CLIM: Contrastive Language-Image Mosaic for Region Representation

Benchmark Model Rank Results
open-vocabulary-object-detection-on-lvis-v1-0CLIM (RN50x64)#8AP novel-LVIS base training: 32.3
open-vocabulary-object-detection-on-mscocoCLIM (RN50)#16AP 0.5: 36.9