14,904 cargo X-ray scans with 14,797 bounding boxes, gathered from 3 public X-ray security datasets and relabelled into one taxonomy, in COCO format, with per-image provenance and licence. The baggage scans live in the sibling repository FrenchCastle/xray-baggages-customs, with the same layout and taxonomy.

The collection was built for MIRO-VLM, a project that evaluates vision-language models as assistants to customs and security officers: reading a scan, finding threats and checking the goods against the declaration. Each source keeps its own split and its own terms; nothing is pooled across sources without a dataset field to tell them apart.

Licensing in one line. This is a redistribution of third-party data under their original terms, which differ per source; some carry no stated licence upstream. Read Licensing before use and cite the original authors.

Quick start

Download everything (metadata and all image zips), then unpack:

hf download FrenchCastle/miro-xray-cargo --repo-type dataset --local-dir xray_cargo
cd xray_cargo && for z in images_*.zip; do unzip -q "$z" && rm "$z"; done

Or only one source:

hf download FrenchCastle/miro-xray-cargo --repo-type dataset --local-dir xray_cargo \
  --include "annotations/cargox_coco.json" "images_cargox.zip" MANIFEST.csv

Load it with pycocotools (any COCO reader works):

from pycocotools.coco import COCO

coco = COCO("xray_cargo/annotations/cargox_coco.json")
img = coco.loadImgs(coco.getImgIds()[0])[0]  # file_name is relative to xray_cargo/
boxes = coco.loadAnns(coco.getAnnIds(imgIds=img["id"]))
print(img["file_name"], img["split"], [(coco.cats[b["category_id"]]["name"], b["bbox"]) for b in boxes])

The Hub dataset viewer is off: the images ship as zips so that each source can be fetched, cited and removed independently.

Dataset structure

Sources and splits

source	images	splits	boxes	images without boxes	zip
CargoX	12,400	test 12,000, train 400	12,400	0	1.6 GB
CargoXray subset (Roboflow)	659	test 65, train 462, val 132	553	289	86.9 MB
X-ray cargo object detection (Roboflow)	1,845	test 184, train 1,292, val 369	1,844	1	92.2 MB

Splits are the ones each source publishes (or, where a source has none, the one its distributor provides); they are not re-drawn. Use them per source. There is no cross-source test set, and no de-duplication across sources was attempted.

Files

path	content
`annotations/<source>_coco.json`	COCO detection file per source, unified class names
`images_<source>.zip`	images of that source, stored (uncompressed); members are `images/<source>/...`, exactly the COCO `file_name`
`MANIFEST.csv`	one row per image: `image_id, file_name, dataset, domain, split, license, width, height, n_boxes`
`label_report.json`	per source, original label to unified class
`classes_mapping.json`	the normalisation table and the unified class ids

COCO fields

Standard COCO detection, with extra keys:

images[]: id, file_name, width, height, plus dataset (source key), domain (cargo), split (train / val / test) and license (the source's terms).
annotations[]: id, image_id, category_id, bbox ([x, y, width, height], pixels), area, iscrowd, segmentation where the source provides masks or polygons, and attributes.original_label, the label as the source wrote it.
categories[]: the full unified taxonomy, shared by both domain repositories, so class ids are stable across files. Classes absent from a file simply have no boxes.

Images without boxes are negatives (nothing of interest annotated), kept as the source ships them.

Taxonomy

Source labels are mapped to one taxonomy: threats and declarable items (knife_cutter, firearm, battery_powerbank, liquid_container...) and, for cargo, goods categories prefixed cargo_. The original label is always kept in attributes.original_label, so any other grouping can be rebuilt.

Class counts in this repository

class	boxes	images containing it
`knife_cutter`	13,400	13,400
`cargo_normal`	844	844
`cargo_fabrics`	349	267
`cargo_shoes`	67	48
`cargo_auto_parts`	65	47
`cargo_household`	30	23
`cargo_tools`	9	7
`cargo_lamps`	8	5
`cargo_bags`	7	3
`cargo_office_supplies`	6	6
`cargo_bicycle`	5	4
`cargo_toys`	4	4
`cargo_unknown`	2	2
`cargo_wheels`	1	1

Label mapping

Original label to unified class, per source

source	original label	unified class
CargoX	`KNIFE_01_0045`	`knife_cutter`
CargoX	`KNIFE_01_4590`	`knife_cutter`
CargoX	`KNIFE_02_0045`	`knife_cutter`
CargoX	`KNIFE_02_4590`	`knife_cutter`
CargoX	`KNIFE_03_0045`	`knife_cutter`
CargoX	`KNIFE_03_4590`	`knife_cutter`
CargoX	`KNIFE_04_0045`	`knife_cutter`
CargoX	`KNIFE_04_4590`	`knife_cutter`
CargoXray subset (Roboflow)	`---- -----`	`cargo_unknown`
CargoXray subset (Roboflow)	`auto parts`	`cargo_auto_parts`
CargoXray subset (Roboflow)	`bags`	`cargo_bags`
CargoXray subset (Roboflow)	`bicycle`	`cargo_bicycle`
CargoXray subset (Roboflow)	`car weels`	`cargo_wheels`
CargoXray subset (Roboflow)	`clohes`	`cargo_fabrics`
CargoXray subset (Roboflow)	`clothes`	`cargo_fabrics`
CargoXray subset (Roboflow)	`fabrics`	`cargo_fabrics`
CargoXray subset (Roboflow)	`lamps`	`cargo_lamps`
CargoXray subset (Roboflow)	`object`	`cargo_unknown`
CargoXray subset (Roboflow)	`office supplies`	`cargo_office_supplies`
CargoXray subset (Roboflow)	`shoes`	`cargo_shoes`
CargoXray subset (Roboflow)	`spare parts`	`cargo_auto_parts`
CargoXray subset (Roboflow)	`table ware`	`cargo_household`
CargoXray subset (Roboflow)	`table warre`	`cargo_household`
CargoXray subset (Roboflow)	`tableware`	`cargo_household`
CargoXray subset (Roboflow)	`tablware`	`cargo_household`
CargoXray subset (Roboflow)	`tetiles`	`cargo_fabrics`
CargoXray subset (Roboflow)	`texstiles`	`cargo_fabrics`
CargoXray subset (Roboflow)	`textile`	`cargo_fabrics`
CargoXray subset (Roboflow)	`textiles`	`cargo_fabrics`
CargoXray subset (Roboflow)	`tools`	`cargo_tools`
CargoXray subset (Roboflow)	`toys`	`cargo_toys`
X-ray cargo object detection (Roboflow)	`Normal-Case`	`cargo_normal`
X-ray cargo object detection (Roboflow)	`Sharp Object`	`knife_cutter`

Licensing

There is no single licence: each image keeps its source's terms, recorded in the license field of the COCO image and in MANIFEST.csv. Source datasets states the terms as each source publishes them. When in doubt, the upstream terms prevail, and anything beyond non-commercial research needs the source authors' permission. The annotation conversion and the label mapping added here are released under CC BY 4.0.

If you are an author of one of these datasets and want your data removed or its terms described differently, open a discussion on this repository and it will be handled promptly.

Source datasets

CargoX (`cargox`)

Synthetic cargo scans from Ewha Womans University: one of 8 knife types (4 shapes x 2 orientation ranges) composited into 768x768 crops of real container scans. One object per image, box and mask.

Upstream: https://huggingface.co/datasets/Thanaporn09/CargoX
Paper: https://doi.org/10.1371/journal.pone.0272961
Terms: No licence stated on the upstream repo (gated, auto-approved); the PLOS ONE paper states the data are freely available. Cite Viriyasaranon et al. 2022.
Caveats: Synthetic threat insertion; no negative images. This repo holds a class-balanced subset of the 64,000 upstream images, not the full set.

CargoXray subset (Roboflow) (`roboflow_cargo_goods`)

Real side-on scans of trucks and railcars with goods regions (textiles, shoes, auto parts, tableware...), a subset of ISSAI's CargoXray re-uploaded to Roboflow Universe.

Upstream: https://universe.roboflow.com/lin-jhhi5/cargo-x-ray-images
Terms: CC BY 4.0 as asserted by the Roboflow uploader. The upstream CargoXray release (ISSAI, Nazarbayev University) carries no licence.
Caveats: Raw labels are noisy (typos such as 'clohes', 'table warre'); they are normalised to cargo_* classes. Many images have no box.

X-ray cargo object detection (Roboflow) (`roboflow_cargo_threat`)

Cargo scans with two labels, Sharp Object and Normal-Case.

Upstream: https://universe.roboflow.com/container-xray/x-ray-cargo-object-detection-lrynu
Terms: CC BY 4.0 (Roboflow Universe).
Caveats: Provenance undocumented, probably derived from CargoX; the Roboflow export includes augmented copies, so near-duplicates exist across its splits.

Dataset creation

Curation rationale. Public X-ray security datasets are scattered across Google Drive, Baidu, Kaggle, Roboflow and Hugging Face, in half a dozen annotation formats and label vocabularies. Evaluating a model across them first needs one format, one taxonomy and exact provenance for every image.

Processing. Each source was downloaded from its official or documented distribution, converted to COCO (from VOC XML, YOLO txt, paired txt or COCO), and its labels mapped to the unified taxonomy through a fixed table (classes_mapping.json). Images are the downloaded files, byte for byte: this repository does not resize, recompress or filter them (a distributor upstream may have, as the caveats note). Large sources may be a class-balanced subset, as stated in their caveats. Splits are kept as distributed.

Annotations. All boxes come from the source datasets (see each paper for its annotation protocol). This repository adds no new manual annotation; it renames labels and records the original.

Personal and sensitive information. The images are X-ray transmission scans of bags and cargo; they contain no faces, names or documents. Some sources are real stream-of-commerce scans from airports or subway stations; none is known to include personal data.

Considerations for use

Intended use. Research on detection and understanding of objects in X-ray security imagery: benchmarking detectors and vision-language models, studying domain shift between scanners and between baggage and cargo, and building tools that assist human inspectors.

Out of scope. Certifying or operating a screening system; any use that would help conceal items from X-ray inspection; commercial use of sources whose terms forbid it.

Known biases and limitations.

Scanners, colour palettes and resolutions differ per source, and a model can learn the source instead of the object. Evaluate per source, or across sources on purpose.
Class balance follows the sources: some classes come from a single source (and therefore a single scanner).
Some sources are staged (items packed for the dataset) or synthetic (threats composited into real scans); these are not stream-of-commerce distributions.
Near-duplicates exist in some sources (augmented copies, dual views); see caveats.
Label granularity is reduced by the mapping (for example, every blade type becomes knife_cutter); use attributes.original_label for the finer label.

Citation

Cite the original datasets you use, and this consolidation if it helped:

@misc{chastel2026miroxray_cargo,
  title        = {MIRO X-ray Security Scans (Cargo): a unified COCO consolidation of public X-ray datasets},
  author       = {Chastel, Fran\c{c}ois},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/FrenchCastle/miro-xray-cargo}}
}

@article{viriyasaranon2022mfanet,
  title   = {{MFA-net}: Object detection for complex X-ray cargo and baggage security imagery},
  author  = {Viriyasaranon, Thanaporn and Chae, Seung-Hoon and Choi, Jang-Hwan},
  journal = {PLOS ONE},
  volume  = {17},
  number  = {9},
  pages   = {e0272961},
  year    = {2022},
  doi     = {10.1371/journal.pone.0272961}
}

@misc{roboflow_cargo_xray_images,
  title        = {cargo x-ray images Dataset},
  author       = {Lin},
  howpublished = {\url{https://universe.roboflow.com/lin-jhhi5/cargo-x-ray-images}},
  note         = {Roboflow Universe; subset of IS2AI CargoXray, https://github.com/IS2AI/cargoxray}
}

@misc{roboflow_xray_cargo_object_detection,
  title        = {X-ray cargo object detection Dataset},
  author       = {{Container Xray}},
  howpublished = {\url{https://universe.roboflow.com/container-xray/x-ray-cargo-object-detection-lrynu}},
  note         = {Roboflow Universe}
}

Maintenance

Maintained by François Chastel (FrenchCastle). Report problems, missing credits or licence questions in the Community tab of this repository.

Downloads last month: 11

Total file size:

1.79 GB

MIRO X-ray Security Scans: Cargo