Detecting Twenty-thousand Classes using Image-level Supervision

Benchmark Model Rank Results
cross-domain-few-shot-object-detection-onDetic-FT#11mAP: 12.0
cross-domain-few-shot-object-detection-onDetic(w/o FT)#13mAP: 0.6
cross-domain-few-shot-object-detection-on-1Detic-FT#8mAP: 22.3
cross-domain-few-shot-object-detection-on-1Detic(w/o FT)#9mAP: 11.4
cross-domain-few-shot-object-detection-on-2Detic-FT#10mAP: 15.4
cross-domain-few-shot-object-detection-on-2Detic(w/o FT)#13mAP: 0.1
cross-domain-few-shot-object-detection-on-3Detic-FT#7mAP: 17.9
cross-domain-few-shot-object-detection-on-3Detic(w/o FT)#10mAP: 0.9
cross-domain-few-shot-object-detection-on-4Detic-FT#3mAP: 16.8
cross-domain-few-shot-object-detection-on-4Detic(w/o FT)#13mAP: 0.0
cross-domain-few-shot-object-detection-on-neuDetic-FT#3mAP: 16.8
cross-domain-few-shot-object-detection-on-neuDetic(w/o FT)#10mAP: 0.0
open-vocabulary-object-detection-on-lvis-v1-0Detic#24AP novel-LVIS base training: 17.8
open-vocabulary-object-detection-on-mscocoDetic#26AP 0.5: 27.8