Why dont you train end-to-end?

[TheSunWillRise]

It seems that the codes in this project are just parts of the total model. You extract features for gallery and probe, and load them as input data for the training of GCN model in this project, right?
So, why dont you train the model end-to-end?

[sjtuzq]

In fact, our entire framework can be end-to-end, but because the code of the whole project is very large, we haven't put them here. The code we release is only the part of the convolution of the graph, which can be embedded in any other person search frameworks. It is very simple and concise.