Skip to main content
. 2022 Nov 21;8:e1161. doi: 10.7717/peerj-cs.1161

Table 2. Experimental results of VGG on CIFAR10/100.

Strategy Pooling Attn. CIFAR10 CIFAR100
Acc.1 (%) Acc.2 (%) Acc.1 (%) Acc.2 (%)
VGG_A max pooling N 91.29 91.61 68.49 69.47
VGG_B FMAPooling N 92.24 92.31 70.97 71.10
VGG_C FMMPooling N 91.76 92.24 70.13 70.93
VGG_D FMAPooling FMAttn 92.82 93.17 71.59 72.29
VGG_E FMAPooling CAM 92.75 93.02 70.47 71.55

Notes:

Acc.1: training with 155 epochs once.

Acc.2: training with 155 epochs twice.

The best results are highlighted in bold font.