Table I.

The quantitative results of infection areas on our uAI dataset compared to the corresponding ground-truth. The best results are highlighted in bold.

Label	Methods	DI	JA	SE	SP	MAE
Scribble	p-UNet [55]	0.683	0.531	0.837	0.836	0.127
	WSOD [54]	0.706	0.558	0.871	0.885	0.093
	S2L [44]	0.745	0.607	0.843	0.956	0.087
	USTM -Net (proposed)	0.762	0.628	0.882	0.984	0.076
Point	WSCL [18]	0.719	0.573	0.826	0.929	0.114
Full	U-Net [49]	0.788	0.663	0.885	0.978	0.071