I need to do a AI surf project with AI. My first idea was to create a system to take surf photos automatically, like these: Flowstate – The Most Advanced AI Video and Photo Capture Platform for Action Sports
. It's for my AI high school discipline, so the most important here is the AI, not the app. My teacher said to run all on pc instead create an app (I don't have enough time also).
The core idea was: detect the surfers with 1x camera -> if there is a surfer in a wave -> zoom to surfer and take photos sequentially -> back to 1x camera.
But, to do this on pc it's strange to me, because I can't simulate the cellphone zoom, what I can do it's a zoom on the image and not the optical zoom.
The goal of that idea was to be able to have surf photos without need another person to take it. The cellphone would be located at the sand of the beach.
So I changed my idea (because I will run on pc). Now I will process videos, if there is a surfer -> record the video. What this solve? Well, it's like a highlight tool, you can send videos from it to "edit automatically" for the parts that has someone surfing.
Anyway, I want to know if I can do something better. Now, I'm training my model, I have 2 classes "surfer" and "surfer_ridding", the images that I'm using to train is something like these: (really small surfers), I'm using these kinds of image because there isn't a dataset available from cellphone pictures took from sand. And I think it simulate.
I didn't decide if i will use yolo-n, or yolo-m to do so. So, if you have some experience, can you help me? Any advice is grate.