Privacy Preservation with Detection Artifact Suppression
Paper | Supplemental | Video
Baowei Jiang*, Bing Bai*, Haozhe Lin*, Yu Wang, Yuchen Guo, Lu Fang
You can download images and annotations from website, and unzip files to data/widerface/.
[Expected directory structure of WIDERFACE (click to expand)]
./data/widerface
└───train
│ └───images
│ | └───0--Parade
│ | │ 0_Parade_marchingband_1_5.jpg
│ | │ ...
│ | └───1--Handshaking
│ | │ 1_Handshaking_Handshaking_1_42.jpg
│ | │ ...
| | ...
└───val
│ └───images
│ | └───0--Parade
│ | │ 0_Parade_marchingband_1_20.jpg
│ | │ ...
│ | └───1--Handshaking
│ | │ 1_Handshaking_Handshaking_1_35.jpg
│ | │ ...
| | ...
└───wider_face_split
| └───wider_face_train_bbx_gt.txt
| └───wider_face_val_bbx_gt.txt
| └───...
You need download labels file from RetinaFace Organise the dataset directory as follows:
[Expected directory structure of WIDERFACE (click to expand)]
./data/widerface
└───train
│ └───images
│ └───label.txt
└───val
│ └───images
│ └───label.txt
python gaussblur.py
python dartblur.py
(Ubuntu/Win10 3090Ti cuda11.8)
conda create -n dartblur python=3.8
pip install torch==1.12.1+cu113 torchvision==0.13.1+cu113 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu113
pip install -r requirements.txt
python train.py
https://github.com/bubbliiiing/retinaface-pytorch
@inproceedings{jiang2023dartblur,
title={DartBlur: Privacy Preservation With Detection Artifact Suppression},
author={Jiang, Baowei and Bai, Bing and Lin, Haozhe and Wang, Yu and Guo, Yuchen and Fang, Lu},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={16479--16488},
year={2023}
}