EResFD: Rediscovery of the Effectiveness of Standard Convolution for Lightweight Face Detection, a real-time CPU face detector.
Joonhyun Jeong1,2, Beomyoung Kim1,2, Joonsang Yu1,3, YoungJoon Yoo1
1 NAVER Cloud, ImageVision
2 KAIST
3 NAVER AI Lab
- 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 |
- download WIDER face dataset
python3 prepare_wider_data.py
according to yourself dataset path,modify data/config.py On widerface,
python3 test_wider.py --wider_root [widerface_directory] --model ./weights/eresfd_16.pth
you can test yourself image
python3 demo.py --model ./weights/eresfd_16.pth
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}
}
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
https://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.