Year: 2025
Type: conference-paper
Abstract: solely on a single trained model to infer transformations between source and target frames, we incorporate 3D guidance from an Image-to-3D model, which bridges this challenging task by explicitly projecting 2D information into 3D space.We design a data generation pipeline to ensure high-quality 3D guidance throughout training.Results show that by integrating these 3D priors, 3D-Fixup effectively supports complex, identity coherent 3Daware edits, achieving high-quality results and advancing the application of diffusion models in realistic image manipulation.The code is provided at https://3dfixup.github.io/.
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
Awards: G1766801816, G1957999595, G2926720356, MRI: Development of an Instrument for Deep Learning Research, RI: Medium: Creating Knowledge with All-Novel-Class Computer Vision, G5620962805, NSF-BSF: RI: Small: Structured Distributions in Deep Nets, CAREER: Learning to Anticipate with Visual Simulation
OpenAlex ID: W4412590239