ICMarkingSynthetic: A synthetic image dataset of defective integrated circuit markings generated by physics-informed guidelines

ICMarkingSynthetic is a synthetic image dataset of defective integrated circuit markings generated by physics-informed guidelines. The guidelines distinguish surface abrasion and deposition from changes to the marking itself, covering scratch, contamination, missing text, smeared text, and combined defects. Images are accompanied by labels and generation provenance. These models are cause-derived image-formation approximations, not first-principles material or optical simulations. This record contains the dataset archive and its companion generator archive. Reproduction instructions and annotation limitations are documented in the archive READMEs. The record metadata are public; access to both archives is restricted. Reviewers and researchers may request access by contacting [e-mail] and describing their intended use. Access requests are reviewed individually. License scopes: image assets are licensed under CC BY-SA 4.0 with per-source attribution; annotations and provenance are licensed under CC BY 4.0; the generator code is licensed under MIT. These licenses apply to their respective components and are not alternative licenses for every file.