First trial feedback

[jaidevd]

Given the following issues, continuing with the whole README is somewhat time consuming, so I'm leaving it here until these are fixed.

Issues

• It's not clear whether github.com/srm-soumya/planet-amazon is to be used as a template,
  or if it is a fully-functioning project by itself.
• In the README, it's not clear where the predict.py file should be located.
• This file should raise NotImplementedErrors initially.
• It's not clear that a Torch model has to be loaded - if this works with Keras or native TF models too,
  please mention that in the README.
• Add link to a sample model in the README
• No clue as to what the three functions in predict.py should be.
  • What are the input arguments? Serialized model paths / model instances / tensors / dataset instances, what?
  • What are the functions supposed to return?
• The functions in predict.py need parsed command line arguments to work - this is not mentioned in the README.

Good to have

• Maybe add an option for GPU in conda installations
• Use cookiecutter for creating project structure, for both torch and Flask related code.