Parallel CNN Network Learning-Based Video Object Recognition for UAV Ground Detection

Table 2

Experimental results based on channel pruning.

Sheared layer	Pruning rate	Pruning time	Proportion parameters	Model accuracy
NULL	0	—	—	77.64%
VGG16	0.2	16 h	67.34%	77.43%
VGG16	0.3	20 h	53.58%	77.05%
VGG16	0.4	36 h	41.63%	75.47%
VGG16	0.5	40 h	31.56%	73.29%
All layers	0.2	18 h	66.50%	76.98%
All layers	0.3	22 h	52.32%	76.42%
All layers	0.4	38 h	39.96%	74.89%
All layers	0.5	40 h	29.47%	72.48%