MvKE-FC

This repository contains more results for the paper:

Multi-view Knowledge Ensemble with Frequency Consistency for Face Cross-Domain Translation

Bing Cao, Qinghe Wang, Pengfei Zhu, Dongwei Ren, Wangmeng Zuo, Qinghua Hu

This paper is the first to address cross-domain face translation problem by ensembling multi-view knowledge from models trained on large-scale database designed for related tasks. And we further design the frequency consistent loss (mPrewitt loss and Gaussian fuzzy loss) to better fuse multi-view knowledge.

FC loss optimization.

Results

We provide more experimental results in the folders:

Each image has two faces:

• Input face
• Translated face

• Photo-to-Sketch Translation on CUFS and CUFSF:

• Sketch-to-Photo Translation on CUFS and CUFSF:

• Photo-to-Drawing Translation on APDrawing:

• Drawing-to-Photo Translation on APDrawing:

• NIR-to-VIS Translation on Oulu-CASIA-NIR-VIS:

The results can also be downloaded from Google Drive, or from Baidu Drive with password 1895 (Results on CUFSF are here).

Databases

• CUFS Database
• CUFSF Database
• APDrawing Database
• Oulu-CASIA-NIR-VIS Database

Contact

If you have any questions or suggestions about the paper, feel free to reach me (caobing@tju.​edu.cn / [email protected]).