List your company
ResearchFormerly WizWand

Multiple Object Grounding on Reasoning Segmentation Overall

48.7gIoU (%)LISA
20.20427.6023542.398May 29, 2025
Updated 8mo ago

Evaluation results

MethodLinks
LISATraining paradigm=Train LLaVA 1.5-7B, Query source=Original2025.0548.748.8
TextRegionQuery source=Interpreted from LLaVA 1.5-7B2025.0542.742.4
OVSegQuery source=Interpreted from LLaVA 1.5-7B2025.0539.731.8
TridentQuery source=Interpreted from LLaVA 1.5-7B2025.0538.439.2
OVSegTraining paradigm=Training-based, VLM usage=No, Query source=Original2025.0526.120.8
TextRegionTraining paradigm=Training-free, VLM usage=No, Query source=Original2025.0525.221.6
SEEMTraining paradigm=Training-based, VLM usage=No, Query source=Original2025.0524.318.7
TridentTraining paradigm=Training-free, VLM usage=No, Query source=Original2025.0523.320
X-DecoderTraining paradigm=Training-based, VLM usage=No, Query source=Original2025.0521.716.3
Grounded-SAMTraining paradigm=Training-based, VLM usage=No, Query source=Original2025.0521.316.4