CNFA: ConvNeXt Fusion Attention Module for Age Recognition of the Tangerine Peel

Table 4

Results of model comparison experiments.

Network	Param (M)	FLOPS (G)	Accuracy (%)	Precision (%)	Recall (%)	F1 (%)	Time (s)
CNN	25.4	4.217	82.59	82.60	81.65	82.12	104.23
ResNet50	25.6	4.158	95.98	94.65	94.92	94.69	93.69
ResNet50-SE	28.1	4.162	96.23	95.40	95.46	95.42	92.02
ResNet50-CBAM	28.1	4.168	96.27	95.53	95.12	95.32	90.97
ReNet50-csRSE	28.1	4.175	96.67	96.67	95.66	96.16	90.85
BiFormer	25.5	4.5	96.38	96.07	95.91	95.99	98.52
ConvNeXt	28.6	4.546	96.70	96.18	96.09	96.13	91.67
CNFA-integrated (ours)	30.1	4.551	97.17	96.71	96.86	96.78	91.02

Bold indicates the best performance.