The origin of this step on pit:
For target detection own training set transfer learning when I write this.
Internet to find Tensorflow object detection API-related training methods (article should be made from May 2018, there is time to reprint misleading rearward, but after a look at the source code to understand).
Unfortunately, the author on the environment to build on the cards for two days.
Tensorflow object detection API's github address is: https://github.com/tensorflow/models/
when I downloaded was in October 2019.
Error cases:
by the next pass environment configuration python,
python object_detection / Builders / model_builder_test.py
has run successfully.
To identify cases of puppies and beaches such as the Master of object_detection_tutorial.ipynb
We look forward to painting the picture to see the tutorial are familiar with, but just can not get out.
I was greeted "service seems to hang, but will immediately restart." Then've been stuck
re interrupt service, and then run or stuck in here.
By observing object_detection_tutorial.ipynb code, which starts on May 18 and 19 October has been quite different.
It seems only chance back in 2018 code. 
From 18 May to move forward to find an Verified label, and as the beginning of the code section. This is submitted.
Then download back on and restart jupyter object_detection_tutorial.ipynb run the same directory.
Familiar with the dog, beach map finally appeared.
Conclusion:
On the Tensorflow object detection API environment to build, if this tutorial did not run through, most likely due to other environmental variables and different linux generated.
We will continue down the train. . . New progress will inform