Performance question

[gasparuff]

Hello,

I'm running this project on a Raspberry Pi4 with 4GB of RAM and it works actually quite well. I'm testing with a bunch of photos I took many years ago and the detection works. Although it takes between 5-10 seconds for detecting the license plate and returning the result. Although sometimes the RPi just "dies" when I execute the detection too many times, I guess the hardware is just reaching it's limits (overheating?).

Is there anything I can do to improve this? I'm thinking in the direction of sacrificing quality for speed. For the particular case I want to solve, I will always get just one photo in very good quality and the license plate will always be approximately in the same region of the photo.

What are my options?

PS: Thanks to all of you for putting together such an awesome project, that allows me to run ANPR without almost any knowledge about ML. Although I'm digging into the whole topic to get a better understanding of everything, it will take me some more time until I get the full picture :).