more object detection models

[romanoss]

As it seems turicreate is lacking more models and a way to train models with more hw support. Especially I would like to see object detection models like yolov3, yolov3 tiny et al with support of documented eGPUs (on a mac), otherwise, this platform will make less and less sense for developing. What's the problem? It does not seem to be that kind of effort. Furthermore, it really makes sense to expose more options for augmentation.

regards

[srikris]

@romanoss This is a great feature request and one that we are actively exploring. If you or any others have any other requests, please add it here and stay tuned!

[jegesh]

I would also like to see model architectures, not just pre-trained models that can be refined