If you want to reuse the existing Yolo 2 network and weights without retraining, another option that I just got working, but is a bit involved, is roughly:

Use the existing YAD2K steps to convert the yolo cfg and weights into an h5 file. Clone the coremltools repo and follow the steps to build from source. Add the coremltools repo top-level directory to your PYTHONPATH. Run python 3 interpreter and import coremltool to check that it’s working. Run your coremltools keras to coreml convert script See the coremltools fail in some file with an error about Lambda’s not being supported Hack the coremltools file that failed by tracing it’s use of conversion methods, and find the correct place to use coreml’s add_reorganize_data() method. That last part is the tricky part and will take some experimenting and re-running your coreml conversion script.

https://apple.github.io/coremltools/generated/coremltools.models.neural_network.html