/ Machine Learning

A PyTorch Implementation of Extremely Tiny Face Detector via Iterative Filter Reuse

A PyTorch Implementation of Extremely Tiny Face Detector via Iterative Filter Reuse

EXTD_Pytorch

EXTD: Extremely Tiny Face Detector via Iterative Filter Reuse

YoungJoon Yoo, Dongyoon Han, Sangdoo Yun

Requirement

  • pytorch 1.0 (checked at 1.0)
  • opencv
  • numpy
  • easydict
  • Python3

Prepare data

WIDER face dataset is used. see the S3FD.pytorch git for more detail.

Train

You can use

python train.py 

Refer the train.py files to check the arguement.
Our setting was

"--batch_size 16 --lr 0.001" 

Evaluation on WIDER Dataset

You should complie the bounding box function. Type

python3 bbox_setup.py build_ext --inplace

Then run

python3 wider_test.py

Demo

you can test your image from

python3 demo.py

References

Citation

@article{yoo2019extd,
  title={EXTD: Extremely Tiny Face Detector via Iterative Filter Reuse},
  author={Yoo, YoungJoon and Han, Dongyoon and Yun, Sangdoo},
  journal={arXiv preprint arXiv:1906.06579},
  year={2019}
}

GitHub