Table 1: Comparison with fully-supervised transfer learning:

CAiD models outperform fully-supervised pre-trained models in 3 downstream tasks.

Initialization	Pre-training dataset	Classification [AUC (%)]		Segmentation [Dice (%)]
Initialization	Pre-training dataset	ChestX-ray14	CheXpert	SIIM-ACR	Montgomery

Random	-	80.31±0.10	86.62±0.15	67.54±0.60	97.55±0.36

Supervised	ImageNet	81.70±0.15	87.17±0.22	67.93±1.45	98.19±0.13
Supervised	ChestX-ray14	-	87.40±0.26	68.92±0.98	98.16±0.05

CAiDMoCo-v2	ChestX-ray14	80.72±0.29	86.86±0.37 ^† ^†	68.16±1.07 ^† ^†	98.19±0.08 ^† ^†
CAiDBarlow Twins	ChestX-ray14	80.86±0.16	87.44±0.33 ^‡ ^†	69.83±0.29 ^‡ ^‡	98.15±0.11 ^†
CAiDSimSiam	ChestX-ray14	79.45±0.42	84.45±0.46	68.35±1.16 ^† ^†	98.01±0.28 ^†

The ‡ and † present the statistically significant (p < 0.05) and equivalent performances, respectively, compared to supervised ImageNet and ChestX-ray14 baselines.