基於PRNet方法開發臉部卡通化換臉應用

A Face Swap Application with Cartoon Style Based on PRNet

莊晉翔(JIN-XIANG ZHUANG)

指導教授 ： 王佳盈

中原大學/電機資訊學院/電機工程研究所/碩士(2021年)

https://doi.org/10.6840/cycu202100586

摘要

本論文主要製作一個人臉變換的應用程式，其中使用了人臉偵測、人臉對齊演算法 PRNet，以及人臉卡通化等相關的技術。使用者只要輸入兩張照片，系統就可以將照片卡通化，並進行兩邊人臉的置換，另外也可以直接選擇整部影片，程式會將影片中的人臉進行置換。系統操作界面簡單、直觀，不需要耗費很多時間就可輕易上手。這個應用程式，主要是做為一個前期研究，未來希望能持續改善並應用於特教領域中的「影片示範教學」。我們希望讓一般的特教老師，可以使用個人電腦設備，就能將示範影片中的人物進行置換，以產出個別化的示範教學影片，並能以此來幫助特教學生進行學習。

關鍵字

人臉置換 ； 換臉 ； 人臉偵測 ； 人臉對齊演算法 ； PRNet

並列摘要

In this thesis, we develop a face swap application based on face detection, face alignment algorithm PRNet, and face cartoonization technologies. For two different face photos, the system can cartoonize the photos and exchange the faces on both sides. In addition, the program can also swap faces in two videos directly. The user interface is easy and intuitive to use. This application is mainly used for a preliminary research, and we hope to apply it to “video modeling” in the field of special education in the future. We hope that special education teachers can use the application to replace the characters in a teaching video to make personalized self-modeling videos. And these videos can be used to help special education students in need.

並列關鍵字

Face swap ； Face recognition ； Face alignment algorithm ； PRNet

參考文獻

[1] Feng, Y., Wu, F., Shao, X., Wang, Y., Zhou, X. (2018). Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network. ECCV.

[2] Kim, J., Kim, M., Kang, H., Lee, K. (2020). U-GAT-IT: Unsupervised Generative Attentional Networks with Adaptive Layer-Instance Normalization for Image-to-Image Translation. ArXiv, abs/1907.10830.

[3] minivision-ai/photo2cartoon. (2021). Retrieved 20 May 2021, from https://github.com/minivision-ai/photo2cartoon

[4] Natsume, R., Yatagawa, T., Morishima, S. (2018). RSGAN: face swapping and editing using face and hair representation in latent spaces. ACM SIGGRAPH 2018 Posters.

[5] Sadu, C., Das, P. (2020). Swapping Face Images Based on Augmented Facial Landmarks and Its Detection. 2020 IEEE REGION 10 CONFERENCE (TENCON). doi: 10.1109/tencon50793.2020.9293884