List your company
ResearchFormerly WizWand

Open-vocabulary semantic segmentation on ADE20K (val)

32mIoUT-REN+
1.7369.59317.4525.307Apr 20, 2026May 2, 2026May 15, 2026May 28, 2026Jun 10, 2026Jun 23, 2026Jul 6, 2026
Updated 4mo ago

Evaluation results

MethodLinks
T-REN+Approach Type=Patch-based vision-language encoder, Backbone=ViT-L, Tokens per image=576, Input Resolution=672p2026.0432
T-RENApproach Type=Patch-based vision-language encoder, Backbone=ViT-L, Tokens per image=576, Input Resolution=384p2026.0430.6
RADSeg+Approach Type=SAM-guided, Backbone=ViT-H, Input Resolution=672p2026.0429.9
TextRegionApproach Type=SAM-guided, Backbone=ViT-H, Input Resolution=672p2026.0427.3
TridentApproach Type=SAM-guided, Backbone=ViT-H, Input Resolution=672p2026.0425.6
DINOv3 dino.txtApproach Type=Patch-based vision-language encoder, Backbone=ViT-L, Tokens per image=5762026.0424.7
CLIP-DINOisertraining=true2026.0720
DINOv2 dino.txtApproach Type=Patch-based vision-language encoder, Backbone=ViT-L, Tokens per image=5762026.0419.2
CLIPixtraining=false2026.0719.1
ResCLIPtraining=false2026.0718
PEApproach Type=Patch-based vision-language encoder, Backbone=ViT-L, Tokens per image=5762026.0417.6
ClearCLIPtraining=false2026.0716.7
SCLIPtraining=false2026.0716.1
TCLtraining=true2026.0714.9
EVA-02-CLIPApproach Type=Patch-based vision-language encoder, Backbone=ViT-L, Tokens per image=5762026.0410.9
SigLIP-2Approach Type=Patch-based vision-language encoder, Backbone=ViT-L, Tokens per image=5762026.0410.8
GroupViTtraining=true2026.079.2
CLIPApproach Type=Patch-based vision-language encoder, Backbone=ViT-L, Tokens per image=5762026.046
CLIPtraining=false2026.072.9