基於衣服轉換之深度虛擬試衣技術

Deep Virtual Try-on with Clothes Transform

陳思穎(Szu-Ying Chen)

指導教授 ： 莊永裕

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2018年)

https://doi.org/10.6342/NTU201802721

摘要

本論文的目標是讓使用者能夠用照片的方式來進行試穿衣服，也就是使用者提供一張自己的照片，以及另一張他人的照片，便能夠將他人身上的衣服換穿到自己的照片上。
其他虛擬試穿的方法都是針對人和衣服的正面照片來做處理，而我們的方法能夠處理正面、微轉向左側、及微轉向右側的方向。相較於其他方法，我們的方法更具有普遍性，而衣服的細節紋理也更為清晰。在用戶研究中，約90%的例子裡，比起其他方法，使用者比較喜歡我們的結果。

關鍵字

卷積神經網路 ； 虛擬試衣 ； 感知損失

並列摘要

The goal of this thesis is to enable users to try on clothes by photos, that is, users provide their own photo and another photo of others, and they can change the clothes of others to their own.
Other virtual try-on methods are focus on front-view of the person and the clothes, and our method can handle front, slightly turn to the left, and slightly turn to the right viewing directions. Compared to other methods, our method is more universal, and the details of the clothes are more clear. In the user study, in about 90% cases, the respondents more like our results than the results of other method.

並列關鍵字

convolutional neural network ； virtual try-on ； perceptual loss

參考文獻

[1] K. E. Ak, A. A. Kassim, J. Hwee Lim, and J. Yew Tham. Learning attribute rep- resentations with localization for flexible fashion search. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition(CVPR), June 2018.

[2] S. Belongie, J. Malik, and J. Puzicha. Shape matching and object recognition using shape contexts. IEEE Transactions on Pattern Analysis and Machine Intelligence, 24(4):509–522, Apr 2002.

[3] K.Gong, X.Liang, D.Zhang, X.Shen, andL.Lin. Look into person: Self-supervised structure-sensitive learning and a new benchmark for human parsing. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), July 2017.

[4] X. Han, Z. Wu, Z. Wu, R. Yu, and L. S. Davis. Viton: An image-based virtual try- on network. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2018.

[5] N. Jetchev and U. Bergmann. The conditional analogy gan: Swapping fashion arti- cles on people images. In The IEEE International Conference on Computer Vision (ICCV) Workshops, Oct 2017.