In Defense of the Triplet Loss for Person Re-Identification

22 Mar 2017  ·  Alexander Hermans, Lucas Beyer, Bastian Leibe ·

In the past few years, the field of computer vision has gone through a revolution fueled mainly by the advent of large datasets and the adoption of deep convolutional neural networks for end-to-end learning. The person re-identification subfield is no exception to this... Unfortunately, a prevailing belief in the community seems to be that the triplet loss is inferior to using surrogate losses (classification, verification) followed by a separate metric learning step. We show that, for models trained from scratch as well as pretrained ones, using a variant of the triplet loss to perform end-to-end deep metric learning outperforms most other published methods by a large margin. read more

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Person Re-Identification CUHK03 TriNet Rank-1 89.63 # 4
Rank-5 99.01 # 3
Person Re-Identification DukeMTMC-reID TriNet Rank-1 72.44 # 54
mAP 53.50 # 55
Person Re-Identification Market-1501 LuNet (RK) Rank-1 84.59 # 66
Rank-5 91.89 # 12
mAP 75.62 # 57
Person Re-Identification Market-1501 LuNet Rank-1 81.38 # 71
Rank-5 92.34 # 11
mAP 60.71 # 74
Person Re-Identification Market-1501 TriNet Rank-1 84.92 # 65
Rank-5 94.21 # 9
mAP 69.14 # 64
Person Re-Identification Market-1501 TriNet (RK) Rank-1 86.67 # 61
Rank-5 93.38 # 10
mAP 81.07 # 52
Person Re-Identification MARS TriNet (RK) mAP 77.43 # 7
Rank-1 81.21 # 8
Rank-5 90.76 # 5
Person Re-Identification MARS LuNet mAP 60.48 # 12
Rank-1 75.56 # 12
Rank-5 89.70 # 6
Person Re-Identification MARS LuNet (RK) mAP 73.68 # 8
Rank-1 78.48 # 10
Rank-5 88.74 # 7
Person Re-Identification MARS TriNet mAP 67.70 # 11
Rank-1 79.80 # 9
Rank-5 91.36 # 4

Methods


No methods listed for this paper. Add relevant methods here