SlackedFace: Learning a Slacked Margin for Low-Resolution Face Recognition

_{Last Update on November 20, 2023.}

This is the official GitHub repository for "SlackedFace: Learning a Slacked Margin for Low-Resolution Face Recognition". This work has been accepted for an oral presentation at the BMVC 2023.

We provide the main contents along with our supplementary materials here.

YouTube Video Presentation

This YouTube video was prepared in conjunction with the BMVC camera-ready submission.

Supplementary Materials

We provide supplementary materials to enhance the understanding and insights into our work, including:

• Theoretical Analysis
• Benchmarking Datasets
• Hyperparameter Analysis and Configuration
• Stability Analysis
• Model Extension

Implementation Materials

For reproducibility purposes, the distractor sets (with 10K and 20K of low-resolution face images) sampled from that of TinyFace are available at:

• TinyFace Distractor Subset - 10K
• TinyFace Distractor Subset - 20K

The complete distractor set (with 153,428 low-resolution face images) is available at the TinyFace Official Portal.

Pre-Trained Models and Extended Experimental Results

We will provide additional materials, including pre-trained models and our extended experimental results for high-resolution/mixed-resolution face recognition in the future.

Citation

C. Y. Low, J. C. L. Chai, J. Park, K. An, M. Cha, "SlackedFace: Learning a Slacked Margin for Low-Resolution Face Recognition," Paper Presented at the BMVC 2023: British Machine Vision Conference, Aberdeen, UK Nov. 2023.

Acknowledgement

This work was supported by the Institute for Basic Science (IBS-R029-C2) Korea.