Created by: TheodoreG

I used official docker image paddle:1.4.1-gpu-cuda9.0-cudnn7 to do cpu infer. The model is official Mask RCNN model. Is there any solution to do some optimization?