A camera is positioned so that it's pointing west at the Manhattan skyline. Snapshots are taken every minute and saved to a hard drive.
The images are compiled into composite grid images of what the sky looks like throughout the day, with one snapshot for every 15 minutes from four hours before to one hour before sunset.
Friends and family are invited to rate sunsets from one to five stars, representing their opinion of the visual quality and beauty of each sunset.
They can view the sunset either as an animated loop or grid of snapshots of before and after sunset time.
The croudsourced sunset ratings are averaged out and matched to that day's composite image.
These composite images labeled with a rating are then used to train an image classification model, using TensorFlow and transfer learning to re-train a MobileNet base model.
Every day, the trained model looks at the composite image of the sky for that day and makes a prediction one hour before sunset.
It posts an image with the date, star rating, and confidence level to the @nycsunsetbot Instagram page.