Ask HN: How to Succeed in Computer Vision

Hello! I'm taking a course in Computer Vision, and I was looking for guidance on generally how to approach solving CV problems. I am pretty fresh in studies and application, but at high level I feel I am developing the engineering and art to establishing an image processing pipeline so that, let's say, you can develop the code to perform Augmented Reality marker tracking in an image. Once the pipeline is created, one can begin tweaking. I want to understand how best to tweak and measure performance. For example, OpenCV has trackbars that can be used in the image viewing UI for twiddling parameters and seeing how changes affect the processing of the image. I can begin examining histograms of pixel intensities to understand how to filter, but there are many other variables to examine, and I'm not experienced enough to know what to keep in my toolbox and what to throw out or replace, or where to buy new tools.

I am looking for advice and new perspective and ways to think when working in this field.

Thank you and looking forward to learning from the community!