.png){kind=icon}
.png){kind=icon}
Return
Exploring a double task learning framework for makeup transfer
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engappai.2025.113142.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Jointly optimizing main makeup transfer and self-supervised auxiliary reconstruction, we propose unsupervised Double Task Makeup Transfer (DTMT), eliminating sub-optimal pseudo ground truth drawbacks. • Our Divide and Conquer Attention (DC-Attention) semantically aligns high-res features via coarse-to-fine processing, enabling efficient high-frequency makeup detail transfer with low overhead. • Experiments on three datasets show DTMT outperforms eight benchmarks quantitatively/qualitatively, boosting AI virtual try-on with strong style generalization and engineering application potential.
Journal
IF:
8
Papers:
5.3K
Citations:
3.5W