Table 1.

Approximate runtime performance for training and prediction runs for all five algorithms on HuBMAP data.

Model	Training on kidney data (n = 15)	Training on colon data (no pretraining on kidney data) (n = 5)	Training on colon data (pretrained on kidney data) (n = 5)	Inference on kidney data (n = 10)	Inference on colon data (n = 2)
Tom	23 h	2 h	2.5 h	1.3 h	1.3 min
Gleb	8 h	4 h	4 h	0.4 hs	2 min
Whats goin on	24 h	0.5 h	0.5 h	0.75 h	1.75 min
DeepLive.exe	51 h	45 h	45 h	1 h	2 min
Deepflash2	8 h	1.6 h	1 h	0.22 h	0.35 min