Skip to content

Gaze-LLE models

Latest

Choose a tag to compare

@fkryan fkryan released this 11 Dec 20:49
· 21 commits to main since this release

We release our DINOv2-based Gaze-LLE models trained on GazeFollow, VideoAttentionTarget, and ChildPlay.

  • gazelle_dinov2_vitb14.pt
  • gazelle_dinov2_vitl14.pt
  • gazelle_dinov2_vitb14_inout.pt
  • gazelle_dinov2_vitl14_inout.pt
  • gazelle_dinov2_vitlb14_inout_childplay.pt
  • gazelle_dinov2_vitl14_inout_childplay.pt