Lightweight Face Detectors

Filter-Pruning of Lightweight Face Detectors Using a Geometric Median Criterion

ML Model

License
Other
Copyright (c) 2024, K. Gkrispanis, N. Gkalelis, V. Mezaris / CERTH-ITI. All rights reserved. This code is provided for academic, non-commercial use only. Please also check for any restrictions applied in the code parts and datasets used here from other sources.

Main Characteristic

This AI Asset is associated with lightweight versions of the face detectors EXTD (Extremely Tiny Face Detector) and EResFD (Efficient ResNet Face Detector). These versions were obtained after applying the Filter Pruning via Geometric Median (FPGM) method in combination with the Soft Filter Pruning (SFP) iterative procedure. Experimental evaluations on the WIDER FACE dataset indicated that the applied approach has the potential to further reduce the model size of already lightweight face detectors, with limited accuracy loss, or even with small accuracy gain for low pruning rates.

Technical Categories
Computer vision
Last updated
06.05.2024 - 14:12
Detailed Description

This repository hosts the code and data for our paper: K. Gkrispanis, N. Gkalelis, V. Mezaris, "Filter-Pruning of Lightweight Face Detectors Using a Geometric Median Criterion", Proc. IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW 2024), Waikoloa, Hawaii, USA, Jan. 2024.

For detailed information about

  • Code requirements
  • Used dataset
  • Usage
  • Android Deployment
  • License and Citation

please refer to the Lightweight Face Detectors repository on Github.

Acknowledgements

This work was supported by the EU Horizon 2020 programme under grant agreement H2020-951911 AI4Media.

GDPR Requirements
None

Related Projects
AI4Media