Skip to main content

Table 5 Accuracy (mAP and rank-1:R-1) and complexity (M: parameters and T: GFLOPS) of baseline and pruning Siamese networks on ReID datasets

From: Exploiting prunability for person re-identification

    Market-1501 DukeMTMC CUHK03-NP
Networks M T mAP rank-1 mAP rank-1 mAP rank-1
ResNet50 23.48 6.32 69.16 85.07 59.46 76.39 47.57 48.43
ResNet34 21.28 6.67 67.44 84.09 58.36 75.45 45.51 47.14
ResNet18 11.12 3.09 61.23 81.18 52.07 71.63 38.27 39.57
L1 11.90 2.96 67.04 84.71 57.51 75.00 44.08 46.50
Taylor 12.09 3.21 66.35 84.44 57.90 75.72 44.40 46.21
Auto-Balanced 11.90 2.96 65.46 83.64 56.45 74.64 41.85 44.21
Entropy 11.90 2.96 65.16 82.39 56.64 74.64 42.44 44.07
PSFP 11.90 2.96 65.92 83.72 56.96 74.66 42.38 45.58