Gaze-LLE models

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

• gazelle_dinov2_vitb14.pt
• gazelle_dinov2_vitl14.pt
• gazelle_dinov2_vitb14_inout.pt
• gazelle_dinov2_vitl14_inout.pt