Skip to main content

Table 6 Theoretical amount of computation in FLOPs and the corresponding acceleration in practice

From: An efficient pruning scheme of deep neural networks for Internet of Things applications

Archi. Original FLOPs Original time (s) Pruned FLOPs Pruned time (s) Acceleration (%)
VGG 3.13×108 0.428 1.59×108 0.259 39.5
ResNet-56 1.25×108 0.256 6.43×107 0.215 16.1
MobileNet 2.97×107 0.421 1.15×107 0.200 52.5
ResNet-50 3.38×109 5.453 1.71×109 4.098 25.0