Sa2VA: Marrying SAM2 with LLaVA for Dense Grounded Understanding of Images and Videos Paper • 2501.04001 • Published 3 days ago • 31