Circle Loss: A Unified Perspective of Pair Similarity Optimization

This paper provides a pair similarity optimization viewpoint on deep feature learning, aiming to maximize the within-class similarity $s_p$ and minimize the between-class similarity $s_n$. We find a majority of loss functions, including the triplet loss and the softmax plus cross-entropy loss, embed $s_n$ and $s_p$ into similarity pairs and seek to reduce $(s_n-s_p)$... (read more)

PDF Abstract CVPR 2020 PDF CVPR 2020 Abstract

Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Metric Learning CARS196 CircleLoss R@1 83.4 # 8
Face Recognition CFP-FP CircleLoss(ours) Accuracy 0.9602 # 2
Metric Learning CUB-200-2011 CircleLoss R@1 66.7 # 1
Face Recognition LFW CircleLoss(ours) Accuracy 0.9973 # 2
Person Re-Identification Market-1501 ResNet50 + CircleLoss(ours) Rank-1 94.2 # 40
MAP 84.9 # 41
Person Re-Identification Market-1501 MGN + CircleLoss(ours) Rank-1 96.1 # 11
MAP 87.4 # 33
Person Re-Identification MSMT17 MGN + CircleLoss(ours) Rank-1 76.9 # 14
mAP 52.1 # 13
Person Re-Identification MSMT17 ResNet50 + CircleLoss(ours) Rank-1 76.3 # 15
mAP 50.2 # 14
Metric Learning Stanford Online Products Circle Loss R@1 78.3 # 4

Methods used in the Paper


METHOD TYPE
Triplet Loss
Loss Functions
Softmax
Output Functions