Anonymizing videos by DSFD: Dual Shot Face Detector

Modified by Jongkuk Lim

History

This repository was forked from FaceDetection-DSFD which is implentation of DSFD: Dual Shot Face Detector by Jian Li, Yabiao Wang, Changan Wang, Ying Tai, Jianjun Qian, Jian Yang, Chengjie Wang, Jilin Li, Feiyue Huang.

Introduction

Simple implementation of video anonymization.

If you are looking for a faster version, check Anonymizing videos by lightDSFD. And, if you are looking for a simpler example, noone video is implemented by only OpenCV examples.

Note that this repository is not designed for training models. If you are looking for training models, please visit original repository FaceDetection-DSFD.

Comparisons

Requirements

CUDA supported enviornment

(Tested on NVIDIA GTX 1060(6GB) and GTX 1080 Ti(8GB))

• Torch >= 0.3.1
• Torchvision >= 0.2.1
• (Tested on torch 1.3.1 and Torchvision 0.4.2)
• Python 3.6

Getting Started

1. Download DSFD model from original repository provided [微云] [google drive]

2. Place WIDERFace_DSFD_RES152.pth to ./weights/.

3. Run ./demo.py to check if it is running.

python demo.py [--trained_model [TRAINED_MODEL]] [--img_root  [IMG_ROOT]] 
               [--save_folder [SAVE_FOLDER]] [--visual_threshold [VISUAL_THRESHOLD]] 
    --trained_model      Path to the saved model
    --img_root           Path of test images
    --save_folder        Path of output detection resutls
    --visual_threshold   Confidence thresh

Usage

usage: blur_video.py [-h] -i INPUT -o OUTPUT [--vertical VERTICAL]
                     [--verbose VERBOSE] [--reduce_scale REDUCE_SCALE]
                     [--trained_model TRAINED_MODEL] [--threshold THRESHOLD]
                     [--cuda CUDA]

Required arguments

Required file paths:
  -i INPUT, --input INPUT
                        Video file path
  -o OUTPUT, --output OUTPUT
                        Output video path

Optional arguments

optional arguments:
  -h, --help            show this help message and exit
  --vertical VERTICAL   0 : horizontal video(default), 1 : vertical video
  --verbose VERBOSE     Show current progress and remaining time
  --reduce_scale REDUCE_SCALE
                        Reduce scale ratio. ex) 2 = half size of the input.
                        Default : 2
  --trained_model TRAINED_MODEL
                        Trained state_dict file path to open
  --threshold THRESHOLD
                        Final confidence threshold
  --cuda CUDA           Use cuda

Citation

If you find DSFD useful in your research, please consider citing:

@inproceedings{li2018dsfd,
  title={DSFD: Dual Shot Face Detector},
  author={Li, Jian and Wang, Yabiao and Wang, Changan and Tai, Ying and Qian, Jianjun and Yang, Jian and Wang, Chengjie and Li, Jilin and Huang, Feiyue},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  year={2019}
}