Open-World Semi-Supervised Learning

Benchmark Model Rank Results
novel-object-detection-on-lvis-v1-0-valORCA Cao et al. (2022)#4All mAP: 2.03Known mAP: 20.57Novel mAP: 0.49
open-world-semi-supervised-learning-on-1ORCA (ResNet-50)#2Novel accuracy (50% Labeled): 72.1Seen accuracy (50% Labeled): 89.1
open-world-semi-supervised-learning-on-cifarORCA (ResNet-18)#3All accuracy (10% Labeled): 84.1All accuracy (50% Labeled): 89.7
open-world-semi-supervised-learning-on-cifar-1ORCA (ResNet-18)#3All accuracy (10% Labeled): 38.6Novel accuracy (10% Labeled): 31.8