The goal of this project is to build software that takes in images of NFL plays, and extracts information about the play.
The first phase is focused on pre-snap information, due to the fact that the "pre-snap" state of a NFL play is relatively static and well defined.
Some types of information I am targetting:
- was the offense in shotgun?
- For each of the following offensive positions, how many were on the field, and how were they aligned?
- RB
- TE
- WR
- Was their motion?
- How was the defense aligned?
- Did the defense show blitz?
- Did the defense actually blitz?
On their own these elements may have limited value, but taken together they provide rich, valuable information that isn't widely available.