Datasets:
Modalities:
Image
Formats:
imagefolder
Size:
< 1K
Tags:
industrial-inspection
anomaly-detection
computer-vision
synthetic-data
defect-synthesis
unsupervised-learning
License:
| license: cc-by-nc-sa-4.0 | |
| task_categories: | |
| - image-classification | |
| - image-segmentation | |
| tags: | |
| - industrial-inspection | |
| - anomaly-detection | |
| - computer-vision | |
| - synthetic-data | |
| - defect-synthesis | |
| - unsupervised-learning | |
| - supervised-learning | |
| - self-supervised-learning | |
| - defect-detection | |
| - mvtec-ad | |
| - industrial-dataset | |
| size_categories: | |
| - 100K<n<1M | |
| modalities: | |
| - images | |
| # ISP-AD: The Industrial Screen Printing Anomaly Detection Dataset | |
|  | |
| The **ISP-AD Dataset** is a large-scale industrial visual anomaly detection benchmark designed for unsupervised, self-supervised, and supervised learning. It features subtle, weakly contrasted surface defects embedded within structured screen-printed patterns with high permitted design variability. | |
| Comprising **559,049** samples, ISP-AD is one of the **largest publicly available industrial anomaly detection datasets to date**, providing both synthetic and real defects collected directly from the factory floor. | |
| It is intended to foster research on **industrially applicable anomaly detection approaches** including: | |
| * Novel **training strategies using real and synthetic defects** as proposed in the accompanying [Paper](https://link.springer.com/article/10.1007/s10845-025-02778-z ), as well as **unsupervised methods**. | |
| * **Zero-shot** or **few-shot defect synthesis** approaches. | |
| * **Emerging vision foundation models** for zero-shot testing or auxiliary fine-tuning for domain adaptation on challenging textured industrial patterns. | |
| --- | |
| ## 📊 Dataset Structure & Modalities | |
| Image patches are provided in **256×256 px** and **512×512 px** resolutions (RGB and grayscale) across **three distinct optical imaging modalities** from a real-world screen-printing process. | |
| --- | |
| ## 📈 Benchmarks | |
| Comprehensive image- and pixel-level evaluation benchmarks for state-of-the-art (SOTA) unsupervised anomaly detection methods—including **EfficientAD**, **U-Flow**, **GLASS**, and **DDAD**—are provided in the accompanying paper. | |
| --- | |
| ## 📂 Dataset Resources | |
| The dataset is publicly available on **Zenodo** under the **CC BY-NC-SA 4.0** license. | |
| Additional information regarding dataset splits, minimal working examples, and evaluation pipelines can be found at: | |
| * **Paper**: [Springer Journal of Intelligent Manufacturing (2026)](https://link.springer.com/article/10.1007/s10845-025-02778-z ) | |
| * **Dataset**: [Zenodo Repository (CC BY-NC-SA 4.0)](https://doi.org/10.5281/zenodo.14911042 ) | |
| * **GitHub Repository**: [p4ulk/isp-ad (PyTorch & Anomalib Pipelines)](https://github.com/p4ulk/isp-ad) | |
| --- | |
| ## 🎓 Citation & Attribution | |
| If you integrate the ISP-AD dataset or the associated code into your scientific research, please cite the primary reference: | |
| ```bibtex | |
| @article{krassnig2026isp, | |
| title={ISP-AD: A large-scale real-world dataset for advancing industrial anomaly detection with synthetic and real defects}, | |
| author={Krassnig, Paul Josef and Gruber, Dieter Paul}, | |
| journal={Journal of Intelligent Manufacturing}, | |
| pages={1--26}, | |
| year={2026}, | |
| publisher={Springer} | |
| } | |
| ``` |