I have an interest in AI, specifically how it relates to aviation. Applications I was thinking of are:
-Flight data extraction from instruments (speed, altitude, heading, pitch, roll, etc. to be used for simulator reference and parameter warnings).
-Aircraft position extraction from navigation display (avoid loss of communication due to position report missed).
-Fuel leak check (compare fuel on board with fuel used and fuel on arrival).
The data extraction centers mostly around text extraction, but this is not always self evident such as with altitude and speed tapes. Pitch and roll extraction from the artificial horizon is an edge case (no pun intended). Example of A320 instruments (video not mine):
Unfortunately I am a total noob regarding AI. I can program in several languages but have only basic math skills. I have looked up how to get started with AI but I don’t have the time to delve into details. I came across one research paper dealing with this sub field but it is way over my head.
So what I am offering is to create a massive database with cockpit pictures and videos in various conditions (day, night, flight phases), specifically from the instruments although I can shoot other angles on request. The fact that I am the pilot will help. Someone who is a bit more AI-savvy can take the training database to teach the AI to recognize certain instruments and extract data from them. The data will be public and free.
Even is no one is interested, I would like to know the following:
-Is video required or are photos just fine. If video is useful, what frame rate and what resolution?
-What resolution photos are best?
-Are wide angle cameras ok? GoPro is the most practical as it can be mounted easily.
-Are blur free images (mechanically stabilized) important?
-Raw or JPG?
Due to database size and processing power requirements, the lower resolution/frame rate, the better.
Let me know if you can help.