Self-supervised deep reconstruction of mixed strip-shredded text documents

作者：

Highlights：

• Paper shreds matching via self-supervised deep learning.

• Training with simulated cuts is effective for real-shredded documents.

• A new public dataset with 100 strip-shredded documents (2292 shreds).

• Accurate (over 90% accuracy) reconstruction of 100 mixed shredded documents.

摘要

•Paper shreds matching via self-supervised deep learning.•Training with simulated cuts is effective for real-shredded documents.•A new public dataset with 100 strip-shredded documents (2292 shreds).•Accurate (over 90% accuracy) reconstruction of 100 mixed shredded documents.

论文关键词：Deep learning,Self-supervised learning,Fully convolutional neural networks,Document reconstruction,Forensics,Optimization search

论文评审过程：Received 17 September 2019, Revised 11 June 2020, Accepted 1 July 2020, Available online 3 July 2020, Version of Record 6 July 2020.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107535