Presentation
Encoder-based Domain Tuning for Fast Personalization of Text-to-image Models
DescriptionOur method uses an encoder to guide text-to-image personalization, teaching a model how to generate specific, personal concepts in 5-15 training steps — roughly three orders of magnitude faster than the state of the art.
Event Type
Technical Paper
TimeThursday, 10 August 20232:22pm - 2:33pm PDT
LocationPetree Hall D
Research & Education
Livestreamed
Recorded
Video
FC
FCS
V
VS
EFC