EResFD (WACV 2024)

EResFD: Rediscovery of the Effectiveness of Standard Convolution for Lightweight Face Detection, a real-time CPU face detector.

Joonhyun Jeong^1,2, Beomyoung Kim^1,2, Joonsang Yu^1,3, YoungJoon Yoo¹

¹ _{NAVER Cloud, ImageVision}
² _KAIST
³ _{NAVER AI Lab}

Performance

• Comparison with other CPU-real time & GPU face detectors on WiderFace.

	Easy MAP	Medium MAP	hard MAP	Latency (480x640)
MTCNN	0.851	0.820	0.607	4.0ms
FaceBoxes	0.885	0.862	0.773	35.7 ms
RetinaFace (MobileNetV1)	0.886	0.870	0.809	58.5ms
EResFD-1x	0.890	0.879	0.804	37.7 ms

Prepare data

1. download WIDER face dataset
2. python3 prepare_wider_data.py

Evalution

according to yourself dataset path,modify data/config.py On widerface,

python3 test_wider.py --wider_root [widerface_directory] --model ./weights/eresfd_16.pth

Demo

you can test yourself image

python3 demo.py --model ./weights/eresfd_16.pth

Result

Citation

If you find that this project helps your research, please consider citing as below:

@article{jeong2022eresfd,
  title={EResFD: Rediscovery of the Effectiveness of Standard Convolution for Lightweight Face Detection},
  author={Jeong, Joonhyun and Kim, Beomyoung and Yu, Joonsang and Yoo, Youngjoon},
  journal={arXiv preprint arXiv:2204.01209},
  year={2022}
}

License

EResFD
Copyright (c) 2022-present NAVER Corp.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

References

• ssd.pytorch