xuekeqing/med_flux2.0

Med Flux 2.0

Dataset Description

Med Flux 2.0 is a collection of publicly available medical datasets organized for medical research, data analysis, and machine learning experiments.

The dataset combines data from multiple public medical data sources. To improve storage and download efficiency, the original files are packaged into TAR shards.

This dataset is intended for research and development purposes only. It must not be used as a substitute for professional medical judgment or as the sole basis for clinical diagnosis or treatment decisions.

Dataset Contents

The dataset may include:

• Medical images
• Classification labels
• Segmentation masks
• Target images
• Metadata
• Dataset annotations
• Supporting files

A typical repository structure is:

.
├── README.md
├── manifest.jsonl.gz
├── shards.json
├── SHA256SUMS
└── shards/
    ├── med-flux2-00000.tar
    ├── med-flux2-00001.tar
    ├── med-flux2-00002.tar
    └── ...

File descriptions:

• shards/*.tar: TAR archives containing the dataset files.
• manifest.jsonl.gz: Mapping between original file paths and TAR shards.
• shards.json: Information about shard sizes and file counts.
• SHA256SUMS: SHA-256 checksums for file integrity verification.
• README.md: Dataset documentation and licensing information.

The original relative directory structure is preserved inside the TAR archives.

Dataset Size

The dataset contains approximately 480,000 original files.

The exact number of files, archive sizes, and shard counts can be found in shards.json.

Data Sources

This dataset was assembled from publicly available medical datasets.

The original data sources should be documented below:

Dataset	Provider	Original License	Source URL	Notes
To be added	To be added	To be added	To be added	To be added

Users must verify the original license, citation requirements, redistribution rules, and terms of use for each source dataset.

Public availability does not automatically permit unrestricted redistribution, commercial use, or relicensing.

Data Processing

The dataset preparation process may include:

1. Collecting publicly available medical datasets.
2. Organizing files into a consistent directory structure.
3. Removing temporary and irrelevant system files.
4. Preserving original relative file paths.
5. Packaging small files into TAR shards.
6. Generating a file-to-shard manifest.
7. Generating shard statistics.
8. Generating SHA-256 checksums.
9. Uploading the processed dataset to Hugging Face.

Unless explicitly stated otherwise, the dataset has not been manually reviewed sample by sample by qualified medical professionals.

Download

Download the dataset with the Hugging Face CLI:

hf download \
  xuekeqing/med_flux2.0 \
  --repo-type dataset \
  --local-dir ./med_flux2

Download it with Python:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="xuekeqing/med_flux2.0",
    repo_type="dataset",
    local_dir="./med_flux2",
)

Verify File Integrity

After downloading the dataset, verify the TAR shards:

cd med_flux2
sha256sum -c SHA256SUMS

Each valid file should be reported as OK.

Extract the Dataset

Extract all TAR shards while preserving the original directory structure:

mkdir -p restored

for shard in shards/*.tar; do
    tar -xf "$shard" -C restored
done

Inspect an archive before extraction:

tar -tf shards/med-flux2-00000.tar | head

Read the Manifest

Each line in manifest.jsonl.gz contains a JSON object:

{
  "path": "example_dataset/train/images/example.png",
  "shard": "shards/med-flux2-00000.tar",
  "size": 123456
}

Example Python code:

import gzip
import json

with gzip.open("manifest.jsonl.gz", "rt", encoding="utf-8") as file:
    for line in file:
        item = json.loads(line)

        print("Original path:", item["path"])
        print("Shard:", item["shard"])
        print("File size:", item["size"])
        break

Intended Uses

Appropriate uses include:

• Medical imaging research
• Computer vision research
• Image classification experiments
• Image segmentation experiments
• Image generation and transformation research
• Medical data preprocessing
• Machine learning model training and evaluation
• Dataset storage and loading research
• Non-clinical academic research
• Educational demonstrations

Users should independently evaluate data quality, labels, source distributions, and task suitability.

Out-of-Scope Uses

This dataset should not be used for:

• Direct clinical diagnosis
• Direct treatment recommendations
• Emergency medical decision-making
• Unsupervised deployment in healthcare environments
• Patient identification or re-identification
• Unauthorized patient profiling
• Discriminatory or unlawful applications
• Uses prohibited by the original source licenses
• Decisions affecting patient health without qualified professional review

Models trained on this dataset must not be assumed to be clinically validated or approved as medical devices.

Privacy and Sensitive Information

Although this dataset was assembled from public medical data, public availability does not guarantee that every file is free of privacy risks.

Reasonable efforts should be made to remove or exclude:

• Patient names
• Patient identifiers
• Government identification numbers
• Contact information
• Home addresses
• Exact dates of birth
• Identifiable metadata
• Protected health information
• Other information that could identify an individual

Because of the size of the dataset, every file may not have been manually inspected.

Users who discover personal information, restricted content, licensing conflicts, or privacy issues should stop distributing the affected files and contact the dataset maintainer.

Limitations

This dataset may contain:

• Inconsistent image quality
• Different acquisition devices and protocols
• Inconsistent annotation standards
• Missing or incomplete labels
• Incorrect labels
• Duplicate or near-duplicate samples
• Class imbalance
• Geographic bias
• Institutional bias
• Demographic bias
• Differences in image resolution and format
• Potential train-test leakage
• Unreviewed or automatically generated annotations

Users should perform their own:

• Data cleaning
• Duplicate detection
• Label validation
• Patient-level split verification
• Bias analysis
• Quality assessment
• External validation

Bias and Fairness

Medical datasets may reflect biases related to:

• Geographic location
• Healthcare institutions
• Imaging equipment
• Clinical workflows
• Patient demographics
• Disease prevalence
• Inclusion and exclusion criteria
• Annotation practices
• Differences in healthcare access

Researchers should evaluate model performance across relevant subgroups whenever possible.

Results obtained from this dataset should not automatically be generalized to other hospitals, countries, devices, populations, or clinical environments.

Independent external validation is strongly recommended.

License

This repository is released under the Apache License 2.0.

Copyright 2026 Med Flux 2.0 Contributors

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this material except in compliance with the License.
You may obtain a copy of the License at:

https://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, material
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

Third-Party Data Notice

The Apache License 2.0 applies only to content for which the repository maintainers have the authority to apply that license, such as:

• Original documentation
• Processing scripts
• Dataset manifests
• Repository organization
• Sharding metadata
• Other original contributions

Third-party medical data included in this repository may remain subject to its original:

• License
• Terms of use
• Citation requirements
• Redistribution restrictions
• Commercial-use restrictions
• Privacy requirements
• Ethical requirements

The Apache License 2.0 declaration for this repository does not automatically relicense third-party data under Apache 2.0.

Users are responsible for confirming that their intended use is permitted by each original data provider.

Citation

When using this dataset, please cite:

1. The Med Flux 2.0 Hugging Face repository.
2. Each original dataset used in the research.
3. The corresponding publications or data providers.

Suggested citation:

@dataset{med_flux_2_2026,
  title        = {Med Flux 2.0},
  author       = {Med Flux 2.0 Contributors},
  year         = {2026},
  publisher    = {Hugging Face},
  repository   = {xuekeqing/med_flux2.0},
  license      = {Apache-2.0}
}

Disclaimer

This dataset is provided on an "AS IS" basis.

No warranties are provided regarding:

• Accuracy
• Completeness
• Medical validity
• Clinical usefulness
• Fitness for a particular purpose
• Correctness of labels or annotations
• Absence of privacy risks
• Compatibility with a particular model or system

This dataset does not constitute medical advice and must not replace evaluation, diagnosis, or treatment by qualified healthcare professionals.

Users are responsible for ensuring compliance with applicable laws, regulations, ethical requirements, institutional policies, source licenses, and privacy obligations.

Maintenance and Issue Reporting

Please report issues involving:

• Corrupted files
• Incorrect checksums
• Missing shards
• Duplicate samples
• Incorrect labels
• Privacy concerns
• Licensing conflicts
• Restricted content
• Missing source attribution
• Files that should not be publicly distributed

Maintainer: @xuekeqing