A Multiscale-Based Adjustable Convolutional Neural Network for Multiple Organ Segmentation

Table 2

The effect of on the performance. We calculate the percentage of FLOPs and memory relative to the baseline U-Net (the first row).

	DSC (%)	FLOPs (G)	Memory (10^6)
Baseline ()	94.34	123.93	1439
0.125	95.43	102.09 (82%)	1313 (91%)
0.25	95.41	83.26 (67%)	1188 (83%)
0.5	95.32	54.28 (44%)	938 (65%)
0.75	95.09	36.84 (30%)	687 (47%)
0.875	94.88	32.45 (26%)	562 (39%)
1.0	93.52	30.98 (25%)	365 (25%)