Created by: littletomatodonkey

1. Combine open image v5 dataset and objects365 dataset to generate a new dataset, based on this we trained some networks: cascade_rcnn_dcn_r101_vd_fpn_generic_server_side and cascade_rcnn_dcn_r50_vd_fpn_generic_server_side
2. One can directly use the two models with 676 categories for inference as we provided the label map.