PoseBox-Reid

A two-stream Re-id CNN network

Paper: https://arxiv.org/abs/1701.07732

Components:

1. Models:

   The folder holds all trained models

2. ExtractFeatures:

   Shows how to extract features using our model

3. Fusion:

   Shows how to merge two pertrain Caffe Models into one two-stream network

4. DataMaker:

   Shows how to generate poseBox data based on pose estimation results and how to prepare hdf5 data for our model training

TODO:

finish how-to-training part