Reconstructing CLIP for Open-Vocabulary Dense Perception
Open paper
Benchmark
Model
Rank
Results
open-vocabulary-object-detection-on-lvis-v1-0
DenseRC (ViT-L/14)
–
AP novel-LVIS base training: 39.6
open-vocabulary-semantic-segmentation-on-ade20k-150
DenseRC (ViT-L/14)
–
mIoU: 41.5
Rank counts only results with a code link.