim ownership of any images or annotations.

What this repository provides instead:

1. A DetectionBench dataset adapter that converts an official NEU-DET download into DetectionBench's canonical training layout, once you have obtained the data yourself.
2. A helper command, detectionbench-download-dataset --dataset neudet, that prints the official download links (Google Drive and Baidu Netdisk) and can fetch the Google Drive copy automatically via gdown -- see Getting the Data.
3. This banner, generated locally from a NEU-DET copy already converted through the adapter, purely to illustrate the dataset's domain and annotation style.

---

Dataset Overview

NEU-DET is a hot-rolled steel strip surface-defect detection benchmark: 1,800 grayscale 200x200 images (300 per class) across six common defect types -- crazing, inclusion, patches, pitted surface, rolled-in scale, and scratches -- with bounding-box annotations. It is used to benchmark automated defect localization for steel-manufacturing quality control, complementing GC10-DET's metallic-surface-defect taxonomy with a different steel-inspection domain.

Verified against a real download. The official release (fetched via detectionbench-download-dataset --dataset neudet) ships as a flat, undivided IMAGES/ + ANNOTATIONS/ (Pascal VOC XML) pair, 1,800 images total -- no official train/val/test split. This adapter applies a deterministic seeded 80/10/10 split (same approach as GC10-DET, which is in the same no-official-split situation): train 1,440 / valid 180 / test 180 images, 3,351 / 396 / 442 boxes respectively.

---

Getting the Data

NEU-DET is not mirrored here. Get it directly from the maintainer:

detectionbench-download-dataset --dataset neudet

Resource	Link
NEU-DET dataset (Google Drive)	https://drive.google.com/open?id=1qrdZlaDi272eA79b0uCwwqPrm2Q_WI3k
NEU-DET dataset (Baidu Netdisk, access code pmqx)	https://pan.baidu.com/s/1nBbO-jWDm1_NHDQsc1dRkg

Once downloaded, convert it into DetectionBench's canonical layout:

detectionbench-prepare-coco --dataset neudet --raw-dir <downloaded_dir> --output-dir <coco_out>
detectionbench-convert-coco-to-yolo --input-dir <coco_out> --output-dir <yolo_out>

Note: the original faculty.neu.edu.cn/yunhyan/... URL cited in older papers is dead; the link above is the dataset maintainer's (Kechen Song) current homepage.

Classes (6)

crazing, inclusion, patches, pitted_surface, rolled-in_scale, scratches

---

Dataset Sources

Original Papers

A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects

Kechen Song, Yunhui Yan

Applied Surface Science, 285:858-864, 2013.

An End-to-end Steel Surface Defect Detection Approach via Fusing Multiple Hierarchical Features

Yu He, Kechen Song, Qinggang Meng, Yunhui Yan

IEEE Transactions on Instrumentation and Measurement, 69(4):1493-1504, 2020.

Official Resources

• Homepage: http://faculty.neu.edu.cn/songkc/en/zdylm/263265/list/index.htm

---

License

No license stated anywhere -- the official homepage requests only a citation, with no explicit terms for reuse or redistribution.

Accordingly:

• No Hugging Face mirror of the data is provided or planned.
• The DetectionBench adapter is provided for local, research use against a copy you download yourself.
• If you need broader rights, contact the dataset maintainer (Kechen Song, Northeastern University).

---

Citation

If you use this dataset, please cite:

@article{song2013noise,
  title={A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects},
  author={Song, Kechen and Yan, Yunhui},
  journal={Applied Surface Science},
  volume={285},
  pages={858--864},
  year={2013},
  publisher={Elsevier}
}

@article{he2020end,
  title={An End-to-end Steel Surface Defect Detection Approach via Fusing Multiple Hierarchical Features},
  author={He, Yu and Song, Kechen and Meng, Qinggang and Yan, Yunhui},
  journal={IEEE Transactions on Instrumentation and Measurement},
  volume={69},
  number={4},
  pages={1493--1504},
  year={2020}
}

---

Acknowledgements

We sincerely thank Kechen Song, Yunhui Yan, and their co-authors for creating and publicly releasing this valuable steel-surface-defect benchmark.