Reduce manual intervention

[dmtroyer]

Currently there are some steps of manual intervention specific to the video that need to go away so this will scale.

• Restrict cameras used to one or two models to ensure focal length and perspective
  • Matt used an iPhone 5 for the original video and we have that mapping
  • Consider validating other phones
  • this allows us not to have to worry about the coordinate system mapping
• Reduce fiddling with threshold
  • Consider looking at differences between frames or dynamically
• Rotation and Crop
  • Currently these are manually determined values
  • Automate this so that the rotation of the physical camera is not an issue
    • Anchor based on features of the video landscape
• YouTube
  • API to give the script a video file that we want to process and possibly upload the resultant annotated video.