Open RusseII opened 3 years ago
Yes but I stopped exploring this avenue cuz @gatesyp indicated it is out of our current budget? I'd be happy investigate again though.
@gatesyp indicated it is out of our current budget.
What if we took the video, converted it into a really low frame rate (like 2-3 frames per second) and used an ML model to compare the static images? Could that bring the cost down?
For example, when we see the below image we know that a game just ended. There is almost always a good clip right before this event.
This is outside of scope for the current implementation, but I wanted to note this possible extension down.