Vision Transformer and Deep Sequence Learning for Human Activity Recognition in Surveillance Videos

Table 2

Different variants of ViT model used for image classification.

Model	Layers	Hidden size	MLP size	Heads	Params (M)
ViT-Base	12	768	3072	12	86
ViT-Large	24	1024	4096	16	307
ViT-Huge	32	1280	5120	16	632

The proposed method for features extraction is represented in bold text.