Using Fully Connected and Convolutional Net for GAN-Based Face Swapping

Bo Shue Lin; Ding Wen Hsu; Chin Han Shen; Hsu-Feng Hsiao

doi:10.1109/APCCAS50809.2020.9301665

Using Fully Connected and Convolutional Net for GAN-Based Face Swapping

Bo Shue Lin, Ding Wen Hsu, Chin Han Shen, Hsu-Feng Hsiao

Department of Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

4 Scopus citations

Abstract

The lifelike results of using face swapping have contributed greatly to the research in computer vision. In this work, we extend the architecture of faceswap-GAN in order to obtain more natural results compared to the original framework. In the original architecture, the self-attention module usually converts the facial features from a source face to the target face with artificial distortion around the facial features. We use a structure of fully connected convolutional layers as a discriminator to approach the problem. The outcome can be smoother and more natural perceptually compared to the results using the original faceswap-GAN.

Original language	American English
Title of host publication	Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020
Editors	Xuan-Tu Tran, Duy-Hieu Bui
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	185-188
Number of pages	4
ISBN (Electronic)	9781728193960
DOIs	https://doi.org/10.1109/APCCAS50809.2020.9301665
State	Published - 8 Dec 2020
Event	16th IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020 - Virtual, Halong, Viet Nam Duration: 8 Dec 2020 → 10 Dec 2020

Publication series

Name	Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020

Conference

Conference	16th IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020
Country/Territory	Viet Nam
City	Virtual, Halong
Period	8/12/20 → 10/12/20

Keywords

Deepfake
fully-connected and convolutional network
Generative adversarial network (GAN)

Access to Document

10.1109/APCCAS50809.2020.9301665

Cite this

Lin, B. S., Hsu, D. W., Shen, C. H., & Hsiao, H.-F. (2020). Using Fully Connected and Convolutional Net for GAN-Based Face Swapping. In X.-T. Tran, & D.-H. Bui (Eds.), Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020 (pp. 185-188). Article 9301665 (Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/APCCAS50809.2020.9301665

Lin, Bo Shue ; Hsu, Ding Wen ; Shen, Chin Han et al. / Using Fully Connected and Convolutional Net for GAN-Based Face Swapping. Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020. editor / Xuan-Tu Tran ; Duy-Hieu Bui. Institute of Electrical and Electronics Engineers Inc., 2020. pp. 185-188 (Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020).

@inproceedings{9bf63f8fc7f54180b5915d28533a5a0a,

title = "Using Fully Connected and Convolutional Net for GAN-Based Face Swapping",

abstract = "The lifelike results of using face swapping have contributed greatly to the research in computer vision. In this work, we extend the architecture of faceswap-GAN in order to obtain more natural results compared to the original framework. In the original architecture, the self-attention module usually converts the facial features from a source face to the target face with artificial distortion around the facial features. We use a structure of fully connected convolutional layers as a discriminator to approach the problem. The outcome can be smoother and more natural perceptually compared to the results using the original faceswap-GAN.",

keywords = "Deepfake, fully-connected and convolutional network, Generative adversarial network (GAN)",

author = "Lin, {Bo Shue} and Hsu, {Ding Wen} and Shen, {Chin Han} and Hsu-Feng Hsiao",

note = "Publisher Copyright: {\textcopyright} 2020 IEEE. Copyright: Copyright 2021 Elsevier B.V., All rights reserved.; 16th IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020 ; Conference date: 08-12-2020 Through 10-12-2020",

year = "2020",

month = dec,

day = "8",

doi = "10.1109/APCCAS50809.2020.9301665",

language = "American English",

series = "Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "185--188",

editor = "Xuan-Tu Tran and Duy-Hieu Bui",

booktitle = "Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020",

address = "United States",

}

Lin, BS, Hsu, DW, Shen, CH & Hsiao, H-F 2020, Using Fully Connected and Convolutional Net for GAN-Based Face Swapping. in X-T Tran & D-H Bui (eds), Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020., 9301665, Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020, Institute of Electrical and Electronics Engineers Inc., pp. 185-188, 16th IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020, Virtual, Halong, Viet Nam, 8/12/20. https://doi.org/10.1109/APCCAS50809.2020.9301665

Using Fully Connected and Convolutional Net for GAN-Based Face Swapping. / Lin, Bo Shue; Hsu, Ding Wen; Shen, Chin Han et al.
Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020. ed. / Xuan-Tu Tran; Duy-Hieu Bui. Institute of Electrical and Electronics Engineers Inc., 2020. p. 185-188 9301665 (Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Using Fully Connected and Convolutional Net for GAN-Based Face Swapping

AU - Lin, Bo Shue

AU - Hsu, Ding Wen

AU - Shen, Chin Han

AU - Hsiao, Hsu-Feng

PY - 2020/12/8

Y1 - 2020/12/8

N2 - The lifelike results of using face swapping have contributed greatly to the research in computer vision. In this work, we extend the architecture of faceswap-GAN in order to obtain more natural results compared to the original framework. In the original architecture, the self-attention module usually converts the facial features from a source face to the target face with artificial distortion around the facial features. We use a structure of fully connected convolutional layers as a discriminator to approach the problem. The outcome can be smoother and more natural perceptually compared to the results using the original faceswap-GAN.

AB - The lifelike results of using face swapping have contributed greatly to the research in computer vision. In this work, we extend the architecture of faceswap-GAN in order to obtain more natural results compared to the original framework. In the original architecture, the self-attention module usually converts the facial features from a source face to the target face with artificial distortion around the facial features. We use a structure of fully connected convolutional layers as a discriminator to approach the problem. The outcome can be smoother and more natural perceptually compared to the results using the original faceswap-GAN.

KW - Deepfake

KW - fully-connected and convolutional network

KW - Generative adversarial network (GAN)

UR - http://www.scopus.com/inward/record.url?scp=85099568212&partnerID=8YFLogxK

U2 - 10.1109/APCCAS50809.2020.9301665

DO - 10.1109/APCCAS50809.2020.9301665

M3 - Conference contribution

AN - SCOPUS:85099568212

T3 - Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020

SP - 185

EP - 188

BT - Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020

A2 - Tran, Xuan-Tu

A2 - Bui, Duy-Hieu

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 16th IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020

Y2 - 8 December 2020 through 10 December 2020

ER -

Lin BS, Hsu DW, Shen CH, Hsiao HF. Using Fully Connected and Convolutional Net for GAN-Based Face Swapping. In Tran XT, Bui DH, editors, Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020. Institute of Electrical and Electronics Engineers Inc. 2020. p. 185-188. 9301665. (Proceedings of 2020 IEEE Asia Pacific Conference on Circuits and Systems, APCCAS 2020). doi: 10.1109/APCCAS50809.2020.9301665

Using Fully Connected and Convolutional Net for GAN-Based Face Swapping

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this