발간년도 : [2021]
| 논문정보 |
|
| 논문명(한글) |
[Vol.16, No.3] Portrait Right Protection System using Real-Time Face Mosaic Technique based on Deep Learning |
|
| 논문투고자 |
Woo-Seok Sim, Byung-Tae Chun |
|
| 논문내용 |
Recently, the number of cases uploading videos taken by individuals on the Internet has increased. Uploading a video that includes not only yourself but also people around you can cause infringement of portrait rights and privacy issues. In general, The portrait rights of others are protected by manually blurring faces of people moving over time using a video editing program. However, this is a very cumbersome task and the existing automatic mosaic system also has limitations because it does not detect all identifiable faces properly in videos and it misses tracked faces frequently. Therefore, to prevent the problem of infringement of portrait rights, this paper proposes a portrait rights protection system using deep learning. The system uses YOLOv3, Face-Net, and SORT algorithm to protect the portrait rights by blurring faces of other people except the users registered in the system. Compared to the existing automatic mosaic system, It can detect all identifiable faces and blur faces more accurately. It can operate in real-time by skipping a certain frame and performing face verification. It is also possible to flexibly respond to various changes of the face reflected on the camera by applying the SORT object tracking algorithm. The proposed system can improve convenience for users by automating the task of manually blurring faces of people and it shows superior performance compared to the existing automatic mosaic system. |
|
| 첨부논문 |
|
|
|
|
|
자료실 > 논문지 
