This repository contains the fine-tuned model proposed in the paper "Identity Document to Selfie Face Matching Across Adolescence".
The model can be download here. It is on MXNet format.
The cleaned Public-IvS that is used in the paper can be found here.
To extract features using the model provided, use insightface_feature_extraction.py.
And to match features, use mult_feature_match_list.py or feature_match_efficient.py (much faster, but requires a lot of memory).
Scripts to plot results can be found here.
You will need a copy of the InsighFace repo.
Point the insightface_feature_extraction.py to the folder deploy inside of the InsighFace repo.
If you use our model, please cite the paper below.
@misc{albiero2019identity,
title={Identity Document to Selfie Face Matching Across Adolescence},
author={Vítor Albiero and Nisha Srinivas and Esteban Villalobos and Jorge Perez-Facuse and Roberto Rosenthal and Domingo Mery and Karl Ricanek and Kevin W. Bowyer},
year={2019},
eprint={1912.10021},
archivePrefix={arXiv}
}
If you use our cleaned version of the Public-IvS, please cite both the paper above and the original paper below.
@article{ivs,
title={Large-scale bisample learning on id versus spot face recognition},
author={Zhu, Xiangyu and Liu, Hao and Lei, Zhen and Shi, Hailin and Yang, Fan and Yi, Dong and Qi, Guojun and Li, Stan Z},
journal={International Journal of Computer Vision},
year={2019},
}