Document steps for PowerAI and Watson Visual Recognition

[krook]

Is your feature request related to a problem? Please describe. Beyond the current open source TensorFlow based implementation, we need to add steps for alternative visual recognition systems, including PowerAI and Watson Visual Recognition.

Describe the solution you'd like DroneAid in general needs to have a flexible architecture that allows different components to implement its 4 core features; 1. Several device type inputs, 2. Several training systems, 3. Several realtime processing systems, and 4. Several downstream reporting/messaging outputs.

Additional context The original system was based on PowerAI. We need to recover those steps if possible.