Table 4.

Result of augmentation and image size based on transformer and transfer learning models.

Model Name	Image Size	Accuracy	Epoch × Time
MobileNetV2	32 × 32	71.98%	195 × 386 s
VGG16	32 × 32	73.22%	165 × 391 s
VGG19	32 × 32	72.88%	177 × 404 s
ResNet50	32 × 32	76.67%	142 × 423 s
Swin Transformer	32 × 32	82.23%	200 × 95 s
Vit	32 × 32	81.56%	200 ×104 s
CCT	32 × 32	84.52%	200 × 78 s