Open LocalJoost opened 3 years ago
Nice article, FYI you could use sqrmagnitude rather than distance to squeeze out some more perf.
Could you explain the values returned by CalculateIndexPinch() and FingerCurl()?
Well - it's both a value between 0 and 1. For curl, 0 in not curled, 1 is as far as it will curl. For pinch, I assume the certainty that a pinch is detected. My threshold values are just empiric - that's a posh word for 'I found this out by trial and error'. Does that answer your question?
Hello, Thanks for this great post, it works perfectly. I wanted to share with you my result.
Thank you for all your contributions
Very glad to hear hear you found my blog useful and thanks for letting me know. Good luck with your further endeavors! 😊
Thank you for this!
@SimonStHilaire you are welcome and thanks. It's nice to hear my blog is not just me shouting into the void ;)
Hi Joost. Thanks a lot for the blog post and it helps me a lot! I am just wondering, do you think it is possible to detect more complex hand gestures, such as a heart shape or a victory shape, with your approach? I know probably there are methods using CV to do so, but I am trying to look for some comparisons of different approaches in terms of speed and accuracy.
Hi @PepperXu , I am not sure, I think you are pretty much on your own here. You might try some algorithmic approach based upon relative position of fingers in a pose, but AFAIK there is nothing out of the box or some code that I am aware of
Hi @LocalJoost I was wondering if you have tried this with MRTK3. Thanks.
Not yet.
Original article: https://localjoost.github.io/Basic-hand-gesture-recognition-with-MRTK-on-HoloLens-2/