Emulated-Inc/copymove

Scientific Image Copy-Move Forgery

Research figure images, some containing copy-move forgeries: one or more regions duplicated elsewhere within the same image. Training images carry instance masks of each duplication event as run-length encodings; test images are unlabelled.

Contents

The data tree lives under data/, exactly as the benchmark environment presents it at /app/copymove. Its own description is at data/README.md.

• 4698 files, 6423113757 bytes (6.0 GiB)
• structure hash (sha256 over sorted path\tsize lines): 04e71b6431ce0a04b33e5ce3b7ef2d8110adaa84ceac504e811aaf2ba5bceaa3
• data.manifest.tsv at the repo root lists every file as sha256 size path

Use

from huggingface_hub import snapshot_download

snapshot_download("Emulated-Inc/copymove", repo_type="dataset", revision="REVISION",
                  local_dir="./copymove", allow_patterns="data/*")

Pin revision to a commit sha rather than a branch if you need reproducibility.

Provenance and licence

Derived from the training split of the Recod.ai/LUC Scientific Image Forgery Detection Kaggle competition (https://www.kaggle.com/competitions/recodai-luc-scientific-image-forgery-detection), organised by Recod.AI (Institute of Computing, University of Campinas) with LUC collaborators, whose data is released under CC BY-SA 4.0. Attribution and share-alike apply to any distribution of this derivative; the terms were reviewed on 2026-09-04 and the dataset published on that basis.