I'm using the TensorFlow Object Detection API with a custom dataset in my own repository (following the project structuring recommended on the Training Custom Object Detector guide).
What the best practices for including the tensorflow/models repository (and its dependencies) in my own project? Should I be using a git submodule, a git subtree, cloning the repo and including/not including it in my project repository when pushing to GitHub, etc?
Would like clarification on this. Thanks.