👀 Are you at #ICCV2025?
Check out Poster #125 where we present ScanDiff:

✅ Generates diverse, realistic human scanpaths
✅ Uses diffusion + ViTs + optional text conditioning
✅ Beats SOTA in free-viewing & task-driven tasks

Come talk with us!
#DiffusionModels #GazePrediction #ICCV2025