Skip to main content

Table 4 Rank-1 accuracy and complexity (M: number of parameters and T: GFLOPS) of baseline and pruned ResNet CNNs on ImageNet dataset [77]

From: Exploiting prunability for person re-identification

  Performance measures
Networks rank-1 T M
ResNet50 76.01 6.32 23.48
ResNet34 73.27 6.67 21.28
ResNet18 69.64 3.09 11.12
L1 71.85 2.96 11.90
Taylor 71.65 3.21 12.09
Auto-Balanced 71.97 2.96 11.90
Entropy 71.46 2.96 11.90
PSFP 71.57 2.96 11.90