Synthesized Multi-GFA Multi-bin Wafer Bin Map Dataset

Citation Author(s):
Dong
Ni
Zhejiang University
Yi
Wang
Zhejiang University
Submitted by:
Dong Ni
Last updated:
Thu, 04/21/2022 - 04:59
DOI:
10.21227/s1je-w708
License:
0
0 ratings - Please login to submit your rating.

Abstract 

This dataset contains 1050 multi-pattern multi-bin wafer bin maps (WBMs) synthesized from WM-811K binary WBM dataset and real world multi-bin WBMs using a trained pix2pix model.

Instructions: 

This dataset contains 1050 multi-pattern multi-bin wafer bin maps (WBMs) synthesized from WM-811K binary WBM dataset and real world multi-bin WBMs using a trained pix2pix model. The WM-811K[1] is a public binary WBM dataset and does not contain multiple bin values, therefore we leverage the rich and realistic spatial patterns in WM-811K and our own propertiary multi-bin WBM data and generated this dataset for researchers interested in WBMs to use.

Pix2pix is one of the conditional generation adversarial networks (cGANs)[2]. We trained a pix2pix model using real-life multi-bin WBM (MBWBM) dataset to map the binary WBMs to MBWBMs. The discriminator learns to classify between the fake MBWBM (synthesized by the generator), binarized WBM tuples (binarized from real MBWBMs), and real MBWBM, binarized WBM tuples. Meanwhile, the generator learns to generate MBWBM as real as possible, and misleads the discriminator. After training the pix2pix model, we fed WM-811K binary WBMs to the trained generator to generate single pattern MBWBMs, and then overlay them to obtain the synthesized MBWBMs.

Each MBWBM is given as a .csv file, where die labels are listed according to die location on the wafer.
The classificaiton labels are listed in the label.xlsx file.

References
1. Wu, Ming-Ju, Jyh-Shing R. Jang, and Jui-Long Chen. “Wafer Map Failure Pattern Recognition and Similarity Ranking for Large-Scale Data Sets.” IEEE Transactions on Semiconductor Manufacturing 28, no. 1 (February 2015): 1–12.
2. P. Isola, J. Zhu, T. Zhou and A. Efros, “Image-to-Image Translation with Conditional Adversarial Networks,” CPVR, 2018.