Michael Dorkenwald

Latest

• SIGMA: Sinkhorn-Guided Masked Video Modeling
• PIN: Positional Insert Unlocks Object Localisation Abilities in VLMs