Disentangling Structure and Appearance in ViT Feature Space

Supplementary Material

SpliceNet Comparisons: SD-Horses

We present SpliceNet results of semantic appearance transfer with respect to the recent baselines: (1) Kim. et al [1] (2) Swapping Autoencoders (SA) [2] (3) WCT2 [3] (4) Splice. SpliceNet manages to transfer complex texture on corresponding semantic parts across images. Note that in some cases, Splice is prone to instabilities during its optimization process, which may lead to incorrect semantic association and poor visual quality. On the other hand, since SpliceNet is trained on a dataset of semantically related image pairs, it results in a better semantic association, higher visual quality, and is more robust to challenging, unaligned input pairs

Appearance	Structure	Kim et al. [1]	SA [2]	WCT2 [3]	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

Appearance	Structure	Kim et al.	SA	WCT2	Splice	SpliceNet

[1] Kunhee Kim, Sanghun Park, Eunyeong Jeon, Taehun Kim, and Daijin Kim. 2022. A Style-aware Discriminator for Controllable Image Translation. In IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[2] Taesung Park, Jun-Yan Zhu, Oliver Wang, Jingwan Lu, Eli Shechtman, Alexei A. Efros, and Richard Zhang. 2020. Swapping Autoencoder for Deep Image Manipu- lation. In Advances in Neural Information Processing Systems

[3] Jaejun Yoo, Youngjung Uh, Sanghyuk Chun, Byeongkyu Kang, and Jung-Woo Ha. 2019. Photorealistic Style Transfer via Wavelet Transforms. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV).