Human Parsing with Discriminant Feature Learning for Person Re-identification

It is urgent key research for person re-identification (ReID) to extract discriminant distinguishing features from pedestrian pictures, optimize the feature extraction ability of the network, and obtain better pedestrian feature representation. By combining the global and detailed features of pedestrians, based on the multi-level attention map generated by human parsing model and soft target obtained by self- teaching way, we design a multi-branch deep network model to distinguish pedestrian key features from different levels for ReID, in particular, each branch is guided by different information to jointly learning complementary global and local feature representation, then a feature selection module used to emphasize important features to further improve discriminative power. Through experimenting on datasets, we verify that human parsing with discriminant feature learning contributes to the performance of ReID. Compared with current state-of-the-art methods, our method achieves promising results.