Presentation
Domain Enhanced Arbitrary Image Style Transfer via Contrastive Learning
SessionTechnical Papers
Event Type
Technical Paper
Research & Education
Virtual
Full Conference Supporter
Full Conference
Virtual Conference Supporter
Virtual Conference
Exhibitor Additional Full Conference
Exhibitor Full Conference
Time
Location
DescriptionWe present Contrastive Arbitrary Style Transfer (CAST), a novel framework for arbitrary image style transfer based on contrastive learning. Our method learns an effective style representation directly from image features without using handcrafted, second-order statistics.
