Table 1 Classification test accuracy across 3 publicly available image datasets after 64 epochs
From: LBCapsNet: a lightweight balanced capsule framework for image classification of porcelain fragments
Method | MNIST (%) | F-MNIST (%) | CIFAR10 (%) |
|---|---|---|---|
AlexNet | 99.14 | 90.31 | 75.92 |
VGG | – | 91.82 | 85.06 |
ResNet | – | 92.37 | 88.65 |
CapsNet(baseline) | 99.50 | 89.80 | 68.53 |
CapsNet [19] | 99.75 | 93.60 | 89.40 |
MS-Capsnet [34] | – | 92.70 | 75.70 |
RS-Capsnet [36] | – | 94.08 | 91.01 |
DeepCaps [35] | 99.72 | 94.46 | 91.01 |
Limit-Caps [33] | 99.50 | 89.80 | 68.53 |
DA-Caps [38] | 99.53 | 93.98 | 85.47 |
GraCapsNet [50] | 99.50 | 93.10 | 82.21 |
LBCapsNet | 99.68 | 94.58 | 91.32 |
