Sa2VA: Marrying SAM2 with LLaVA for Dense Grounded Understanding of Images and Videos
Release notes
fixed masking