I'm currently trying to learn about computer vision stuff and I want to try it on a simple project of real time tracking of figures of players in a foosball game while the camera can be moving a little bit above the table. The first goal I want to achieve is detecting (and tracking) the bars on which the figures are attached (so then I can easily find the figures on them).
My idea was to use Hough transform to find the lines that fit to the bars, but I'm wondering whether it couldn't be done using a neural network. Would it have better performance? If so, how could it be done, what approach can be used here? So I could give images of the table from different angles as inputs to the NN and it'd give me 8 lines that fit to the bars. Using the usual objects detection algorithms probably isn't the right way as this seems to be a simpler problem (finding just bars/lines). If you could just point me to some direction or show me some papers or tutorials that solve a similar problem, I'd be grateful.
Note: To train the NN a was thinking about the mentioned Hough transform to find the lines in the training examples.