Dataset description:
MIIC is a novel dataset of real microscopic images of integrated circuits (ICs), to benchmark the IAD algorithms. The MIIC dataset includes 25,160 normal and 116 anomalous high-resolution IC images obtained by ScanningElectron Microscopy (SEM). The SEM images are taken at the metal layer of a manufactured IC and are in gray-scale with a dimension of 512x512 pixels. For each image containing anomalies, we provide differ-ent types of annotations, including the bounding box and pixel-wise ground truth mask for them, which enables future research toward various computer vision applications.
Dataset Link: Download MIIC.
All data is subject to copyright and may only be used for non-commercial research.
Contact Bihan Wen (bihan.wen@ntu.edu.sg) for any questions.
In case of use, please cite our publication: L. Huang, D. Cheng, X. Yang, T. Lin, Y. Shi, K. Yang, B.-H. Gwee, and B. Wen, "Joint Anomaly Detection and Inpainting for Microscopy Images via Deep Self-Supervised Learning," in Proc. IEEE Int. Conf. Image Processing (ICIP), 2021.
Bibtex:
@inproceedings{huang2021,
author={Huang, Ling and Cheng, Deruo and Xulei, Yang and Tong, Lin and Yiqiong, Shi and Kaiyi Yang and Bah-Hwee, Gwee and Bihan, Wen},
title={Joint Anomaly Detection and Inpainting for Microscopy Images via Deep Self-Supervised Learning},
year={2021},
booktitle={IEEE International Conference on Image processing (ICIP)}
}