Center Prediction Loss for Re-identification

Abstract

The training loss function that enforces certain training sample distribution patterns plays a critical role in building a re-identification (ReID) system. Besides the basic requirement of discrimination, i.e., the features corresponding to different identities should not be mixed, additional intra-class distribution constraints, such as features from the same identities should be close to their centers, have been adopted to construct losses. Despite the advances of various new loss functions, it is still challenging to strike the balance between the need of reducing the intra-class variation and allowing certain distribution freedom. Traditional intra-class losses try to shrink samples of the same class into one point in the feature space and may easily drop their intra-class similarity structure. In this paper, we propose a new loss based on center predictivity, that is, a sample must be positioned in a location of the feature space such that from it we can roughly predict the location of the center of same-class samples. The prediction error is then regarded as a loss called Center Prediction Loss (CPL). Unlike most existing metric learning loss functions, CPL involves learnable parameters, i.e., the center predictor, which brings a remarkable change in the properties of the loss. In particular, it allows higher freedom in intra-class distributions. And the parameters in CPL will be discarded after training. Extensive experiments on various real-world ReID datasets show that the proposed loss can achieve superior performance and can also be complementary to existing losses.