List your company

Referring Expression Grounding on OCID-Ref

94.6Accuracy (Overall)VC-1
62.04870.49978.9587.401May 9, 2024

Evaluation results

MethodLinks
VC-1Architecture=ViT-based, Type=vision-encoder-only2024.0594.693.796.593.7
SD-VAEType=vision-encoder-only2024.0594.393.296.393.4
SCRprotocol=frozen, Backbone=U-Net mid/downsampling layers2024.0592.991.195.991.8
Voltron2024.0592.596.991.890.2
SCR-FTprotocol=fine-tuned, Backbone=U-Net mid/downsampling layers2024.0591.890.194.890.8
CLIPArchitecture=ViT-based2024.0568.160.376.667
R3M2024.0563.355.368.363.3