Thanks for your reply, I did what you advice me, thanks, now I have other problems,

I can not run train.py and eval.py at the same time, since no enough memory, although I hv 8 GB GPU memory.

I converted the training file to .pb file, but it can not detect anything when I tested with object_detection_tutorial.ipynb

To make it easy for u, please have a look here https://stackoverflow.com/questions/45802672/train-tensorflow-with-my-own-images-successfully-but-still-have-problems

)