[Retracted] Research on Efficient Deep Learning Algorithm Based on ShuffleGhost in the Field of Virtual Reality

Table 4

The trainable parameters, FLOPS, training time (CPU), and training time (GPU) from different structures.

Structure	Trainable parameters	FLOPs	Training time (CPU)	Training time (GPU)
Original VGG	14.728 M	314.032 M	1181 s	102.58 s
+ GhostNet-like module	7.388 M (-49%)	158.401 M	960 s	91.37 s
+ ShuffleGhost module	7.370 M (-50%)	157.295 M	931 s	90.84 s
+ Ghost bottleneck	8.337 M (-43%)	193.593 M	1111 s	110.32 s
+ ShuffleGhost BottleneckV1	7.795 M (-46%)	168.494 M	1112 s	114.82 s
+ ShuffleGhost BottleneckV2	7.670 M (-48%)	166.157 M	1097 s	122.05 s