I would also be willing to do testing!
My position is not optimized, and as discussed up thread, the opportunity to use your technology to suggest positional changes, front hydration placement, etc, would be super valuable, and a welcome “first step” before formal - and more expensive - real world testing.
With this kind of validation on the repeatability of videoing I think I would have seen enough to jump in and get some tests of my own done using your software.
I need to work out how many bottles to put where this season and this seems like a really good way of going about it rather than just guessing and way easier / less expensive than formal testing.
I’m also interested in stack height too but that’s a bit more complicated as there’s an aero-comfort-power tradeoff to consider - that will be a rabbit hole for the winter I think.
That’s the best way to optimize, precisely what I’m doing, as I went from Triathlon to gravel racing and end up solo TTing after getting dropped in a single track option. It’s these tradeoffs and just knowing that can give confidence to your setup and help “cheat” the wind in ways that are specific to the person.
I’m absolutely going to learn what I aggregate from doing these tests to optimize the heck out of my setups 
Oh if I didn’t mention it (it’s been a minute since I’ve looked at this thread), the site is https://wind-tunnel.ai, it’s $30 for now as each test takes a bit over an hour to run (hourish of powerful workstation doing the conversions and simulation, and 20ish minutes of manual effort running it, fixing things in 3D, etc.).
Right now, we’ve only had a handful of users so it’s easy to keep up with the tests, if we start to get a lot more I’ll raise the price until I can train my teammembers to take the workload and get some of the computing on the cloud, just an FYI.
Trust me I want to but it’s very cold and currently snowing, and I don’t have the lighting and such inside, I have reproessed the same mode from different images from the same video (I only use 300ish images from the video, and there’re thousands of frames I could parse out, and, in doing this, the results are nearly identical each time.
Once things start warming up a bit in Wisconsin I’ll have so many more tests to showcase.
I’ve seen enough to have a go - if I can get time this weekend to film my setup then I will.
Looks very cool!
I just saw this, so, thought I should mention some points.
I do a ton of random tests of things of this nature on the Insta (I post a few here): https://www.instagram.com/windtunnel_ai/
For something like different water bottle comparisions, it would be trivial and quite high accuracy.
I’d probably just do it from one video too, with no water bottles, then just shoot me an image of a water bottle (20 second thing to model) and I’d just test a whole bunch of configurations re-using the exact same model of the person/bike so it’s controlled to only the water bottle.
If you did want to use separate videos, the reconstructed model turns out very similar each time, easily testing within 1% of each other, typically even better.
Cool product! I think there would be tremendous demand if users could submit a single video, and you could return the current CdA, as well as several (3-4?) suggested position changes and their predicted CdA. I realize you need to have some constraints, so maybe just looking at elbow pad drop, reach, and forearm angles initially. Something along those lines would easily be worth $100 IMOO.
-Scott (also in Wisconsin)