Open-vocabulary Object Detection via Vision and Language Knowledge Distillation

Benchmark Model Rank Results
open-vocabulary-object-detection-on-lvis-v1-0ViLD-ensemble w/ ALIGN (Eb7-FPN)#13AP novel-LVIS base training: 26.3
open-vocabulary-object-detection-on-lvis-v1-0ViLD-ensemble (R152-FPN)#23AP novel-LVIS base training: 18.7
open-vocabulary-object-detection-on-lvis-v1-0ViLD-ensemble (R50-FPN)#26AP novel-LVIS base training: 16.6
open-vocabulary-object-detection-on-lvis-v1-0ViLD (R50-FPN)#27AP novel-LVIS base training: 16.1
open-vocabulary-object-detection-on-mscocoViLD#27AP 0.5: 27.6