Table 3.

Comparative analysis of model complexity.

Model	Trainable parameters (M)	FLOPs (G)	Fusion time (s)	Peak GPU memory (GB)
DWT-BL	N/A (non-learnable)	N/A	0.42	N/A
IFCNN	0.08	0.56	0.02	0.31
DenseFuse	0.07	0.51	0.55	0.28
U2Fusion	0.66	4.83	0.53	0.92
MSML-DenseXmer	6.52	18.42	2.46	3.74