Video-Related Golf Analysis Application Needed

I have a partially developed, golf-related analysis application that is currently written in the R and Shiny languages. While the present intent is to get this application up and running as a web application as soon as humanly possible, a desktop version will likely be needed also. So if starting with a desktop version is deemed to be more advantageous for the project as a whole, I can consider that. The successful application does not have to be in R. In fact, the application currently runs extremely slowly, appears to have other issues in that programming, and might have to be rewritten anyhow even if initially successfully formulated in R. (There are some nice visuals possible within the application, though, and I do not know if they can be comparatively duplicated in other languages). Someone who can at least skillfully decipher and transpose these languages efficiently would seemingly be helpful. Personal users start by recording videos of their golfing pre-swing motions (which could last as much as 15 seconds or more for some golfers). A selected video is uploaded (if web based) to the application. The original video format (which could vary) is reformatted (to MP4 I think) by the application, and frame numbers are added (time code could be used instead if more efficient). The reformatted video is next analyzed essentially frame by frame. (This can currently be done somewhat decently in certain browsers but not others [not sure if this is a deficiency of R], so an option exists for downloading the reformatted video for analysis in a local video player). One or more shorter segments of the video are selected to be further analyzed. The determined segment(s) is then input into an analysis section of the application and specified visual and mathematical parameters are output. What needs to be accurately tracked and analyzed is the pre-swing golf club movement (before an actual swing even begins). This has been done thus far in the application by a color-tracking computer vision program, with the program tracking two colored markers placed at known positions on a club before recording a video. An auto-tracking feature by the computer vision program is needed, yet it has never yet been 100% correct in placing tracking markers when reviewing its accuracy, even for video segments comprising as few as ten frames. (A feature for reviewing marker placements is included and needed in the application). But a strictly manual marker placement is problematical also, as even very short (time-wise) video segments can have high frame rates and be much too cumbersome for marking all frames manually. So some type of "hybrid" tracking system would seem to be in order at this time, consisting of an auto-tracking system coupled with a feature for efficiently reviewing and manually altering marker placements as needed. The video tracking aspect has not been satisfactorily achieved yet, thus data from the analysis section of the application has little value to this point. I do not know if the R language and/or computer vision aspects as examples are not well suited to the video analysis needs of the application. I am told that something like Artificial Intelligence or machine learning video tracking might work faster in R, and I am open to that if it can be accomplished at a comparable time and cost of color tracking. But I have heard that such methods might be even less accurate than color-tracking. Plus a critical issue(s) with the video analysis aspect goes beyond, and appears unrelated to, just its speed of operation. This project is already beyond the point of desperation in its development. It needs to be successfully completed as soon as possible (and at as little additional cost as possible at least until proven that it can operate successfully). I please need help to successfully complete the development of the application and deploy it online. Hourly or project rates can be considered. Thank you very much.