Table 1. Rank-1 recognition rate (%) performance comparison between the baseline FNM and the proposed Multi-Image FNM across different numbers of input images (n). The proposed Multi-Image FNM was trained using two input images.

number of input images	FNM [1]	Multi-Image FNM (maxpolling)	Multi-Image FNM (avgpooling)
n=1	87.39	87.39	87.39
n=2	-	94.47	93.91
n=3	-	95.63	95.81
n=4	-	95.82	96.61
n=5	-	95.89	97.15
n=6	-	95.94	97.28
n=7	-	96.01	97.43
n=8	-	95.94	97.63