Generative AI
Building controllable generation pipelines with diffusion and flow-matching models.
I am interested in controllable visual generation systems that understand user intent, image structure, and fine-grained editing constraints.
Building controllable generation pipelines with diffusion and flow-matching models.
Editing images with semantic control, spatial consistency, and faithful preservation of visual details.
Modeling garment fit, body shape, and realistic try-on behavior across diverse subjects.
Recent releases and updates.
A growing list of research projects and papers.
A fit-aware virtual try-on framework that models garment-body size control and evaluates fitting fidelity with the FittingEffect3K benchmark and a fit-oriented VLM protocol.