First of all I am very new to deep learning and want to ask for advice.
I am planing to train a model for detecting dart arrows on a dart target.
My idea is to make my own dataset. I would place a stationary camera to face my target and I would simply stick darts on to it on all possible points you could hit from different angles. Imagine a dart target has 20 * 4 + 2 = 82 fields. Two sigles, a double and a triple for each number, bullseye and bull. I thought I could make about 30 images with different arrow angles and different lighting in the room resulting in 82 * 30 = 2460 images. I would also need images with multiple arrows. I thought I could make the same batch of images with two arrows, where I would kind of randomly place the second arrow on a different location as the first one. And I would do the same for 3 arrows, where I would place 2 arrows randomly. I would result in a dataset of 2460 * 3 = 7380 images.
The question is now if you could recommend me something I could do with this dataset. I want that my model can predict how many arrows are on the target and predict for each arrow the corresponding points. As there are 82 fields on the target, that would be my number of classifications for arrows I wish. I don’t know for sure how to accomplish this. Can someone give me an advice or recommend me example projects to this topic or libraries?