FACIAL INPAINTING IN UNALIGNED FACE IMAGES USING GENERATIVE ADVERSARIAL NETWORK WITH FEATURE RECONSTRUCTION LOSS

Avin Maulana, Chastine Fatichah, Nanik Suciati

Abstract


Facial inpainting or face restoration is a process to reconstruct some missing region on face images such that the inpainting results still can be seen as a realistic and original image without any missing region, in such a way that the observer could not realize whether the inpainting result is a generated or original image. Some of previous researches have done inpainting using generative network, such as Generative Adversarial Network. However, some problems may arise when inpainting algorithm have been done on unaligned face. The inpainting result show spatial inconsistency between the reconstructed region and its adjacent pixel, and the algorithm fail to reconstruct some area of face. Therefore, an improvement method in facial inpainting based on deep-learning is proposed to reduce the effect of the stated problem before, using GAN with additional loss from feature reconstruction and two discriminators. Feature reconstruction loss is a loss obtained by using pretrained network VGG-Net, Evaluation of the result shows that additional loss from feature reconstruction loss and two type of discriminators may help to increase visual quality of inpainting result, with higher PSNR and SSIM than previous result.


Full Text:

PDF

References


L. Haofu, G. Funka-Lea, Z. Yefeng, L. Jiebo, and S. K. Zhou, "Face Completion with Semantic Knowledge and Collaborative Adversarial Learning," in Proc. Asian Conference on Computer Vision, Dec. 2018, pp. 382–397, doi: 10.1007/978-3-030-20887-5_24.

M. A. Qureshi, M. Deriche, A. Beghdadi, and A. Amin, "A critical survey of state-of-the-art image inpainting quality assessment metrics," J. Vis. Commun. Image Represent., vol. 49, pp. 177–191, 2017, doi: 10.1016/j.jvcir.2017.09.006.

I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. MIT Press, 2016.

Y. Li, S. Liu, J. Yang, and M.-H. Yang, "Generative Face Completion," in Proc. IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 5892–5900, doi: 10.1109/CVPR.2017.624.

K. Simonyan and A. Zisserman, "Very Deep Convolutional Networks for Large-Scale Image Recognition," in Proc. International Conference on Learning Representations, 2015, pp. 1–14.

X. Hou, L. Shen, K. Sun, and G. Qiu, "Deep Feature Consistent Variational Autoencoder," in Proc. IEEE Winter Conference on Applications of Computer Vision, 2017, pp. 1133–1141, doi: 10.1109/WACV.2017.131.

L. Ziwei, L. Ping, W. Xiaogang, and T. Xiaoou, "Deep learning face attributes in the wild," in Proc. IEEE International Conference on Computer Vision, 2015, pp. 3730–3738, doi: 10.1109/ICCV.2015.425.

T. Tanaka, N. Kawai, Y. Nakashima, T. Sato, and N. Yokoya, "Iterative applications of image completion with CNN-based failure detection," J. Vis. Commun. Image Represent., vol. 55, no. May, pp. 56–66, 2018, doi: 10.1016/j.jvcir.2018.05.015.

I. J. Goodfellow et al., "Generative Adversarial Nets," Adv. Neural Inf. Process. Syst., vol. 27, pp. 4089–4099, 2014.

X. Hou, K. Sun, L. Shen, and G. Qiu, "Improving variational autoencoder with deep feature consistent and generative adversarial training," Neurocomputing, vol. 341, pp. 183–194, 2019, doi: 10.1016/j.neucom.2019.03.013.

M. Arjovsky, S. Chintala, and L. Bottou, "Wasserstein GAN," arXiv, vol. abs/1701.0, 2017, [Online]. Available: http://arxiv.org/abs/1701.07875.

K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit., vol. 2016-Decem, pp. 770–778, 2016, doi: 10.1109/CVPR.2016.90.

D. Pathak, P. Krahenbuhl, J. Donahue, T. Darrell, and A. A. Efros, "Context Encoders: Feature Learning by Inpainting," in Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit., pp. 2536–2544, 2016, doi: 10.1109/CVPR.2016.278.




DOI: http://dx.doi.org/10.12962/j24068535.v18i2.a1004

Refbacks

  • There are currently no refbacks.


Creative Commons License
JUTI (Jurnal Ilmiah Teknologi Informasi) by Department of Informatics, ITS is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. JUTI is accordance with CC BY-SA.