Authors: zhaohanyue, litong, yuanzhibo, daipengwen

Description: 1. Use YOLOv3 as basic framework, adjust parameter and anchor box
2. Training dataset augmentation