Web video golf club tracking / analysis application

I have been struggling getting a project developed to the extent that the initial twenty-year patent period is now down to about four years and I still do not have it successfully developed. It does not seem to be a particularly exasperating project, in part requiring a pre-swing period of a golfer (before a swing even begins) to be analyzed and an existing rotation point location along the length of a club determined during a unique club motion traditionally termed a waggle. The point location can vary notably from golfer to golfer. For about the first decade I tried almost exclusively using an inertial-type sensor(s). It was eventually abandoned primarily due to data drifting errors over time (presumably still existent) when trying to determine an object's location in space. A pre-swing period could last as much as 10-15 seconds or more and contain numerous individual waggles throughout. Comparing them can be valuable, though a single waggle is of a much shorter duration and all that is foundationally needed at a most basic level. So a few years ago I switched to trying a video-based tracking system, and while some progress has been made it is not functionally workable yet. Video has its downsides for sure, but also has some great upsides that inertial sensors cannot touch. Golf club tracking, display, and calculations using computer vision really only need to be performed in 2D when analyzed along the proper plane of view. I went right to attempting a web-based application, maybe out of desperation to try to salvage a little more from the project before it is totally too late. And perhaps this has complicated and slowed things a bit over starting with just a desktop application first. But it has been started like that, so hopefully it can be successfully completed well in that manner, with any comments appreciated. The last couple of contractors have really been terrific, yet neither was able to achieve workable success to this point. The first was highly specialized in the R/Shiny platform or language. His actual tracking system was really nice and pretty effective, and some of his application graphics were great. Among other issues, however, video frame handling was abnormally slow and ineffective on multiple levels. And the platform seemed unable to produce certain functions, such as an ability to effectively manually replace tracking markers too inaccurately placed by the auto-tracking system before performing further calculations. With that experience I searched out and worked with another fine contractor who worked in a version or derivative of C (maybe C++) at least for the base application. And indeed, he was able to correct and/or improve just about all of the features that R/Shiny struggled with. But unfortunately, what appears to be the best feature of R, the actual tracking system, appears to be the worst feature in the C++ version. It has been far too inaccurate and inconsistent to the point where most every tracking point placed by the auto-tracking system needs to be manually replaced, a very tedious process. As one of the core functions of the application, this is the biggest (though not the sole) reason that version cannot be deployed either. Someone well versed in both languages or platforms that might be able to analyze them and correct/improve what is still needed (and that I can afford with the scant funds I have remaining) would seem ideal at this point to hopefully finally make the project a success. Many different additional details are of course available for any interested and qualified parties. Somewhat ironically, studying what has been accomplished with the video attempts thus far reveals that inertial-type sensor use might be quite successful and once again be considered instead, but at this point I feel it is more important to complete a video version of the application successfully first. Project or hourly quotes can be considered. Thank you very much.